Directional bird damage expelling method and device for power transmission line

Abstract

The invention provides a directional bird damage expelling method and device for a power transmission line, and relates to the technical field of power transmission lines, the device comprises a multi-mode sensing unit, a central control box and a plurality of distributed neural interference nodes, bird behaviors are collected through the multi-mode sensing unit, a behavior phase model of the central control unit is used for predicting a high-risk phase of'intention to land 'of birds, and the high-risk phase of'intention to land' of the birds is obtained. And a game scheduler cooperates with a reinforcement learning algorithm to generate a non-periodic and dynamically optimized repelling strategy, so that the problems that a traditional bird repelling signal mode is single and birds are easy to adapt to are solved, and an execution end cooperatively drives a miniature electromagnetic field emission array and a directional polarized light projector to achieve the purpose of repelling birds. According to the method, accurate and cross-sensory interference on a geomagnetic navigation system and a visual system of birds is achieved, indiscriminate interference of traditional acousto-optic driving is remarkably eliminated, in addition, through the dynamic movement effect of an electromagnetic interference field, orderly guiding is carried out on group birds, and the secondary risk that the birds collide with wires due to disordered driving is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of power transmission line technology, specifically to a method and apparatus for directional bird removal from power transmission lines. Background Technology

[0002] According to Chinese Patent No. CN113875743A, a bird damage fault detection and early warning system for transmission lines includes a fault detection system database, a bird damage fault record database, and an instant fault early warning module. The fault detection system database includes a real-time monitoring module, a fault detection module, and a real-time feedback module. The bird damage fault record database includes statistics on the number of bird damages, statistics on fault types, and tracking and monitoring of bird damage types. This invention also discloses a method for operating the transmission line bird damage fault detection and early warning system. This transmission line bird damage fault detection and early warning system, through the combined effect of a real-time monitoring module, a fault detection module, a real-time feedback module, and routine manual inspections, achieves strong detection and early warning effects for transmission lines and has good performance.

[0003] The aforementioned patent documents and prior art have the following technical problems when used:

[0004] Problem 1: Traditional methods often involve broad-spectrum interference, such as omnidirectional sound waves or flashes, which lack the ability to identify bird behavior, resulting in low deterrence efficiency and an excessively large range of action, causing unnecessary noise or light pollution to unrelated animals.

[0005] Question 2: Traditional bird deterrence methods, such as fixed-frequency sound waves, simple flashlights, and static raptor bionic devices, have a single deterrence signal pattern and a fixed effect. They usually become ineffective within a few weeks after installation. Birds easily become accustomed to fixed-frequency and single-pattern deterrence signals. Birds can quickly establish a conditioned reflex of "disturbing and harmless" and eventually become completely adapted, making the deterrence effect unstable and unsustainable, resulting in a decline in long-term effectiveness.

[0006] Thirdly, when faced with large flocks of migratory birds or groups crossing power transmission corridors, traditional independent devices can only disperse them in a disorderly manner, which can easily cause the birds to be startled and scatter, thus increasing the secondary risk of birds hitting the power lines. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a method and apparatus for directional bird removal from power transmission lines, which solves the problems mentioned in the background section.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a bird-directional deterrent device for power transmission lines, comprising a crossbar, a central control box on the surface of the crossbar, a communication synchronization module, a central control unit, and an energy storage module inside the central control box, a multimodal sensing unit on the surface of the crossbar, a neural interference node on the bottom surface of the crossbar, a miniature electromagnetic field emitting array and a directional polarized light projector on the surface of the neural interference node, a directional ultrasonic pulser on the bottom surface of the neural interference node, and an electromagnetic induction pickup on the surface of the crossbar.

[0011] Preferably, the multimodal sensing unit includes a frequency-modulated continuous wave radar and a visual camera, used to provide the bird's three-dimensional coordinates, velocity, and species classification.

[0012] Preferably, the micro electromagnetic field emitting array adopts an orthogonal coil array structure to precisely control the frequency, intensity, and direction of the emitted magnetic field and radio frequency pulses, and the frequency range includes extremely low frequency and very low frequency.

[0013] Preferably, the directional polarized light projector employs a narrow-wavelength light source with a rapidly switchable polarization angle to project a randomly rotating and switching polarization pattern onto the bird's estimated landing area.

[0014] Preferably, one side of the electromagnetic induction pickup is provided with a power transmission line, and the power transmission line maintains a safe gap to achieve non-contact electromagnetic energy pickup.

[0015] Preferably, the method includes the following steps:

[0016] Sp1: Collects the location, speed, flight trajectory, and estimated landing point of bird flocks through a multimodal sensing unit;

[0017] Sp2: The central control unit runs a behavioral phase model, abstracts bird behavior into a phase sequence, and predicts its navigation intentions;

[0018] Sp3: The central control unit operates the game scheduler, which generates aperiodic electromagnetic interference patterns and spatiotemporal illusion sequences according to the navigation intent, and outputs phase codes;

[0019] Sp4: Drives multiple distributed neural interference nodes in a coordinated manner based on phase codes;

[0020] Sp5: The neural interference node drives a miniature electromagnetic field emission array to emit extremely low-power non-periodic magnetic field pulses and radio frequency pulses toward the estimated landing point and group targets, interfering with the birds' geomagnetic navigation or neural perception.

[0021] Sp6: The neural interference node simultaneously drives the directional polarized light projector to project a high refresh rate random rotating polarization pattern, which, combined with directional acoustic pulses, creates a spatial illusion to prevent birds from landing.

[0022] Sp7: Based on birds' responses to electromagnetic pulse frequency and polarization pattern switching rate, the strategy is adjusted in real time using reinforcement learning algorithms to continuously break birds' habituation.

[0023] Preferably, in Sp4, the central control unit drives multiple neural interference nodes to start according to a predetermined phase sequence and time difference through phase codes, so that the electromagnetic field pulses generate a dynamic movement effect in space, forming a movable deterrent barrier, and realizing the orderly guidance of the bird flock.

[0024] Preferably, the behavioral phase model in the central control unit adopts a probabilistic finite state machine to distinguish between the intrusion behavior, resting behavior and foraging behavior of bird flocks, and to provide state input for the game scheduler.

[0025] Preferably, in Sp5, the essential function of the extremely low-power aperiodic magnetic field pulse and radio frequency pulse interference is to disrupt the bird's accurate navigation and positioning perception of the geomagnetic field, and the pulse intensity is strictly controlled within a threshold that is harmless and does not interfere with communication.

[0026] Beneficial effects

[0027] This invention provides a method and apparatus for directional bird removal from power transmission lines. It offers the following advantages:

[0028] 1. This invention combines a multimodal sensing unit with a behavioral phase model, achieving an improvement from "blindly driving away" to "target locking," and achieving a balance between precise orientation and eco-friendliness. The system only activates when it detects a high-risk phase in which birds "intend to land." Furthermore, the miniature electromagnetic field emission array and directional polarized light projector used are "silent attacks" targeting specific bird senses, successfully avoiding the indiscriminate interference of traditional sound and light repelling on ground animals and the environment. This upgrades bird repelling from "regional pollution" to "surgical intervention," ensuring the repelling effect while achieving harmonious coexistence with the surrounding ecosystem.

[0029] 2. This invention employs game-theoretic scheduling and reinforcement learning algorithms to construct a continuously evolving anti-adaptive deterrent, breaking the curse of traditional devices' "three-week failure." It transforms the system from a rigid "fixed pattern" into a "dynamic game" agent, continuously disrupting bird cognition through non-periodic strategy combinations and cross-sensory collaborative interference. Simultaneously, the reinforcement learning algorithm adjusts the weights of the dispersal strategies in real time based on bird responses, automatically eliminating ineffective old strategies. This ensures the system possesses strong adaptive capabilities, with its evolutionary speed always outpacing the birds' adaptation speed, guaranteeing the long-term stability and sustained effectiveness of the dispersal effect.

[0030] 3. This invention upgrades the "panic dispersal" to "orderly guidance" through the "dynamic movement effect" supported by the communication synchronization module. Instead of trying to frighten the flock, it drives multiple distributed neural interference nodes to start in a predetermined phase sequence and time difference to form a controllable and continuously moving virtual electromagnetic barrier in front of the flock's flight path. This method enables precise "herding" management of large-scale flocks, forcing the entire group to turn around and leave the power transmission corridor safely without being frightened. It effectively avoids secondary accidents such as panic dispersal of the flock and bird collisions with power lines that may be caused by traditional disorderly dispersal. Attached Figure Description

[0031] Figure 1 This is a flowchart of the expulsion method of the present invention;

[0032] Figure 2 This is a structural diagram of the deflection device of the present invention;

[0033] Figure 3 This is an isometric view of the deflection device of the present invention;

[0034] Figure 4 This is a cross-sectional view of the central control box of the present invention.

[0035] The components include: 1. Crossbar; 2. Central control box; 3. Multimodal sensing unit; 4. Communication synchronization module; 5. Central control unit; 6. Energy storage module; 7. Neural interference node; 8. Miniature electromagnetic field emission array; 9. Directional polarized light projector; 10. Directional ultrasonic pulser; 11. Electromagnetic induction pickup. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Example 1:

[0038] like Figures 1 to 4 As shown, a bird-directional deterrent device for power transmission lines includes a crossbar 1. A central control box 2 is mounted on the surface of the crossbar 1. The central control box 2 contains a communication synchronization module 4, a central control unit 5, and an energy storage module 6. A multimodal sensing unit 3 is mounted on the surface of the crossbar 1. A neural interference node 7 is mounted on the bottom surface of the crossbar 1. A miniature electromagnetic field emission array 8 and a directional polarized light projector 9 are mounted on the surface of the neural interference node 7. A directional ultrasonic pulser 10 is mounted on the bottom surface of the neural interference node 7. An electromagnetic induction pickup 11 is mounted on the surface of the crossbar 1. The multimodal sensing unit 3 includes... Frequency-modulated continuous wave radar and visual cameras are used to provide three-dimensional coordinates, velocity, and species classification of birds. The miniature electromagnetic field emission array 8 adopts an orthogonal coil array structure to precisely control the frequency, intensity, and direction of the emitted magnetic field and radio frequency pulses, with a frequency range including extremely low frequency and very low frequency. The directional polarized light projector 9 uses a narrow wavelength light source with a rapidly switchable polarization angle to project randomly rotating and switching polarization patterns onto the bird's estimated landing area. The electromagnetic induction pickup 9 is fixed on the crossbar 1 and maintains a safe gap with the power transmission line to achieve non-contact electromagnetic energy pickup.

[0039] Horizontal bar 1, as the main horizontal support structure of the transmission tower, serves as the physical mounting base for all functional modules. All structures are fixed to horizontal bar 1 using insulating clamps, ensuring a sufficient safe distance from the high-voltage conductors. Central control box 2 is fixed to the surface of horizontal bar 1, serving as the system's central data processing and strategy scheduling center. Central control unit 5 is located inside central control box 2. This unit is an embedded high-performance processor that fuses multimodal sensing data in real time, runs complex behavioral phase models, game schedulers, and reinforcement learning algorithms. This unit is the source of decision generation and command issuance, undertaking three core tasks: first, real-time fusion of high-bandwidth data from multimodal sensing unit 3; second,... Second, the core algorithm, namely the behavioral phase model and the game scheduler, is run. Third, precise microsecond-level instructions are sent to all neural interference nodes 7 through the communication synchronization module 4. The communication synchronization module 4 is located inside the central control box 2. This module establishes a high-speed, anti-interference industrial bus connection with each neural interference node 7 to achieve precise clock synchronization at the microsecond level, ensuring the phase consistency of pulse transmission of each node during dynamic expulsion. The internal link is connected to each neural interference node 7 through a highly reliable industrial bus to ensure low latency and high anti-interference of instructions. The external link is used to communicate with the cloud platform through a 4G or 5G wireless module for remote data reporting, receiving policy updates, or manual intervention.

[0040] The electromagnetic induction pickup 11 is fixed to the surface of the crossbar 1. The electromagnetic induction pickup 11 is an open-clamp current transformer (CT) using a high-insulation composite material clamp. Its induction coil surrounds the transmission line but does not make any physical contact with it. It generates electricity by inducing a strong alternating magnetic field produced by the alternating current in the transmission line, without exerting any mechanical pressure on the transmission line itself. It safely senses the alternating magnetic field around the transmission line, obtains electricity, and transmits the energy to the energy storage module 6. The power extraction capacity of the electromagnetic induction pickup 11 is limited and fluctuates with the line load; therefore, the energy must first be stored in the energy storage module 6. The energy storage module 6 is located inside the central control box 2, storing the energy from the electromagnetic induction pickup 11 and providing the system with the instantaneous high-power electrical energy required during extremely low-frequency electromagnetic pulse or ultrasonic pulse emission. This module includes a supercapacitor and a lithium titanate battery. The supercapacitor is used to cope with the instantaneous high-power pulse discharge during decoupling, and the lithium titanate battery... This is used for the long-term stable power supply of the system and can withstand extreme temperature changes on the tower. The multimodal sensing unit 3 is fixed on the surface of the crossbar 1, facing the power transmission corridor. This unit is connected to the central control unit 5 through a high-speed data interface. It includes a frequency-modulated continuous wave radar and a visual camera. The frequency-modulated continuous wave radar is responsible for all-weather, wide-area coarse surveillance. It is responsible for detection, wake-up and vector tracking. It works with a low duty cycle and provides accurate three-dimensional coordinates, velocity and acceleration vectors of birds. The visual camera is activated after the radar detects the target. It is responsible for high-precision species classification and behavioral posture recognition. The image data it collects is used to distinguish whether the bird is a high-risk crow or a protected bird of prey, and to identify whether the bird is in a "landing intention" posture such as deceleration and claw extension. It is responsible for answering "What kind of bird is this?" and "What does it want to do?". It locks the target species through image recognition and uses posture estimation to assist the behavioral phase model in determining whether the bird is circling, foraging or preparing to land.

[0041] Multiple neural interference nodes 7 are distributed along the bottom surface of the crossbar 1. Each node is a highly integrated, IP68-rated, fully sealed execution unit that receives synchronization commands from the central control box 2 via a communication bus. A miniature electromagnetic field emission array 8 is located on the surface of the neural interference nodes 7. This array consists of an array of orthogonal coils and is designed to target the bird's navigation system. The central control unit 5 can precisely generate non-periodic magnetic field pulses with specific directions and vectors in space by controlling the current phase and amplitude of three sets of orthogonal coils. This array emits extremely low frequency (VLF) / ELF signals to interfere with the bird's geomagnetic navigation and simultaneously emits low-power radio frequency (RF) pulses to interfere with its neural perception. The extremely low frequency and very low frequency (ELF / VLF) bands are specifically used to interfere with the geomagnetic perception system that birds rely on for navigation, creating a kind of "magnetic noise" that disables the bird's built-in compass, making it unable to accurately locate the tower. The RF pulse is a type of neural perception interference designed to create a low-power but high-frequency electromagnetic field around the bird just before it lands, which may cause slight discomfort to the nerve endings in its skin or feathers, causing it to instinctively abandon landing. A directional polarized light projector 9 is located on the neural interference node. On the surface of point 7, the projector uses a high-brightness, narrow-wavelength LED as the light source. A liquid crystal or mechanically variable polarizer is mounted in front of it. Strictly following instructions, it rapidly and randomly switches the polarization angle of the light at a high refresh rate, projecting a randomly rotating polarization pattern onto the estimated landing area to create spatial illusion. This is designed for birds' visual systems, as birds are highly sensitive to light polarization. Using a narrow-wavelength light source with rapidly switchable polarization angles, the central control unit 5 can apply different voltages to the liquid crystal, causing the polarization angle of the emitted light to rotate or switch randomly at high speed within milliseconds. To birds, this originally solid landing point (the surface of crossbar 1) appears as a "shimmering water surface" or an "unstable space," thus generating a strong spatial illusion and avoidance response. The directional ultrasonic pulser 10 is located on the bottom surface of the neural interference node 7. This pulser consists of multiple piezoelectric transducers, forming an ultrasonic phased array to achieve high-level focusing of ultrasonic energy, avoiding the indiscriminate interference problem of traditional sound wave bird deterrence. By controlling the emission phase of each transducer, it can focus ultrasonic energy into an extremely narrow beam, accurately striking the target and achieving directional targeting without disturbing people. Specific Implementation Example 2:

[0043] like Figures 1 to 4 As shown, based on the content of the above specific embodiments, the following content is further disclosed:

[0044] The expulsion method includes the following steps:

[0045] Sp1: The multimodal sensing unit 3 collects the position, speed, flight trajectory and estimated landing point of the bird flock;

[0046] Acquisition and wake-up: When the system is in low-power sleep mode, the electromagnetic induction pickup 11 continuously charges the energy storage module 6. The frequency-modulated continuous wave radar in the multimodal sensing unit 3 scans with a 10% duty cycle. Once the radar detects a moving target and confirms that its vector is pointing towards the tower, the central control unit 5 is immediately woken up and the visual camera is activated.

[0047] This is a three-step process of "detection-tracking-prediction," entirely completed collaboratively by the multimodal sensing unit 3 and the central control unit 5. For position and velocity acquisition (detection): the frequency-modulated continuous wave radar is the first line of defense, scanning the protected area 24 / 7 with an extremely low power LED duty cycle (energy saving). Once an object enters, it immediately locks onto it and operates at a high refresh rate, acquiring the target's precise three-dimensional coordinates (x, y, z) and velocity vector (v) in real time. x ,v y ,v z At this stage, the system only knows that "something exists," but not "what it is." Flight trajectory acquisition (tracking): The central control unit 5 does not only look at a single data point. It receives a continuous data stream from the radar and runs a Kalman filter algorithm. This algorithm smooths out noise and jitter in individual data points, fitting discrete (x,y,z) coordinates into a smooth flight trajectory. Simultaneously, the radar signal wakes up the visual camera, which initiates target recognition and tracking algorithms to confirm that the target species is "birds." The visual tracking data is then fused with the radar trajectory to ensure the continuity and accuracy of the trajectory. Landing point prediction (forecasting): This is the most crucial step. The central control unit 5 acquires the smooth trajectory and velocity vector. The system extrapolates the trajectory. Simultaneously, the system has pre-stored three-dimensional digital models of the tower and crossbar 1 (i.e., the "safe zone" model). The system calculates in real time: if the bird's speed (v) decreases, its altitude (z) decreases, and its extrapolated trajectory is about to collide with the surface of the tower's three-dimensional model, then this collision point is defined as the "predicted landing point."

[0048] Sp2: The central control unit 5 runs a behavior phase model, which abstracts bird behavior into a phase sequence and predicts its navigation intention. The behavior phase model in the central control unit 5 uses a probabilistic finite state machine to distinguish between the intrusion behavior, resting behavior and foraging behavior of bird groups, and provides state input for the game scheduler.

[0049] Modeling and prediction: The central control unit 5 receives real-time data streams from radar and cameras, the Behavioral Phase Model (PFSM) is activated, the data stream is parsed into S2: approach state, and P(S3: intended landing) = 85% is predicted;

[0050] PFSM acts as the system's "bird psychologist," and "phase" is a "discrete label" for bird behavior. The central control unit 5 performs cluster analysis on the continuous data (velocity, acceleration, altitude, and attitude) collected in Sp1, namely: high speed + vector pointing towards the tower = "intrusion phase," low speed + random vector + low-altitude circling = "foraging phase," sharp decrease in speed + decrease in altitude + trajectory pointing towards the crossbar = "intended landing phase," and zero speed + position on the tower = "resting phase." The core of PFSM is "probability transition." When a bird is in an "intrusion phase," the model calculates the probability of it transitioning to other phases in real time. If the trajectory prediction of Sp1 shows "imminent collision," and the camera detects the bird "spreading its claws" or "folding its wings" through attitude recognition, this data is input into PFSM. PFSM will instantly calculate: P(transition to "intended landing phase") > 95. This high-probability phase transition prediction is the "navigation intention" determined by the system. At this time, the system no longer needs to wait for the bird to actually land, but triggers Sp3 in advance.

[0051] SP3: The central control unit 5 operates the game scheduler, which generates non-periodic electromagnetic interference patterns and spatiotemporal illusion sequences according to the navigation intent, and outputs phase codes;

[0052] Game Theory and Decision Making: A high-probability S3 prediction triggers the game scheduler, which queries the Q-table of the reinforcement learning (RL) module, selects a high-value non-periodic policy, i.e., script A, for the current target (e.g., "magpie"), and generates a "phase code" containing precise timestamps and action sequences.

[0053] The game-theoretic scheduler acts as the system's "general staff," its sole objective being to "break the mold," ensuring the birds can never predict its next move. The PFSM's "high-probability intention to land" prediction serves as the trigger signal. To combat "habituation," the birds are highly intelligent; if each expulsion is accompanied by three beeps, they will land on the horn after three days. Non-periodicity and pseudo-randomness are the only ways to ensure long-term effectiveness. The scheduler has an internal "policy library." When the PFSM reports "intention to land," the scheduler doesn't always use the strongest policy; instead, it runs an ∈-Greedy (Epsilon greedy) algorithm, selecting a policy with a 90% probability. The currently known optimal strategy is (Strategy A: moderate magnetic field + rotating polarized light), but there is a 10% probability of deliberately choosing an "exploratory strategy" (Strategy B: no magnetic field + strong ultrasonic waves, or Strategy C: allow landing and then suddenly punish). This mechanism ensures that birds can never establish a stable "causal relationship". The "random weights" of this algorithm are adjusted by the next reinforcement learning algorithm. The "phase code" is not a simple "on" signal, but a high-precision data packet. This data packet is broadcast by the central control unit 5 through the industrial bus of the communication synchronization module 4 and contains all the information required to execute the strategy.

[0054] Sp4: Drives multiple distributed neural interference nodes in a coordinated manner based on phase codes;

[0055] Collaborative drive: The central control unit 5 broadcasts the "phase code" to three neural interference nodes (7A, 7B, 7C) near the target landing point via the high-speed bus of the communication synchronization module 4;

[0056] All neural interference nodes 7 simultaneously received the "phase code" from Sp3. Since the communication synchronization module 4 achieved microsecond-level clock synchronization, nodes 7A and 7B will strictly follow the TimeStamp in the data packet to start their respective micro electromagnetic field emission array 8 and directional polarized light projector 9 at the same time. This is "coordination".

[0057] In Sp4, the central control unit 5 drives multiple neural interference nodes 7 to start according to a predetermined phase sequence and time difference through phase codes, causing the electromagnetic field pulses to generate a dynamic movement effect in space, forming a movable deterrent barrier to achieve orderly guidance of the bird flock. This is an advanced strategy for group targets (bird flocks). When the radar of Sp1 detects a "area target" (large and dispersed signal) rather than a "point target," that is, a flock of birds attempting to cross the line, the game scheduler will switch to "herding mode." The game scheduler will not execute the "fixed-point strike" of Sp5, but will instead execute the "dynamic movement effect" strategy, sending a series of "phase codes" with time differences to multiple nodes on a line (e.g., 7A, 7B, 7C, 7D). The operating logic is as follows:

[0058] T = 0.0s: Node 7A pulse;

[0059] T = 0.2s: Node 7A is closed, Node 7B pulses;

[0060] T = 0.4s: Node 7B is turned off and Node 7C is pulsed. This creates a "moving, invisible electromagnetic interference wall" in front of the flock's flight path, forcing the entire flock to change course and achieve orderly guidance, rather than scattering the flock and causing chaos.

[0061] Because "dispersing" flocks of birds would be disastrous, as the frightened birds would scatter and potentially collide with the power lines, the goal is "orderly guidance." This "movable barrier" functions similarly to a sheepdog; it doesn't "bite" the flock, but rather creates pressure on one side of the flock's flight vector, forcing the entire flock to turn as a whole and safely "herd" away from the power transmission corridor.

[0062] Sp5: The neural interference node 7 drives the miniature electromagnetic field transmitting array 8 to emit extremely low-power non-periodic magnetic field pulses and radio frequency pulses towards the estimated landing point and group targets, interfering with the birds' geomagnetic navigation or neural perception. The essential function of the extremely low-power non-periodic magnetic field pulses and radio frequency pulse interference is to disrupt the birds' accurate navigation and positioning perception of the geomagnetic field. The pulse intensity is strictly controlled within the threshold that is harmless and does not interfere with communication.

[0063] Neural interference (electromagnetic): The miniature electromagnetic field emission arrays 8 on nodes 7A, 7B, and 7C are activated synchronously according to the instructions of the phase code, emitting extremely low-frequency non-periodic magnetic field pulses to create a "magnetic vacuum" or "magnetic storm" in the landing area, completely disrupting the birds' geomagnetic navigation perception.

[0064] Extremely low-power aperiodic magnetic field pulses and radio frequency pulses can achieve "harmless" and "non-interfering" communication for birds. The orthogonal coil array (triaxial coil) within the miniature electromagnetic field emission array 8 is driven, and the central control unit 5, by controlling the current magnitude and phase of the triaxial coils, can synthesize magnetic field pulses in space with arbitrary vector directions. Birds (especially pigeons and migratory birds) rely on the Earth's magnetic field for navigation; their brains can sense weak magnetic field tilt angles and intensities, acting like a built-in biological compass. The "extremely low frequency (ELF / VLF) aperiodic magnetic field pulses" of Sp5, with frequencies close to Earth's magnetic field fluctuations, when this "fake" magnetic field signal is emitted at extremely low power (but at close range),... When applied, it "overwhelms" the weak, stable Earth's magnetic field signal. For birds, this is equivalent to a human trying to see a distant landscape when suddenly a high-frequency flashing neon light appears in front of them. It doesn't harm the eyes, but it completely "blinds" navigation, causing a strong sense of disorientation and anxiety. Radio frequency pulses (RF) are another mechanism. Through electromagnetic induction, they induce a weak current on the bird's feathers or skin surface (especially the claws, which are rich in nerve endings). This current is far from enough to cause harm, but it is enough to trigger its nerve endings, producing a feeling of "numbness," "itching," or "static" discomfort. This is a kind of "discomfort" that birds cannot understand and originates from themselves, causing them to instinctively flee the area.

[0065] Sp6: The neural interference node 7 simultaneously drives the directional polarized light projector 9 to project a high refresh rate random rotating polarization pattern, which, combined with directional acoustic pulses, creates a spatial illusion to prevent birds from landing.

[0066] Spatial illusion (visual and acoustic): At the same time, the node's directional polarized light projector 9 projects a high refresh rate random rotating polarization pattern toward the landing point, and the directional ultrasonic pulser 10 also assists by emitting a focused sound beam. When the bird attempts to land, it simultaneously encounters compass failure, ground shaking, and piercing noise, and is forced to abort the landing.

[0067] The directional polarized light projector 9 contains a high-brightness LED (typically in the green or blue band, which birds are sensitive to). The light passes through a liquid crystal polarizer, which is crucial. The central control unit 5 can change the arrangement of molecules within milliseconds by altering the voltage applied to the liquid crystal, thus instantaneously changing the polarization angle of the emitted light. While the human eye can barely distinguish polarized light, a bird's eye can. To a bird, the polarized light reflected from a solid, rough landing surface (bar 1) is stable and predictable. When the directional polarized light projector 9 illuminates the estimated landing point with a "high refresh rate, randomly rotating polarization pattern," this "solid ground" appears to the bird. The sudden transformation into a "shimmering, rapidly flashing, and extremely unstable water surface" creates a spatial illusion of "solid turning into liquid," triggering a depth perception disorder in birds and making them "afraid to put their feet down." Combined with directional acoustic pulses, which constitute "cross-sensory interference," just as the birds are confused by the "liquid ground," the directional ultrasonic pulser 10 fires a focused, high-energy ultrasonic "bullet" at them. The birds cannot hear it (non-auditory), but they can feel it (physical impact). The birds simultaneously experience "navigation failure," "ground liquefaction," and a "mysterious attack." This multi-sensory synergistic interference causes significant cognitive confusion, leading them to mark the location as "extremely dangerous," thus achieving long-term deterrence.

[0068] Sp7: Based on the birds' responses to electromagnetic pulse frequency and polarization pattern switching rate, the strategy is adjusted in real time through reinforcement learning algorithms to continuously break the birds' habituation.

[0069] Feedback and Learning: The multimodal perception unit 3 detects that the bird has turned and entered S5: escape state. The central control unit 5 records the complete log of this (state, action, reward +10) and updates the reinforcement learning Q table. The system clears the cache and returns to the low-power alert state of Sp1.

[0070] This is the system's "cerebral cortex," ensuring that the system "evolves" faster than birds; this is a closed-loop reinforcement learning (RL) process.

[0071] Step 1: Action: The game scheduler selects a strategy, namely strategy A (magnetic field + polarized light);

[0072] Step 2: Observation: The system will not stop working, and the multimodal sensing unit 3 will continue to track the driven target;

[0073] Step 3: Feedback: The system automatically evaluates the results.

[0074] High positive reward (+10): The bird immediately flees at high speed and does not return that day;

[0075] Medium positive reward (+3): The bird is driven away, but tries again after 5 minutes;

[0076] Negative Reward (-10): The bird ignores the disturbance and lands successfully (resting phase);

[0077] These "reward" values ​​are fed back to the reinforcement learning algorithm in the central control unit 5. The core of this algorithm is maintaining a "Q-Table," which records the expected total reward as (bird species + weather + strategy). Example:

[0078] Q(Crow, Sunny Day, Strategy A) = 30;

[0079] Q(Crow, Sunny Day, Strategy B) = 80;

[0080] If strategy A fails three times in a row (-30 points), its Q value will drop rapidly. When the game scheduler queries the Q value table in the next game, it will find that the Q value of strategy B is much higher than that of strategy A. Therefore, the probability of choosing strategy B will increase dynamically in real time. In this way, the system automatically "eliminates" the strategies that the birds have adapted to and "promotes" new and effective strategies, thus continuously breaking habituation. Specific Implementation Example 3:

[0082] like Figures 1 to 4 As shown, based on the content of the above specific embodiments, the following content is further disclosed:

[0083] The core algorithm and model of this method mainly include the following:

[0084] Behavioral Phase Model (PFSM): A probabilistic finite state machine, this is an interpretable model for discretizing and probabilistically predicting the continuous flight and landing behaviors of birds. It is more suitable for low-power embedded devices than traditional black-box neural networks and can provide explicit "next move intention" signals for game schedulers. The core concepts are state definition and transition probabilities. The state definition defines bird behavior as a finite number of stages: S0: Standby (no target in the area), S1: Alert (target detected by radar, far away, no threat), S2: Approaching (target vector points towards the tower, entering the threat zone), S3: Circling / Intending to Land (target decelerates, altitude decreases, camera identifies landing attitude), S4: Landed (target stationary on crossbar 1 or insulator). S5: Escape (target vector high speed away from the tower). The key to the model is "probability". It is not just saying "the bird is in S2", but based on sensor data (speed, acceleration, flight changes, attitude) to calculate in real time: "the probability of being in S2 is 90% and the probability of moving to S3 within 3 seconds is 75%". It is used to distinguish between intrusion, resting and foraging. When the multimodal sensing unit 3 transmits data, the PFSM immediately classifies it. If the radar shows that the target hovers at low speed for a long time and the camera recognizes the head-down pecking action, the PFSM will judge it as foraging behavior. If the target approaches in a straight line at high speed and decelerates, the PFSM will judge it as intrusion / resting behavior. This "phase prediction" is crucial and is the input signal to trigger the next step of "game scheduler".

[0085] The core of PFSM is to define a set of behavioral states S and a transition probability matrix P between states. The behavioral state set S is:

[0086] S = S 警戒 ,S 接近 ,S 意图着陆 ,S 觅食 ,S 逃逸 ,S 停歇 ;

[0087] Input features F: Based on the input of the multimodal sensing unit 3, including the bird's three-dimensional coordinates X, velocity vector V, acceleration A, and attitude features P. pose and species type C spec ;

[0088] Core output: Current state S t Next, transition to the next state S. t+1 The probability P(S) t+1 |S t ,F t );

[0089] PFSM uses Bayesian inference to update the state. The model is trained on historical data to determine each state S. iThe characteristic probability distribution P(F|S) i State transition probability formula:

[0090] P(S t+1 |F t )∝P(F t |S t+1 )·P(S t+1 |S t )

[0091] P(S t+1 |F t ): In the current observation feature F t The target is now in the next state S. t+1 The posterior probability is the final prediction result of the central control unit 5;

[0092] P(F t |S t+1 Likelihood function: represents the likelihood function if the target is in state S. t+1 State, current feature F observed t The probability, that is, if S t+1 If it is an "intended landing", then its V is close to zero and its attitude feature P pose The probability of it including claws is very high;

[0093] P(S t+1 |S t Prior probability (or transition matrix): represents the target's transition from state S. t Naturally transferred to S t+1 The probability of this matrix reflects the general behavioral habits of birds;

[0094] The core function of PFSM is to identify the "intended landing" phase, once P(S 意图着陆 |F t Exceeding the set threshold θ action The system immediately triggers the game scheduler.

[0095] The game scheduler is the system's strategy decision-making center. Its sole purpose is to break the birds' habitual behavior. Essentially, it's a strategy library manager based on game theory, with "non-periodicity" and "strategy randomization" as its core principles. Birds are highly intelligent animals; if they are repeatedly subjected to the same stimulus, they will quickly learn to ignore it. The game scheduler prevents birds from "learning the pattern." When the Behavioral Phase Model (PFSM) reports "Target is about to enter S3: Intended landing," the game scheduler will not always execute the same action. Its strategy library contains multiple "scripts": Script A (Standard): Electromagnetic interference + polarized light; Script B (Deterrence): High-frequency RF pulse + polarized light. +Directional ultrasound; Script C (Deception): Inaction, allowing birds to land; Script D (Punishment): After a 2-second delay following bird landing (entering S4), suddenly activate Script B. The game scheduler selects the script to execute based on a pseudo-random algorithm. Each time a deportation task is triggered, the scheduler does not always choose the best strategy, but rather selects a strategy pseudo-randomly based on the weights assigned by the reinforcement learning module. This non-periodic decision-making pattern is key to preventing birds from developing habits and adaptations. It does not run complex learning algorithms, but is responsible for parsing and executing the output of the reinforcement learning (RL) optimizer. Input: PFSM prediction result P(S) 意图着陆 ), current environmental state E (e.g., species C) spec The policy value table Q generated by the weather (W) and the RL optimizer is used to determine the number of executable removal policies A = A1, A2, ..., A n In this process, based on the value of the Q-table, the optimal strategy A is selected in a non-periodic, randomized but biased manner. * ;

[0096] The game scheduler uses the ∈-Greedy strategy to find a balance between exploration (random strategy) and exploitation (optimal strategy) to achieve aperiodicity;

[0097]

[0098] A * The optimal strategy to be executed in this expulsion mission (i.e., the expulsion scenario, including the combination and timing of SP5 and SP6);

[0099] Q(E,A): The long-term expected value (Q value) of executing policy A in environment state E, provided by the reinforcement learning optimizer;

[0100] argmax A∈A Q(E,A): Select the strategy with the highest current Q value (exploitation);

[0101] ∈: Exploration rate. ∈ is a value between 0 and 1. When the probability is ∈, the system will randomly select a strategy (exploration) to continuously test the bird's adaptability and prevent it from becoming habitual.

[0102] Reinforcement Learning (RL) Optimizer: This is the system's adaptive learning engine, used to evaluate the long-term effectiveness of the game scheduler's policies and dynamically adjust its policy selection weights. The core of reinforcement learning is the reward mechanism and Q-value iteration. Each expulsion attempt is regarded as a "training" exercise. By assigning rewards or penalties to the results, the long-term value (Q-value) of each policy in the scheduler is iteratively updated. Here, the state is obtained by the reinforcement learning model, i.e. (species: crow, phase: S3, weather: sunny); the action is selected by the game scheduler, i.e., script D (penalty) is executed. Feedback: After the system executes the action, the multimodal perception unit 3 immediately observes the result; Reward: The system gives a reward based on the result: Reward +10: The bird is frightened and enters S5: Escape state; Reward -1: The bird ignores the interference and the state remains unchanged; Reward -5: The bird successfully lands and enters S4: Landed state. The reinforcement learning algorithm updates its internal "Q table" based on this (state, action, reward) triple. If the system finds that for the species "crow", the long-term total reward brought by scenario D is much higher than that of other scenarios, it will increase the probability of the game scheduler choosing scenario D next time.

[0103] The model type is the Q-Learning algorithm (or a variant thereof), and the input is the policy A executed by the game scheduler. t Current environmental status E t and the reward value R after execution t Iteratively update the Q table to ensure that when the scheduler encounters a similar state E again, it can choose the non-periodic strategy with the highest long-term reward.

[0104] The reinforcement learning optimizer updates the Q-value of each state-action pair using an iterative form of the following Bellman equation:

[0105] Q(E t A t )←Q(E t A t )+α[R t +γmax A Q(E t+1 ,A)-Q(E t A t )]

[0106] Q(E t A t ): In state E t The following strategy A will be adopted. t The current expected value;

[0107] α: Learning rate, which determines the extent to which new information covers old information. The larger α is, the faster the system adapts to new situations.

[0108] R t Instant reward: Given immediately after the multimodal perception unit 3 determines the expulsion result; successful expulsion (transfer to S) is rewarded. 逃逸 R t =+10; Expulsion failed (transferred to S) 停歇 )R t =-5:

[0109] γ: Discount factor, which determines the importance of future rewards. The closer γ is to 1, the more the system values ​​the long-term disengagement effect and avoids pursuing only immediate success.

[0110] max A Q(E t+1 A): Next state E t+1 The maximum expected value of all possible strategies;

[0111] Through continuous iteration, the system has learned to select which electromagnetic-visual cooperative strategy (A) under different environmental conditions (species, weather, historical adaptability). t This allows for the optimization of strategies in Sp7 based on bird responses, maximizing the long-term deterrent effect Q. Specific Implementation Example 4:

[0113] like Figures 1 to 4 As shown, based on the content of the above specific embodiments, the following content is further disclosed:

[0114] To further verify the feasibility of the technical solution in this application, the following case study is provided:

[0115] Case 1: "Game Theory and Delayed Punishment" among Highly Intelligent Individuals;

[0116] Scenario: At dusk, an adult crow (highly intelligent, exhibiting probing behavior) needs to be prevented from resting, and a long-term deterrent needs to be established to break its "probing-adaptation" cycle.

[0117] Sp1: Data Acquisition and Early Warning: 16:30:01, The frequency-modulated continuous wave radar of the multimodal sensing unit 3 detected a target with v>2m / s in sleep mode and woke up the system. 16:30:02, The radar locked onto the target and provided three-dimensional coordinates and velocity vectors. The visual camera was activated, locked onto the target and performed species identification: {Species: Crow, Confidence: 98%}.

[0118] Sp2: Modeling and Intent Prediction: 16:30:05, the crow slows down and hovers above bar 1; trajectory analysis shows it is observing. 16:30:08, the crow suddenly swoops down; trajectory extrapolation calculates the estimated landing point is directly above neural interference node 7; the camera recognizes the "claws outstretched" posture. 16:30:09, Behavioral Phase Model (PFSM) determination: P (transfer to S). 意图着陆 ) = 97%;

[0119] Sp3: Game Scheduler Decision: The game scheduler in the central control unit 5 is triggered. The system queries the reinforcement learning (RL) Q-value table. The record shows that the crows in this area have become desensitized to the "standard magnetic field interference" (strategy A) (Q value = 2.1, which is low). In order to break the habituation, the scheduler does not choose strategy A and starts the "exploration" mode of the Epsilon greedy algorithm, or chooses a "non-standard" strategy with a high Q value: "strategy D: deception and delayed penalty". The system sends a delayed trigger instruction packet to node 7.

[0120] Sp4-Sp6: Collaborative Execution (Deception and Punishment): 16:30:10 (T=0.0s), the crow lands, the system deliberately does nothing (deception), 16:30:12 (T=2.0s), the crow feels safe and begins to preen its feathers, 16:30:12.1 (T=2.1s), punishment is initiated! Neural interference node 7 strictly executes the delayed instruction: the miniature electromagnetic field emission array 8 instantly emits RF radio frequency pulses at maximum power, the crow's claws (nerve endings) feel a strong "static electricity" or "numbness", the directional polarized light projector 9 simultaneously projects a randomly rotating polarized pattern at the highest refresh rate (e.g., 50Hz), the horizontal bar in the crow's eyes instantly "liquefies", the directional ultrasonic pulser 10 simultaneously emits a focused sound beam, the crow, in its most relaxed state, simultaneously encounters three incomprehensible attacks from its feet, its eyes, and its body, this "betrayal" establishes extremely strong cognitive confusion and fear, the crow flies away in terror;

[0121] Sp7: Feedback and Reinforcement Learning: 16:30:13, Multimodal Perception Unit 3 confirms the target enters S 逃逸 Phase, system record: (State: Raven, S) 停歇 (Sunny day) + (Action: Policy D) -> (Reward: +15), the reinforcement learning algorithm greatly improved the Q value of "Policy D: Delayed Penalty" for crows. The system "learned" that "pretending to be weak and then launching a surprise attack" is far more effective than "driving them away immediately" when dealing with smart crows.

[0122] Case Study 2: Dynamic Herding and Collaborative Guidance of Group Goals;

[0123] Scenario: In the morning, a flock of pigeons (about 30 birds, exhibiting group behavior) needs to be guided to fly away in an orderly manner to prevent them from scattering due to fright (which could lead to them hitting the line).

[0124] SP1: Data Acquisition and Early Warning: 09:15:01, Frequency Modulated Continuous Wave Radar detects a large area, low density RCS signal, which is identified as a "group target". The system is activated. 09:15:02, Visual Camera confirms {Species: Pigeon Flock}.

[0125] Sp2: Modeling and Intent Prediction: 09:15:05, the central control unit 5 does not calculate the "estimated landing point", but instead calculates the "centroid" and "group velocity vector" of the flock. 09:15:06, the trajectory prediction shows that the flock will cross the entire crossbar 1 from left (node ​​5) to right (node ​​9). PFSM decision: P (transfer to S_intrusion / group).

[0126] Sp3: Game Scheduler Decision: The scheduler is triggered, and the Q-value table shows that for (state: pigeon flock, S_intrusion), the strategy with the highest Q value is "strategy F: dynamic movement barrier". The scheduler works with the communication synchronization module 4 to generate a series of phase codes with time differences, ready to execute the "dynamic movement effect" in Sp4. The instructions are pre-issued to all relevant nodes (5,6,7,8,9) on the bottom of the horizontal bar 1.

[0127] Sp4-Sp6: Cooperative Execution (Dynamic Moving Barrier): 09:15:10 (T=0.0s), the "leader bird" of the flock approaches the defense zone of the fifth node. T=0.0s: the miniature electromagnetic field emission array 8 of the fifth node is activated, emitting ELF extremely low frequency pulses. The lead bird's geomagnetic navigation system is instantly overwhelmed by "noise". T=0.2s: the fifth node is turned off, and the sixth node activates the ELF pulse. T=0.4s: the sixth node is turned off, and the seventh node activates the ELF pulse. T=0.6s: the seventh node is turned off, and the eighth node activates... From the perspective of the flock of pigeons, they are flying towards an "invisible wall that interferes with navigation and is moving to the right at high speed". The lead bird's instinct is to avoid this moving barrier, so they collectively turn outward.

[0128] Sp7: Feedback and Reinforcement Learning: 09:15:13, Multimodal Perception Unit 3 confirms that the "group velocity vector" of the entire flock has turned, deviating from the tower and entering S. 逃逸 Phase, system record: (State: pigeon flock, S_intrusion) + (Action: policy F) -> (Reward: +10). The reinforcement learning algorithm confirms that "policy F: dynamic moving barrier" is the optimal, low-energy, and high-safety (without causing panic among the flock) policy for dealing with group targets. Its Q value is consolidated. The system has successfully achieved a high-tech "bird herding" rather than a crude "bird driving".

[0129] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0130] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A bird-directional deterrent device for power transmission lines, comprising a crossbar (1), characterized in that: The surface of the crossbar (1) is provided with a central control box (2), the interior of the central control box (2) is provided with a communication synchronization module (4), a central control unit (5) and an energy storage module (6), the surface of the crossbar (1) is provided with a multimodal sensing unit (3), the bottom surface of the crossbar (1) is provided with a neural interference node (7), the surface of the neural interference node (7) is provided with a miniature electromagnetic field emission array (8) and a directional polarized light projector (9), the bottom surface of the neural interference node (7) is provided with a directional ultrasonic pulser (10), and the surface of the crossbar (1) is provided with an electromagnetic induction pickup (11).

2. The bird-directional deterrent device for transmission lines according to claim 1, characterized in that: The multimodal sensing unit (3) includes a frequency-modulated continuous wave radar and a visual camera, used to provide the bird's three-dimensional coordinates, velocity, and species classification.

3. The bird-directional deterrent device for transmission lines according to claim 1, characterized in that: The micro electromagnetic field emitting array (8) adopts an orthogonal coil array structure to precisely control the frequency, intensity and direction of the emitted magnetic field and radio frequency pulses. The frequency range includes very low frequency and extremely low frequency.

4. The bird-directional deterrent device for transmission lines according to claim 1, characterized in that: The directional polarized light projector (9) uses a narrow wavelength light source with a rapidly switchable polarization angle to project a randomly rotating and switching polarization pattern onto the bird's estimated landing area.

5. The bird-directional deterrent device for transmission lines according to claim 1, characterized in that: The electromagnetic induction pickup (9) is provided with a power transmission line on one side, and it maintains a safe gap with the power transmission line to achieve non-contact electromagnetic energy pickup.

6. The method corresponding to the bird-directional repelling device for transmission lines according to any one of claims 1-5, characterized in that: The method includes the following steps: Sp1: The location, speed, flight trajectory and estimated landing point of the bird flock are collected through the multimodal sensing unit (3); Sp2: The central control unit (5) runs a behavioral phase model, abstracts bird behavior into a phase sequence and predicts its navigation intention; Sp3: The central control unit (5) runs the game scheduler, generates aperiodic electromagnetic field interference patterns and spatiotemporal illusion sequences according to the navigation intention, and outputs phase codes; Sp4: Drives multiple distributed neural interference nodes in a coordinated manner based on phase codes; Sp5: The neural interference node (7) drives the micro electromagnetic field emission array (8) to emit extremely low-power non-periodic magnetic field pulses and radio frequency pulses to the estimated landing point and group targets, interfering with the birds' geomagnetic navigation or neural perception. Sp6: The neural interference node (7) simultaneously drives the directional polarized light projector (9) to project a high refresh rate random rotating polarization pattern, combined with directional acoustic pulses, to create a spatial illusion to prevent birds from landing; Sp7: Based on birds' responses to electromagnetic pulse frequency and polarization pattern switching rate, the strategy is adjusted in real time using reinforcement learning algorithms to continuously break birds' habituation.

7. The method corresponding to the bird-directional deterrence device for transmission lines according to claim 6, characterized in that: In the Sp4, the central control unit (5) drives multiple neural interference nodes (7) to start according to a predetermined phase sequence and time difference through phase code, so that the electromagnetic field pulse generates a dynamic movement effect in space, forming a movable repelling barrier, and realizing the orderly guidance of the bird flock.

8. The method corresponding to the bird-directional repelling device for transmission lines according to claim 6, characterized in that: The behavior phase model in the central control unit (5) adopts a probabilistic finite state machine to distinguish between the intrusion behavior, resting behavior and foraging behavior of bird flocks, and to provide state input for the game scheduler.

9. The method corresponding to the bird-directional deterrence device for transmission lines according to claim 6, characterized in that: In Sp5, the essential function of the extremely low-power aperiodic magnetic field pulses and radio frequency pulse interference is to disrupt birds' accurate navigation and positioning perception of the geomagnetic field. The pulse intensity is strictly controlled within a threshold that is harmless and does not interfere with communication.

Images