Bat repelling and bird repelling method and device for multi-mode wind turbine generator

Abstract

The invention relates to the technical field of wind power generation, in particular to a bat repelling and bird repelling method and device for a multi-mode wind turbine generator. Comprising a shell, and a monitoring cabin, a repelling cabin and a control cabin are formed in the shell through partition plates; a millimeter-level radar is arranged at the top of the shell, a dual-spectrum camera is arranged on the front side of the shell, and the millimeter-level radar and the dual-spectrum camera correspond to the position of the monitoring cabin; a first module, a second module and an AI identification module are arranged in the monitoring cabin; a loudspeaker is installed on the front side of the shell, a stroboscope lamp and a laser spotlight are installed on the two sides of the shell, the loudspeaker, the stroboscope lamp and the laser spotlight are arranged corresponding to the positions of the repelling cabin, an ultrasonic transmitter is installed in the repelling cabin, and a multi-mode repelling system is arranged in the repelling cabin; photovoltaic panels are fixedly mounted at two ends of the top of the shell; a power supply module, a hollow module and a communication module are arranged in the control cabin; and efficient, accurate and intelligent repelling of birds and bats is realized.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a method and device for repelling bats and birds in a multi-mode wind turbine. Background Technology

[0002] With the booming development of the global wind power industry and the continuous expansion of power grids, collisions and nesting activities between birds and bats and wind turbines and transmission lines are increasing, posing a serious threat to bird and bat populations, the safe operation of wind farms, and the safe and stable operation of power production. Traditional methods of bird and bat repellency, such as visual disturbances (ribbons, reflective strips, etc.), sound repellency (bird deterrents, firecrackers, etc.), and physical barriers, have problems such as limited coverage, unstable repellency effects, potential negative environmental impacts, and difficulty in effectively targeting specific species like bats. These methods often lack specificity and are difficult to effectively address the differences in habits among different bird and bat species and the complex and variable weather conditions, especially in the unique environment of towering wind turbines, where their limitations are even more pronounced.

[0003] While existing technologies attempt to utilize radar monitoring, infrared imaging, and acoustic monitoring to assist in bird and bat control, these systems typically rely on single sensors or bird control patterns, lacking intelligent and automated functions. They struggle to respond in real-time to the varying habits of different birds and bats and complex environments. Furthermore, the nesting and roosting of birds and bats on power transmission line towers can directly threaten the safe operation of power production, causing significant economic losses and social impact. Existing bird and bat control devices often lack the ability to dynamically predict and accurately locate bird and bat behavior, and are also unsuitable for adapting to different environmental conditions (such as wind speed, wind direction, and light intensity). Simultaneously, the lack of a continuous evaluation and optimization mechanism for bird and bat control effectiveness makes it difficult to adjust strategies based on actual conditions, hindering the improvement of bird and bat control results.

[0004] Therefore, there is an urgent need to provide a multi-mode wind turbine bat and bird avoidance method and device, which can achieve efficient, accurate and intelligent avoidance of birds and bats compared with existing technologies. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a method and device for bat and bird avoidance in multi-mode wind turbines.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A multi-mode wind turbine bat and bird avoidance device includes an outer shell, inside which a monitoring chamber, a avoidance chamber, and a control chamber are formed by partitions; a millimeter-level radar is installed on the top of the outer shell, and a dual-spectrum camera is installed on the front of the outer shell, the millimeter-level radar and the dual-spectrum camera corresponding to the position of the monitoring chamber; the monitoring chamber contains a first module, a second module, and an AI recognition module; a speaker is installed on the front of the outer shell, and strobe lights and laser spotlights are installed on both sides of the outer shell, the speaker, the strobe lights, and the laser spotlights corresponding to the position of the avoidance chamber, an ultrasonic transmitter is installed in the avoidance chamber, and a multi-mode avoidance system is installed in the avoidance chamber; photovoltaic panels are fixedly installed at both ends of the top of the outer shell; the control chamber contains a power supply module, a hollow module, and a communication module.

[0007] Furthermore, the AI ​​recognition module includes a wingspan recognition module, a flight attitude recognition module, a thermal imaging feature recognition module, and an acoustic feature analysis module. The wingspan recognition module, the thermal imaging feature recognition module, and the acoustic feature analysis module are used to sequentially identify the category of the organism, and the flight attitude recognition module is used to identify the flight patterns and movement modes of the identified organism.

[0008] Furthermore, the multi-mode avoidance system includes an optical avoidance module, an acoustic avoidance module, and an ultrasonic avoidance module. The speaker is electrically connected to the acoustic avoidance module, the strobe light and the laser spotlight are electrically connected to the optical avoidance module, and the ultrasonic generator is electrically connected to the ultrasonic avoidance module.

[0009] Furthermore, the power supply module includes an MPPT controller, which connects the tilted photovoltaic panel and the lithium iron phosphate battery pack via a ring-shaped copper busbar; the communication module adopts a hierarchical architecture, and real-time control commands are transmitted through a time-sensitive network.

[0010] A method for repelling bats and birds in a multi-mode wind turbine, using a multi-mode wind turbine bat and bird repelling device, includes the following steps: S1. Data is collected through millimeter-level radar and dual-spectrum camera. The data collected by millimeter-level radar includes flight speed, acceleration, wingbeat frequency, and formation. The data collected by dual-spectrum camera includes infrared thermal imaging data. S2. The wingspan recognition module extracts the contour features of biological targets based on the point cloud data acquired by millimeter-wave radar and a target detection algorithm based on deep learning, and performs a preliminary classification of the biological categories. S3. The thermal imaging feature recognition module extracts the surface temperature distribution characteristics of organisms based on infrared thermal imaging data, and further calibrates the biological category identified in step S2. S4, the acoustic feature analysis module, collects echolocation pulses through a high-frequency microphone array, extracts frequency domain features (including center frequency, bandwidth, and pulse interval) through fast Fourier transform, and classifies the organisms identified in step S3 using a support vector machine classifier to obtain the final classification result. S5. The flight attitude recognition module extracts information on flight speed, acceleration, wingbeat frequency, and formation, obtaining the flight patterns and movement patterns of the organisms identified in step S4; including the following steps: S51. Obtain the flight patterns of the organisms identified in step S4, including the target trajectory prediction, the estimated wingbeat frequency, and the formation quantification. S52. Input the extracted flight speed, acceleration, and wingbeat frequency into the long short-term memory network to identify motion patterns; S6. Based on the biological classification obtained in step S4 and the flight patterns and movement patterns of each biological obtained in step S5, match the optimal avoidance scheme in the mapping database. S7. Generate corresponding control commands through the central control module and send them to each module to execute the optimal avoidance scheme obtained in step S5.

[0011] Furthermore, the target trajectory prediction value in step S51 is calculated using the following formula: ; in, Indicates time k before, based only on The predicted values ​​of the state based on the data at and before time are the state vector (including position and velocity). Indicates time Previously, based solely on The predicted value of the state based on data from the current time and earlier; Here is the state transition matrix. , For controlling input.

[0012] Furthermore, the estimated value of the wingbeat frequency in step S51 is calculated using the following formula: ; In the above formula, The Fourier transform result of signal s(t) represents the amplitude and phase of the signal at frequency f. Its absolute value is the spectrum, and the frequency corresponding to the peak frequency is the estimated value of the flapping frequency; f represents frequency, t represents time, and j represents the imaginary unit.

[0013] Furthermore, the method for obtaining the quantization value of the formation pattern in step S51 is as follows: a density-based spatial clustering algorithm (such as DBSCAN) is used to segment the multi-target point cloud at the same time, and the geometric center C of the group and the average distance between individuals are calculated. And the formation's main axis direction angle θ, thereby quantifying the formation pattern and obtaining the quantified value of the formation pattern.

[0014] Furthermore, in step S52, LSTM units in the Long Short-Term Memory network are used to perform calculations at each time step t. Finally, the hidden state at the end of the network is concatenated with the quantized value of the formation pattern and fed into a multilayer perceptron classifier for motion pattern recognition, represented as: ; In the above formula, The output represents the probability distribution of various movement patterns, including migration, predation, hovering, and attack. This represents the weight matrix in the classifier. This represents the bias term in the classifier. Indicates the hidden state at the end of the network. The quantified value representing the formation pattern.

[0015] Furthermore, in step S6, acoustic and optical repellency are used for gregarious birds; ultrasonic interference is used as the main method and ultraviolet pulses as a supplementary method for nocturnal bats; and a graded and enhanced repellency mode is used for the five solitary bird species.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention combines an intelligent monitoring system and a repellency system to achieve real-time monitoring and effective repellency of flying objects around wind farms, thereby improving the operational safety and reliability of wind farms. The intelligent monitoring system can comprehensively monitor the flying objects within the sweep range of wind turbine blades and identify bats and birds through an AI recognition module, providing accurate target information for the repellency system. Through the built-in AI recognition module, it makes precise responses based on the different species, behaviors, and environments of bats and birds, and implements targeted repellency measures, making up for the limitations of traditional single repellency methods and avoiding harm to certain specific species of birds or bats. The repellency system plays the calls of birds and bat predators through a loudspeaker to acoustically repel flying objects, while using laser spotlights and ultrasonic transmitters for optical and ultrasonic repellency, forming a multi-mode, multi-level repellency method, effectively avoiding the limitations that may exist with a single repellency method.

[0017] (2) The power supply system of this invention adopts a combination of solar panels and batteries, realizing the utilization of green energy and reducing the operating cost and environmental impact of the device. The solar panels supply power to the batteries, ensuring the stable operation of the device under various weather conditions.

[0018] (3) The control system of the present invention realizes centralized control of intelligent monitoring system, avoidance system, power supply system, etc. through microcontroller, and realizes remote monitoring and data analysis through connection with cloud platform, which facilitates remote management and optimization of device.

[0019] (4) The self-learning algorithm module of the present invention can adjust and optimize the avoidance strategy in real time according to the effect of the avoidance action, thereby improving the adaptability and effectiveness of the avoidance effect. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0021] Figure 2 This is a schematic diagram showing the internal structure of the present invention.

[0022] Explanation of reference numerals in the attached figures: 1. Outer shell; 101. Monitoring compartment; 102. Avoidance compartment; 103. Control compartment; 2. Photovoltaic panel mounting bracket; 3. Photovoltaic panel; 4. Millimeter-wave radar; 5. Dual-spectrum camera; 6. Strobe light; 7. Laser spotlight; 8. Speaker; 9. Ultrasonic transmitter; 10. Central control module; 11. Power supply module; 12. Communication module; 13. First module; 14. Second module; 15. AI recognition module; 16. Optical avoidance module; 17. Acoustic avoidance module; 18. Ultrasonic avoidance module. Detailed Implementation

[0023] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention. It should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0024] like Figure 1 , Figure 2As shown, this invention provides a multi-mode wind turbine bat and bird avoidance device, including a housing 1 and a monitoring compartment 101, a avoidance compartment 102, and a control compartment 103 integrated inside the housing 1. Multiple compartments are formed inside the housing 1 by partitions, with the compartments from top to bottom being the monitoring compartment 101, the avoidance compartment 102, and the control compartment 103. Adjacent compartments are electrically connected via waterproof aviation connectors and shielded cables. The housing 1 includes a top cover and a base, which are detachably connected. The housing 1 is made of aluminum alloy, and its surface is treated with anti-corrosion coating.

[0025] A millimeter-wave radar 4 is mounted on the top of the outer shell 1, and a dual-spectrum camera 5 is installed on the front of the outer shell 1. The dual-spectrum camera 5 supports 360° horizontal and ±90° pitch angle adjustment. The millimeter-wave radar and the dual-spectrum camera 5 are positioned corresponding to the monitoring cabin 101. The monitoring cabin 101 contains a first module 13, a second module 14, and an AI recognition module 15. The first module 13 is electrically connected to the millimeter-wave radar, and the second module 14 is electrically connected to the dual-spectrum camera 5. The first module 13 acquires data collected by the millimeter-wave radar and transmits the acquired data to the AI ​​recognition module 15. The second module 14 acquires data collected by the dual-spectrum camera 5 and transmits the acquired data to the AI ​​recognition module 15. The AI ​​recognition module 15 identifies the species of organism based on the data transmitted by the first module 13 and the second module 14, formulates an avoidance plan based on the identified species, and transmits it to the avoidance module. The AI ​​recognition module 15 includes a wingspan recognition module, a flight attitude recognition module, a thermal imaging feature recognition module, and an acoustic feature analysis module. The wingspan recognition module is used to perform preliminary classification of organisms; the flight attitude recognition module is used to identify the flight patterns and movement modes of organisms; the thermal imaging feature recognition module is used to further distinguish the categories of organisms based on the preliminary classification obtained by the wingspan recognition module, thereby enhancing the species differentiation capability; and the acoustic feature analysis module is used to perform species matching on organisms with acoustic features identified by the thermal imaging feature recognition module.

[0026] A speaker 8 is installed on the front of the outer casing 1, and a strobe light 6 and a laser spotlight 7 are installed on the sides of the outer casing 1, respectively. The speaker 8, strobe light 6, and laser spotlight 7 are all positioned corresponding to the avoidance compartment 102. Ultrasonic transmitters 9 are installed at the four corners of the avoidance compartment 102. The avoidance compartment 102 is equipped with a multi-mode avoidance system, including an optical avoidance module 16, an acoustic avoidance module 17, and an ultrasonic avoidance module 18. The speaker 8 is electrically connected to the acoustic avoidance module 17, the strobe light 6 and the laser spotlight 7 are each electrically connected to one optical avoidance module 16, and the ultrasonic transmitter 9 is electrically connected to the ultrasonic avoidance module 18. The strobe light 6 and the laser spotlight 7 are controlled according to the instructions of the optical avoidance module 16; the speaker 8 is controlled according to the instructions of the acoustic avoidance module 17; and the ultrasonic transmitter 9 is controlled according to the instructions of the ultrasonic avoidance module 18.

[0027] The top of the outer shell 1 is fixedly connected to photovoltaic panel 3 mounting brackets 2 at both ends. Each photovoltaic panel 3 mounting bracket 2 has a photovoltaic panel 3 installed on its upper end, and the two photovoltaic panels 3 are arranged in a figure-eight shape. The control cabin 103 is equipped with a power supply module 11, a central control module 10, and a communication module 12. The power supply module 11 includes an MPPT controller, which connects the tilted photovoltaic panels 3 and the lithium iron phosphate battery pack through a ring copper busbar, supporting priority switching and load balancing. The power supply for different cabins adopts a star topology. The communication module 12 adopts a hierarchical architecture. Real-time control commands are transmitted through a time-sensitive network (TSN) to ensure millisecond-level response of laser steering and ultrasonic frequency control. Sensing data is transmitted to the cloud to train the model through a 5G uplink, and algorithm update packets are broadcast to the local device group via LoRa. The control cabin 103 is equipped with dual redundant communication channels. When the main channel fails, it can automatically switch to the backup link to ensure continuous system operation.

[0028] This invention also provides a method for bat and bird avoidance in multi-mode wind turbines, comprising the following steps: S1. Data is collected through millimeter-level radar and dual-spectrum cameras. The data collected by the millimeter-level radar includes flight speed, acceleration, wingbeat frequency, and formation, while the data collected by the dual-spectrum cameras includes infrared thermal imaging data.

[0029] S2. The wingspan recognition module extracts the contour features of biological targets based on the point cloud data acquired by millimeter-wave radar and a target detection algorithm based on deep learning, and performs a preliminary classification of the biological species.

[0030] Point cloud data is the set of coordinates in three-dimensional space for each detected target point (such as a bird or bat) within the radar scanning area. Each data point contains information such as the target's distance, azimuth, elevation angle, and velocity (measured through the Doppler effect). The set of these points constitutes the target's contour features and trajectory.

[0031] S3. The thermal imaging feature recognition module extracts the surface temperature distribution features of organisms based on infrared thermal imaging data, such as average body temperature and temperature gradient. Combined with the ambient temperature, it further calibrates the biological category identified in step S2.

[0032] S4, the acoustic feature analysis module, collects echolocation pulses through a high-frequency microphone array, extracts frequency domain features (including center frequency, bandwidth, and pulse interval) through a fast Fourier transform, and classifies the organisms identified in step S3 using a support vector machine classifier to form the final classification result.

[0033] S5. The flight attitude recognition module extracts information on flight speed, acceleration, wingbeat frequency, and formation pattern to obtain the flight patterns and movement patterns of the organisms identified in step S4. The flight patterns include the predicted target trajectory, the estimated wingbeat frequency, and the formation pattern. Specifically, it includes the following steps: S51. Dynamic Feature Extraction: Millimeter-wave radar acquires point cloud data of the target using high-frequency sampling (e.g., 100Hz). For each tracked target, its dynamic features over time are calculated. Velocity and Acceleration: Radial velocity vr is directly measured based on Doppler frequency shift. The target trajectory is smoothed and predicted using Kalman filtering to obtain the predicted target trajectory value. The process model can be expressed as follows: ; in, Indicates time k before, based only on The predicted values ​​of the state based on the data at and before time are the state vector (including position and velocity). Indicates time Previously, based solely on The predicted value of the state based on data from the current time and earlier; Here is the state transition matrix. , For controlling input.

[0034] By analyzing the periodic fluctuations in the wingtip echo intensity or the temporal variation of the target's radar cross-section (RCS), the dominant frequency is extracted. A Fourier transform is then performed on the dominant frequency time-series signal s(t) to obtain an estimate of the wingbeat frequency. ; In the above formula, The Fourier transform result of signal s(t) represents the amplitude and phase of the signal at frequency f. Its absolute value is the spectrum, and the frequency corresponding to the peak frequency is the estimated value of the flapping frequency; f represents frequency, t represents time, and j represents the imaginary unit.

[0035] Density-based spatial clustering algorithms (such as DBSCAN) are used to segment multi-target point clouds at the same time, and the geometric center C of the group and the average distance between individuals are calculated. The formation shape is quantified by the direction angle θ of the main axis of the formation, and the quantified value of the formation shape is obtained. The formation shape includes V-shape, linear or disordered.

[0036] S52. Temporal Pattern Recognition and Classification: The extracted feature sequences are input into a Long Short-Term Memory (LSTM) network to capture their temporal dependencies and identify motion patterns. The LSTM unit is computed at each time step t as follows: ; ; ; ; ; ; In the above formula, The input feature vector at time t represents velocity, acceleration, vibration frequency, etc. , , These are the forget gate, input gate, and output gate, generated by the sigmoid function, with values ​​in the range [0,1]. They respectively control how much of the old cell state of the current input information needs to be forgotten, how much needs to be stored in the cell state, and how much of the current cell state needs to be output to the hidden state. , , , These represent the weight matrices for the forget gate, input gate, output gate, and candidate cell states, respectively. These are parameters that need to be learned during model training. express The hidden state at all times , , , These represent the bias vectors for the forget gate, input gate, output gate, and candidate cell state, respectively. This represents the candidate cell state at time t. This represents the cell state at time t. This represents the hidden state at time t; This represents the sigmoid activation function. This indicates element-wise multiplication.

[0037] Finally, the hidden state at the network end is concatenated with the quantized value of the formation pattern and fed into a multilayer perceptron (MLP) classifier for motion pattern recognition, represented as: ; In the above formula, The output represents the probability distribution of various movement patterns, including migration, predation, hovering, and attack. This represents the weight matrix in the classifier. This represents the bias term in the classifier. Indicates the hidden state at the end of the network. The quantified value representing the formation pattern.

[0038] S6. Based on the biological classification obtained in step S5 and the flight patterns and movement modes of each biological obtained in step S5, match the optimal avoidance scheme in the mapping database and execute the optimal avoidance scheme through the central control module.

[0039] For gregarious birds (such as black-backed gulls): Acoustic-optical combined avoidance is prioritized. Given their susceptibility to sound interference during flock flight, predator calls (such as peregrine falcon attack calls) are played to induce panic-driven turning in the flock. Simultaneously, dynamic light spot projection creates visual deterrence, leveraging the herd mentality to amplify the avoidance effect. When millimeter-wave radar detects a flock of black-backed gulls (15-20 birds, flight speed 12 m / s) flying in a V-formation at a distance of 200 meters, the AI ​​identification module confirms the species type through wingspan characteristics (1.2-1.5 meters) and flight rhythm analysis. Simultaneously, infrared thermal imaging data (body temperature 41.2±0.5℃) from a dual-spectrum camera is used to verify the target's attributes. Based on a group flight hydrodynamic model, the central control module predicted that the flock would enter the blade sweep zone in 18 seconds and immediately activated a coordinated acoustic and optical avoidance strategy: the loudspeaker played the peregrine falcon attack call stored in the database (frequency concentrated in 3-5kHz, sound pressure level 115dB@50m), while a strobe light projected a dynamically flashing spot (spot diameter 0.8m, moving speed 15m / s) at a frequency of 0.5Hz 50 meters in front of the flock's flight path. Actual measurements showed that this combined avoidance strategy caused the flock's formation to deflect 35° within 6 seconds, successfully avoiding the wind turbine's danger zone.

[0040] For nocturnal bats: A strategy primarily using ultrasonic interference, supplemented by ultraviolet pulses. Based on their reliance on echolocation, wideband-sweeping ultrasonic waves disrupt their sonar systems. Simultaneously, the bats' unique sensitivity to ultraviolet light (due to the abundance of S-opsin photosensitive protein in their retina) is utilized for visual interference, achieving multi-sensory synergistic inhibition. When a dual-spectrum camera captures thermal imaging signals of three greater brown bats (wingspan 0.28 m, flight altitude 30 m) in low-light conditions, millimeter-wave radar simultaneously detects their distinctive spiral ascent trajectory (angular velocity 22° / s). The AI ​​identification module confirms the species through sonar feature analysis (echolocation pulse frequency 82-86 kHz, interval 3 ms) and identifies them as being in a hovering state preying on insects. The central control module immediately activated the bat-specific mode: the bottom ultrasonic array emitted dynamically matched interference waves (center frequency 85kHz, bandwidth ±5kHz sweep frequency, sweep period 4ms) to disrupt their echolocation ability; simultaneously, the strobe light switched to the ultraviolet band (wavelength 365nm, pulse frequency 12Hz) to utilize the special photosensitive cells in the bat's retina for visual interference. On-site monitoring data showed that after the avoidance was triggered, the bat flock dispersed within 2.3 seconds, increasing their flight altitude to over 55 meters, exceeding the range of the leaf sweep.

[0041] For solitary raptor species (such as the white-tailed eagle): A tiered enhanced deterrence mode is activated. Based on their high aggression and adaptability, a composite acoustic impact of high-frequency alarm sounds and low-frequency vibration waves is employed, combined with laser tracking to create visual pressure. Simultaneously, the ultrasonic intensity is dynamically adjusted based on real-time distance (precise sound pressure gradient control via a PID controller), ensuring both deterrence effectiveness and avoiding excessive stress responses. When the millimeter-wave radar detects a rapidly diving target with a 2.3-meter wingspan (acceleration 4 m / s²) at a distance of 80 meters, the AI ​​identification module uses wingtip vortex characteristics and thermal imaging body temperature data (39.8℃) for species identification. Based on historical behavior data, the central control module predicted that the target would enter the danger zone at the blade root in 4.2 seconds and immediately activated the highest-level avoidance scheme: the acoustic avoidance module released a specific frequency band of expulsion sound waves (a mixture of high-frequency alarm sounds and low-frequency vibration waves, covering a frequency range of 0.8-18kHz), the optical avoidance module simultaneously activated the laser tracking mode, and the laser spotlight tracked the target's head with 532nm green light and a 20Hz frequency flash, while the ultrasonic transmitter formed a conical sound pressure field (sound pressure gradient from 120dB@1m to 95dB@50m). Actual measurements showed that the target completed an emergency climb within 2.8 seconds under the combined acoustic and optical effects, and the closest distance between its flight path and the blade increased from the predicted 1.2 meters to 9.7 meters. During this process, the central control module dynamically adjusted the avoidance intensity. When the millimeter-wave radar 4 detected that the target was starting to move away, the central control module commanded the ultrasonic avoidance control module to reduce the ultrasonic intensity from 15V to 8V, ensuring both avoidance effectiveness and energy consumption.

[0042] S7, the central control module controls the multi-mode avoidance system to execute the optimal avoidance scheme.

[0043] This invention combines an intelligent monitoring system and a repellency system to achieve real-time monitoring and effective repellency of flying objects around wind farms, improving the operational safety and reliability of wind farms. The intelligent monitoring system comprehensively monitors flying objects within the sweep range of wind turbine blades, using an AI recognition module to identify bats and birds, providing accurate target information for the repellency system. Through the built-in AI recognition module, it reacts precisely based on the different species, behaviors, and environments of bats and birds, implementing targeted repellency measures to overcome the limitations of traditional single repellency methods and avoid harm to certain bird or bat species. The repellency system uses loudspeakers to play the calls of birds and bat predators for acoustic repellency, while simultaneously utilizing laser spotlights and ultrasonic transmitters for optical and ultrasonic repellency, forming a multi-mode, multi-layered repellency approach that effectively avoids the limitations of single repellency methods.

[0044] This invention's power supply system utilizes a combination of solar panels and batteries, achieving the utilization of green energy and reducing the operating costs and environmental impact of the device. The solar panels power the batteries, ensuring stable operation of the device under various weather conditions.

[0045] The control system of this invention uses a microcontroller to centrally control the intelligent monitoring system, avoidance system, power supply system, etc., and through connection with the cloud platform, it realizes remote monitoring and data analysis, which facilitates remote management and optimization of the device.

[0046] The self-learning algorithm module of this invention can adjust and optimize the avoidance strategy in real time based on the effect of the avoidance action, thereby improving the adaptability and effectiveness of the avoidance effect.

[0047] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A multi-mode wind turbine bat and bird avoidance device, characterized in that, The system includes an outer shell, inside which a monitoring compartment, an avoidance compartment, and a control compartment are formed by partitions. A millimeter-level radar is mounted on the top of the outer shell, and a dual-spectrum camera is mounted on the front of the outer shell, with the radar and camera corresponding to the positions of the monitoring compartment. The monitoring compartment contains a first module, a second module, and an AI recognition module. A speaker is mounted on the front of the outer shell, and strobe lights and laser spotlights are mounted on both sides of the outer shell, corresponding to the positions of the avoidance compartment. An ultrasonic transmitter is installed inside the avoidance compartment, which also contains a multi-mode avoidance system. Photovoltaic panels are fixedly mounted at both ends of the top of the outer shell. The control compartment contains a power supply module, a hollow module, and a communication module.

2. The multi-mode wind turbine bat and bird avoidance device according to claim 1, characterized in that, The AI ​​recognition module includes a wingspan recognition module, a flight attitude recognition module, a thermal imaging feature recognition module, and an acoustic feature analysis module. The wingspan recognition module, the thermal imaging feature recognition module, and the acoustic feature analysis module are used to sequentially identify the category of the organism, and the flight attitude recognition module is used to identify the flight patterns and movement patterns of the identified organism.

3. The multi-mode wind turbine bat and bird avoidance device according to claim 1, characterized in that, The multi-mode avoidance system includes an optical avoidance module, an acoustic avoidance module, and an ultrasonic avoidance module. The loudspeaker is electrically connected to the acoustic avoidance module, the strobe light and the laser spotlight are electrically connected to the optical avoidance module, and the ultrasonic generator is electrically connected to the ultrasonic avoidance module.

4. The multi-mode wind turbine bat and bird avoidance device according to claim 1, characterized in that, The power supply module includes an MPPT controller, which connects the tilted photovoltaic panel and the lithium iron phosphate battery pack via a ring copper busbar; the communication module adopts a hierarchical architecture, and real-time control commands are transmitted through a time-sensitive network.

5. A method for repelling bats and birds in a multi-mode wind turbine, characterized in that, Using the multi-mode wind turbine bat and bird avoidance device according to any one of claims 1-4, the process includes the following steps: S1. Data is collected through millimeter-level radar and dual-spectrum camera. The data collected by millimeter-level radar includes flight speed, acceleration, wingbeat frequency, and formation. The data collected by dual-spectrum camera includes infrared thermal imaging data. S2. The wingspan recognition module extracts the contour features of biological targets based on the point cloud data acquired by millimeter-wave radar and a target detection algorithm based on deep learning, and performs a preliminary classification of the biological categories. S3. The thermal imaging feature recognition module extracts the surface temperature distribution characteristics of organisms based on infrared thermal imaging data, and further calibrates the biological category identified in step S2. S4, the acoustic feature analysis module, collects echolocation pulses through a high-frequency microphone array, extracts frequency domain features (including center frequency, bandwidth, and pulse interval) through fast Fourier transform, and classifies the organisms identified in step S3 using a support vector machine classifier to obtain the final classification result. S5. The flight attitude recognition module extracts information on flight speed, acceleration, wingbeat frequency, and formation, obtaining the flight patterns and movement patterns of the organisms identified in step S4; including the following steps: S51. Obtain the flight patterns of the organisms identified in step S4, including the target trajectory prediction, the estimated wingbeat frequency, and the formation quantification. S52. Input the extracted flight speed, acceleration, and wingbeat frequency into the long short-term memory network to identify motion patterns; S6. Based on the biological classification obtained in step S4 and the flight patterns and movement patterns of each biological obtained in step S5, match the optimal avoidance scheme in the mapping database. S7. Generate corresponding control commands through the central control module and send them to each module to execute the optimal avoidance scheme obtained in step S5.

6. A method for repelling bats and birds in a multi-mode wind turbine according to claim 5, characterized in that, The target trajectory prediction value in step S51 is calculated using the following formula: in, Indicates time k before, based only on The predicted values ​​of the state based on the data at and before time are the state vector (including position and velocity). Indicates time Previously, based solely on The predicted value of the state based on data from the current time and earlier; Here is the state transition matrix. , To control the input.

7. A method for repelling bats and birds in a multi-mode wind turbine according to claim 5, characterized in that, The estimated value of the wingbeat frequency in step S51 is calculated using the following formula: In the above formula, The Fourier transform result of signal s(t) represents the amplitude and phase of the signal at frequency f. Its absolute value is the spectrum, and the frequency corresponding to the peak frequency is the estimated value of the flapping frequency; f represents frequency, t represents time, and j represents the imaginary unit.

8. A method for repelling bats and birds in a multi-mode wind turbine according to claim 5, characterized in that, The method for obtaining the quantization value of the formation in step S51 is as follows: a density-based spatial clustering algorithm (such as DBSCAN) is used to segment the multi-target point cloud at the same time, and the geometric center C of the group and the average distance between individuals are calculated. And the formation's main axis direction angle θ, thereby quantifying the formation pattern and obtaining the quantified value of the formation pattern.

9. A method for repelling bats and birds in a multi-mode wind turbine according to claim 8, characterized in that, In step S52, the LSTM unit in the Long Short-Term Memory network is used to perform calculations at each time step t. Finally, the hidden state at the end of the network is concatenated with the quantized value of the formation pattern and fed into the multilayer perceptron classifier for motion pattern recognition, as shown below: In the above formula, The output represents the probability distribution of various movement patterns, including migration, predation, hovering, and attack. This represents the weight matrix in the classifier. This represents the bias term in the classifier. Indicates the hidden state at the end of the network. The quantified value representing the formation pattern.

10. A method for repelling bats and birds in a multi-mode wind turbine according to claim 5, characterized in that, In step S6, acoustic and optical deterrence are used for gregarious birds; ultrasonic interference is used as the main deterrence method and ultraviolet pulse is used as a supplementary deterrence method for nocturnal bats; and a graded and enhanced deterrence mode is used for the five solitary bird species.

Images