A method and system for preventing small animals from climbing power lines and tripping breakers

Abstract

The present application belongs to the technical field of power transmission protection, and relates to a protection method and system for preventing small animals from climbing and tripping of a power collection line, comprising: collecting multi-source data; performing weighted fusion on the multi-source data to obtain a target feature vector; obtaining a target type based on the target feature vector through a trained intelligent recognition model; generating a corresponding driving strategy according to the target type and outputting the driving strategy to a corresponding execution mechanism for driving. The present application obtains a target feature vector by performing weighted fusion on multi-source data, obtains a target type based on the target feature vector through a trained intelligent recognition model, realizes accurate identification of the target type, improves the identification accuracy, generates a corresponding driving strategy according to the target type, outputs the driving strategy to a corresponding execution mechanism for driving, realizes multi-dimensional driving, improves the driving effect, and solves the problems of low target type identification accuracy and poor driving effect.

Description

Technical Field

[0001] This invention relates to the field of power transmission protection technology, and specifically discloses a protection method and system for preventing small animals from climbing and tripping power lines. Background Technology

[0002] With the continuous growth of wind power installed capacity, 35kV collection lines play a crucial role in power transmission, especially in mountainous wind farms, such as loess plateaus and hilly areas at altitudes of 1300m-1400m. Due to the unique ecological environment and widespread crop distribution, these areas easily attract birds such as sparrows and pigeons to perch on the collection line towers, which in turn attracts rodents (such as weasels and rats), snakes, and other small animals. These small animals are mainly active at night, and when they climb up the towers to the crossarms of the terminal towers to forage for food, they may not maintain a safe distance from exposed parts such as isolating switches and conductors, causing instantaneous grounding discharge and tripping of the line. Currently, cameras are typically used to collect detection images, and the target type (i.e., the type of small animal) is identified through these images. Different frequencies of ultrasonic waves are then selected to drive away the small animal. However, relying on a single detection image for target type identification results in low accuracy, and using a single ultrasonic wave for repelling leads to poor repelling effect. Summary of the Invention

[0003] The purpose of this invention is to provide a protection method and system for preventing small animals from climbing and tripping power lines, thereby solving the problems of low accuracy in target type identification and poor deterrence effect in existing methods.

[0004] The specific solution of the present invention is as follows:

[0005] A method for preventing small animals from climbing and tripping power lines includes:

[0006] Collect data from multiple sources;

[0007] Weighted fusion of multi-source data yields the target feature vector;

[0008] Based on the target feature vector, the target type is obtained through a trained intelligent recognition model;

[0009] Generate a corresponding expulsion strategy based on the target type, and output the expulsion strategy to the corresponding execution agency for expulsion.

[0010] Preferably, the collection of multi-source data includes:

[0011] The visual sensing unit monitors in real time whether a target enters the monitoring range. If no target is detected, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are not activated.

[0012] If the detection is positive, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are activated, and the collected visual, auditory, olfactory, and tactile data are transmitted to the protection control module. When the collection time reaches the preset time threshold, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are deactivated.

[0013] Preferably, the visual perception unit includes:

[0014] Infrared thermal imaging sensors and microwave radar sensors;

[0015] The system detects heat sources using infrared thermal imaging sensors and movement using microwave radar sensors. When both heat sources and movement are detected simultaneously, it is determined that a target has entered the monitoring range, and visual data is obtained. Otherwise, it is determined that no target has entered the monitoring range, and monitoring continues.

[0016] Preferably, the target feature vector includes:

[0017] Multi-source data includes visual data, auditory data, olfactory data, and tactile data;

[0018] Based on visual data, Doppler features of microwave radar data are extracted using short-time Fourier transform, and temperature features of thermal imaging images are extracted using image segmentation methods to generate visual feature sub-vectors. Based on auditory data, voiceprint features are extracted using Mel-frequency cepstral coefficients to obtain auditory feature sub-vectors. Based on olfactory data, odor features are extracted using principal component analysis to obtain olfactory feature sub-vectors. Based on tactile data, vibration features are extracted using wavelet transform to generate tactile feature sub-vectors.

[0019] Based on the visual feature vectors, auditory feature vectors, olfactory feature vectors, tactile feature vectors, and preset visual weights, auditory weights, olfactory weights, and tactile weights, the target feature vector is obtained by linear fusion through a weight adjustment matrix.

[0020] Preferably, the formula for calculating the target feature vector is:

[0021] ,

[0022] in, For the target feature vector, For visual feature sub-vectors, For auditory feature subvectors, For olfactory feature sub-vectors, For tactile feature sub-vectors, For visual weight, For auditory weight, For olfactory weight, For tactile weight, For feature cascade.

[0023] Preferably, the training of the intelligent recognition model uses a reward function, the configuration of which includes:

[0024] When the target type output by the intelligent recognition model is completely correct, the reward score is the first reward value;

[0025] When the target type output by the intelligent recognition model is completely wrong, the penalty score is the first penalty value;

[0026] When the target type output by the intelligent recognition model is neither completely correct nor completely wrong, if the target type is correctly identified, the reward score is the second reward value; if the target type is incorrectly identified, the penalty score is the second penalty value; if the target type is missed, the penalty score is the third penalty value.

[0027] The first reward value is greater than the second reward value, the first penalty value is greater than the second penalty value, and the second penalty value is greater than the third penalty value.

[0028] Preferred driving strategies include:

[0029] Target types include birds, snakes, and rodents;

[0030] If the target is a bird, use eagle calls, red and blue flashing lights, and barbs to scare it away;

[0031] If the target is a snake, use ultrasound, gas, or barbs to drive it away;

[0032] If the target is a rodent, use ultrasound, wolf howls / cat meows, gas, red and blue flashing lights, or barbs to drive it away.

[0033] Preferably, it further includes a preprocessing step, which includes:

[0034] Perform time alignment, outlier removal, and standardization on multi-source data.

[0035] This invention also relates to a protection system for preventing small animals from climbing and tripping power lines, used to implement the above-mentioned protection method for preventing small animals from climbing and tripping power lines, comprising:

[0036] The system comprises an energy module, a sensing module, a protection and control module, and an actuator. The actuator includes a barb assembly, an ultrasonic generator, a sound simulator, a strobe LED light, and an odor generator. The protection and control module is electrically connected to the energy module, the sensing module, the ultrasonic generator, the sound simulator, the strobe LED light, and the odor generator, respectively.

[0037] The perception module includes a visual perception unit, an auditory perception unit, an olfactory perception unit, and a tactile perception unit;

[0038] The protection and control module includes a preprocessing unit, a multi-source fusion unit, a target type identification unit, and a drive-away strategy generation unit.

[0039] Preferably, the energy module includes:

[0040] When the power provided by the solar photovoltaic modules cannot meet the load demand, the power supply is switched to the energy storage battery modules.

[0041] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0042] This invention reduces energy consumption and saves energy through layered sensing of the sensing module; it obtains target feature vectors by weighted fusion of multi-source data, and obtains target types based on the target feature vectors through a trained intelligent recognition model, achieving accurate target type recognition and improving recognition accuracy; it improves recognition accuracy by training the intelligent recognition model through a reward function, and encourages the intelligent recognition model to maintain stability and consistency in continuous recognition tasks; it generates corresponding driving strategies according to target types, and outputs the driving strategies to the corresponding actuators for driving, realizing multi-dimensional driving through ultrasound, simulated sound, red and blue flashing lights, gas, and barbs, thereby improving the driving effect. Attached Figure Description

[0043] Figure 1 This is a flowchart of a protection method for preventing small animals from climbing and tripping power lines, according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of a protection system for preventing small animals from climbing and tripping the circuit breaker in an embodiment of the present invention.

[0045] Figure 3 This is a block diagram of the protection control module in an embodiment of the present invention.

[0046] Attached reference numerals: 1-Energy module, 2-Sensing module, 3-Protection and control module, 4-Ultrasonic generator, 5-Sound simulator, 6-Strobe LED light, 7-Odor generator. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0048] A method for preventing small animals from climbing and tripping power lines, such as... Figure 1As shown, it includes the following steps:

[0049] S1. Collect data from multiple sources;

[0050] The visual sensing unit monitors in real time whether a target enters the monitoring range. If no target is detected, the auditory, olfactory, and tactile sensing units are not activated. If a target is detected, the auditory, olfactory, and tactile sensing units are activated, and the collected visual, auditory, olfactory, and tactile data are transmitted to the protection control module 3. When the collection time reaches the preset time threshold, the auditory, olfactory, and tactile sensing units are deactivated.

[0051] Since the protection system is powered by photovoltaic energy or energy storage, the power supply is limited by weather conditions and the energy storage is limited by battery capacity. Therefore, by using trigger-type switchable auditory sensing units, olfactory sensing units, and tactile sensing units, energy can be effectively saved and the system stability and reliability can be improved.

[0052] The visual perception unit includes an infrared thermal imaging sensor and a microwave radar sensor. The infrared thermal imaging sensor generates thermal images through infrared radiation to detect the presence of heat sources. The microwave radar sensor emits microwave signals and receives the echoes reflected from targets, generating a Doppler frequency shift. This Doppler frequency shift is used to determine the presence, speed, and distance of moving targets to detect motion. Only when both a heat source and motion are detected simultaneously is it determined that a target has entered the monitoring range, and visual data is obtained; otherwise, it is determined that no target has entered the monitoring range, and monitoring continues. This dual-sensor collaborative decision-making reduces the false alarm rate in detecting target entry into the monitoring range.

[0053] Multi-source data includes visual data, auditory data, olfactory data, and tactile data; visual data includes thermal imaging images acquired by infrared thermal imaging sensors and microwave radar data acquired by microwave radar sensors; tactile data includes acceleration data acquired by MEMS accelerometers and vibration data acquired by vibration sensors.

[0054] To improve data accuracy, multi-source data is preprocessed, including time alignment, outlier removal, and standardization.

[0055] Time alignment includes: determining whether the timestamps of visual, auditory, olfactory, and tactile data are consistent according to the sampling frequency. If the determination is not consistent, then visual, auditory, olfactory, and tactile data at the same time are obtained using a linear method. If the determination is consistent, no processing is performed. Time alignment ensures the temporal consistency of visual, auditory, olfactory, and tactile data.

[0056] Outlier removal includes: Since thermal imaging images are easily affected by ambient light, equipment noise, and high-temperature objects (such as exposed metal), a grayscale thresholding method is used to remove pixels with grayscale values ​​exceeding the normal range of 0-255, thus filtering out image noise and light interference; Since microwave radar data is easily affected by insects flying by, wind blowing debris, electromagnetic interference, etc., resulting in false motion data, a Kalman filter algorithm is used to filter out transient interference signals (such as false motion data generated by insects flying by, wind blowing debris, etc.) and retain continuous and stable motion data.

[0057] Since the auditory data collected by ultrasonic sensors is easily interfered with by environmental noise such as wind noise, vehicle horn noise, and ground vibration transmission noise, low-pass filtering algorithm and high-pass filtering algorithm are used to filter the auditory data for noise.

[0058] Because olfactory data is easily affected by environmental factors such as the smell of fertilizers in farmland, industrial exhaust gas, and dampness from rainwater, the range anomaly removal method is used to remove abnormal data outside the range.

[0059] Since MEMS accelerometers and vibration sensors are susceptible to environmental vibration interference, a wavelet transform algorithm is used to decompose the acceleration and vibration data, remove the acceleration and vibration data corresponding to environmental vibration interference, and retain the small acceleration and vibration data generated by the small animal climbing.

[0060] Standardization is used to eliminate differences in units between different data. Visual, auditory, olfactory, and tactile data are standardized separately. The formula for standardization is as follows:

[0061] ,

[0062] in, For standardized data, For the data before standardization, The mean of the data before standardization. This represents the standard deviation of the data before standardization.

[0063] S2. Weighted fusion of multi-source data to obtain the target feature vector;

[0064] Based on visual data, Doppler features of microwave radar data are extracted using short-time Fourier transform, and temperature features of thermal imaging images are extracted using image segmentation methods to generate visual feature sub-vectors. Based on auditory data, voiceprint features are extracted using Mel frequency cepstral coefficients to obtain auditory feature sub-vectors. Based on olfactory data, odor features are extracted using principal component analysis to obtain olfactory feature sub-vectors. Based on tactile data, vibration features are extracted using wavelet transform to generate tactile feature sub-vectors.

[0065] The visual feature subvector, auditory feature subvector, olfactory feature subvector, and tactile feature subvector are all 8-dimensional subvectors.

[0066] Based on the visual feature vector, auditory feature vector, olfactory feature vector, tactile feature vector, and preset visual weights, auditory weights, olfactory weights, and tactile weights, a 32-dimensional target feature vector is obtained by linear fusion through a weight adjustment matrix.

[0067] The formula for calculating the target feature vector is:

[0068] ,

[0069] in, For the target feature vector, For visual feature sub-vectors, For auditory feature subvectors, For olfactory feature sub-vectors, For tactile feature sub-vectors, For visual weight, For auditory weight, For olfactory weight, For tactile weight, For feature cascade.

[0070] S3. Based on the target feature vector, obtain the target type through a trained intelligent recognition model;

[0071] A large amount of historical visual, auditory, olfactory, and tactile data is acquired to generate a training set. Based on the training set, an intelligent recognition model is trained using reinforcement learning methods to obtain a well-trained intelligent recognition model, which is a multi-branch fusion convolutional neural network (CNN).

[0072] Reinforcement learning methods include reward functions, and the configuration of reward functions includes:

[0073] When the target type output by the intelligent recognition model is completely correct, the reward score is the first reward value;

[0074] When the target type output by the intelligent recognition model is completely wrong, the penalty score is the first penalty value;

[0075] When the target type output by the intelligent recognition model is neither completely correct nor completely wrong, if the target type is correctly identified, the reward score is the second reward value; if the target type is incorrectly identified (for example, rodents are identified as snakes), the penalty score is the second penalty value; if the target type is missed (for example, the target type to be identified includes rodents and birds, but the intelligent recognition model only identifies birds and not rodents), the penalty score is the third penalty value.

[0076] The first reward value is greater than the second reward value, the first penalty value is greater than the second penalty value, and the second penalty value is greater than the third penalty value.

[0077] Target types include: birds, snakes, and rodents, with rodents including mice and weasels.

[0078] For example, during the training of the intelligent recognition model, the real target types are birds and rodents. If the target types output by the intelligent recognition model are birds and rodents, it means that the intelligent recognition model has perfect recognition ability. In this case, the highest reward value, namely the first reward value, is set to strongly encourage the intelligent recognition model to maintain a perfect score. The first reward value can be 10 points.

[0079] If the target type output by the intelligent recognition model is birds or unknown other types, it means that the intelligent recognition model has recognized a part of the real target types. The intelligent recognition model has mastered the types of some targets, but it is not comprehensive enough. In this case, a second reward value lower than the first reward value is set to encourage the intelligent recognition model to continue to learn the parts that were missed. The second reward value can be 3 points.

[0080] If the target type output by the intelligent recognition model is a snake, it means that the recognition ability of the intelligent recognition model has completely failed. In this case, the highest penalty value, i.e. the first penalty value, is set to strongly suppress the behavior of the intelligent recognition model outputting completely wrong values. The first penalty value can be -10 points.

[0081] If the target type output by the intelligent recognition model is birds or snakes, it means that the intelligent recognition model has incorrectly identified the target type. In this case, a second penalty value lower than the first penalty value is set to take into account the severity of the intelligent recognition model's recognition error. The second penalty value can be -7 points.

[0082] If the target type output by the intelligent recognition model is birds, it means that the intelligent recognition model has missed recognizing rodents. In this case, a third penalty value lower than the second penalty value is set to warn the intelligent recognition model to complete the missed part. The third penalty value can be -5 points.

[0083] By setting a first reward value, the intelligent recognition model is ensured to tend to output complete and correct target types, rather than just a single category. A second reward value provides positive feedback to the intelligent recognition model when partially correct, thereby learning to complete the missing target types. Different levels of penalty values ​​help the intelligent recognition model distinguish the severity of errors and prioritize correcting the most serious errors (such as complete errors). Through repeated iterative training, the intelligent recognition model will continuously improve its target recognition accuracy under the drive of maximizing rewards and minimizing penalties. In other words, training the intelligent recognition model through a reward function can improve the recognition accuracy and generalization ability of the intelligent recognition model, and encourage the intelligent recognition model to maintain stability and consistency in continuous recognition tasks.

[0084] The target feature vector is input into the trained intelligent recognition model. The model performs feature recognition on visual, auditory, olfactory and tactile senses through different CNN branches. The features are then fused in the fully connected layer to output the target type.

[0085] By using multi-dimensional features (visual features, auditory features, olfactory features, and tactile features), the target type can be identified, thereby improving the accuracy of identification.

[0086] S4. Generate the corresponding expulsion strategy according to the target type, and output the expulsion strategy to the corresponding execution mechanism for expulsion.

[0087] The expulsion strategies include:

[0088] If the target is a bird, use eagle calls, red and blue flashing lights, and barbs to scare it away;

[0089] If the target is a snake, use ultrasound, gas, or barbs to drive it away;

[0090] If the target is a rodent, use ultrasound, wolf howls / cat meows, gas, red and blue flashing lights, or barbs to drive it away.

[0091] The gases include sulfur gas, camphor gas, etc.

[0092] This invention also relates to a protection system for preventing small animals from climbing and tripping power lines, used to implement the aforementioned protection method for preventing small animals from climbing and tripping power lines, such as... Figure 2 As shown, it includes:

[0093] The system comprises an energy module 1, a sensing module 2, a protection and control module 3, and an actuator. The actuator includes a barb assembly, an ultrasonic generator 4, a sound simulator 5, a strobe LED light 6, and an odor generator 7. The protection and control module 3 is electrically connected to the energy module 1, the sensing module 2, the ultrasonic generator 4, the sound simulator 5, the strobe LED light 6, and the odor generator 7, respectively.

[0094] The barbed assembly is used to block small animals from climbing by gripping or using their limbs for leverage. Made of 5cm long stainless steel, the assembly is corrosion-resistant and anti-aging, suitable for complex outdoor environments such as high temperatures, heavy rain, and humidity. Its barbs are densely distributed and sharpened, with a spacing of no more than 2cm between adjacent barbs, forming a physical barrier without any climbing leverage points. The barbed assembly is installed from 1m above the ground to 30cm below the first crossarm, and continuously laid along the entire length of the pole's footing, covering all gripping points. The barbed assembly is secured with wire, with a binding tensile force of no less than 50N to ensure a firm installation. Installation of the barbed assembly does not require power outages to the power supply lines, and the construction time for a single pole should not exceed 4 hours.

[0095] Energy Module 1 is used to provide electrical energy. Energy Module 1 includes solar photovoltaic modules and energy storage battery modules; the solar photovoltaic modules are used for power supply first, and when the power provided by the solar photovoltaic modules cannot meet the load demand, the system switches to the energy storage battery modules for power supply; the photoelectric conversion efficiency of the solar photovoltaic modules is not less than 18%, providing stable power for the entire system; the energy storage battery modules have a capacity of not less than 12Ah, which can ensure the normal operation of the system during 3-5 consecutive days of cloudy and rainy weather; Energy Module 1 is fixed to the crossarm of the terminal tower using a fixed bracket, and the installation location avoids bird nesting areas to reduce interference with the ecological environment.

[0096] The sensing module 2 is used to collect visual, auditory, olfactory, and tactile data. The sensing module 2 includes a visual sensing unit, an auditory sensing unit, an olfactory sensing unit, and a tactile sensing unit. The visual sensing unit includes an infrared thermal imaging sensor and a microwave radar sensor, both of which are located in the upper middle part of the tower. The auditory sensing unit includes an ultrasonic sensor, which is located in the upper middle part of the tower. The olfactory sensing unit includes a gas sensor, which is located at the guy wire. The tactile sensing unit includes a MEMS accelerometer and a vibration sensor, which are located at the guy wire.

[0097] Protection control module 3 is installed at the crossarm of the terminal tower, such as Figure 3 As shown, it includes a preprocessing unit, a multi-source fusion unit, a target type identification unit, and a repelling strategy generation unit. The preprocessing unit is used to perform time alignment, outlier removal, and standardization on visual, auditory, olfactory, and tactile data. The multi-source fusion unit is used to perform weighted fusion on multi-source data to obtain a target feature vector. The target type identification unit is used to obtain the target type based on the target feature vector and a trained intelligent identification model. The repelling strategy generation unit is used to generate a corresponding repelling strategy according to the target type.

[0098] Ultrasonic generator 4 is used to generate ultrasonic waves to repel small animals, and it is located in the middle of the tower. The operating frequency of ultrasonic generator 4 is adapted to the hearing sensitivity range of rodents and snakes. By generating vibrations and ultrasonic waves, it creates a biological deterrent, prompting small animals to actively move away from the protected area.

[0099] Sound simulator 5 is used to generate simulated sounds, including eagle cries, wolf howls, and cat meows, and it is located in the middle of the tower.

[0100] The strobe LED 6 is used to generate red and blue strobe lights, and it is installed at the crossarm of the terminal tower.

[0101] Odor generator 7 is used to release gases with specific odors, including sulfur gas, camphor gas, etc., and it is installed at the bottom of the tower.

[0102] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for preventing small animals from climbing and tripping power lines, characterized in that, include: Collect data from multiple sources; Weighted fusion of multi-source data yields the target feature vector; Based on the target feature vector, the target type is obtained through a trained intelligent recognition model; Generate a corresponding expulsion strategy based on the target type, and output the expulsion strategy to the corresponding execution agency for expulsion.

2. The method for preventing small animals from climbing and tripping power lines according to claim 1, characterized in that, The collection of multi-source data includes: The visual sensing unit monitors in real time whether a target enters the monitoring range. If no target is detected, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are not activated. If the detection is positive, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are activated, and the collected visual, auditory, olfactory, and tactile data are transmitted to the protection control module. When the collection time reaches the preset time threshold, the auditory sensing unit, olfactory sensing unit, and tactile sensing unit are deactivated.

3. The method for preventing small animals from climbing and tripping power lines according to claim 2, characterized in that, The visual perception unit includes: an infrared thermal imaging sensor and a microwave radar sensor; The system detects heat sources using infrared thermal imaging sensors and movement using microwave radar sensors. When both heat sources and movement are detected simultaneously, it is determined that a target has entered the monitoring range, and visual data is obtained. Otherwise, it is determined that no target has entered the monitoring range, and monitoring continues.

4. The method for preventing small animals from climbing and tripping power lines according to claim 1, characterized in that, The obtained target feature vector includes: The multi-source data includes visual data, auditory data, olfactory data, and tactile data; Based on visual data, Doppler features of microwave radar data are extracted using short-time Fourier transform, and temperature features of thermal imaging images are extracted using image segmentation methods to generate visual feature sub-vectors. Based on auditory data, voiceprint features are extracted using Mel-frequency cepstral coefficients to obtain auditory feature sub-vectors. Based on olfactory data, odor features are extracted using principal component analysis to obtain olfactory feature sub-vectors. Based on tactile data, vibration features are extracted using wavelet transform to generate tactile feature sub-vectors. Based on the visual feature vectors, auditory feature vectors, olfactory feature vectors, tactile feature vectors, and preset visual weights, auditory weights, olfactory weights, and tactile weights, the target feature vector is obtained by linear fusion through a weight adjustment matrix.

5. A method for preventing small animals from climbing and tripping power lines according to claim 4, characterized in that, The formula for calculating the target feature vector is: ， in, For the target feature vector, For visual feature sub-vectors, For auditory feature subvectors, For olfactory feature sub-vectors, For tactile feature sub-vectors, For visual weight, For auditory weight, For olfactory weight, For tactile weight, For feature cascade.

6. The method for preventing small animals from climbing and tripping power lines according to claim 1, characterized in that, The training of the intelligent recognition model uses a reward function, the configuration of which includes: When the target type output by the intelligent recognition model is completely correct, the reward score is the first reward value; When the target type output by the intelligent recognition model is completely wrong, the penalty score is the first penalty value; When the target type output by the intelligent recognition model is neither completely correct nor completely wrong, if the target type is correctly identified, the reward score is the second reward value; if the target type is incorrectly identified, the penalty score is the second penalty value; if the target type is missed, the penalty score is the third penalty value. The first reward value is greater than the second reward value, the first penalty value is greater than the second penalty value, and the second penalty value is greater than the third penalty value.

7. A method for preventing small animals from climbing and tripping power lines according to claim 1, characterized in that, The expulsion strategy includes: The target types include birds, snakes, and rodents; If the target is a bird, use eagle calls, red and blue flashing lights, and barbs to scare it away; If the target is a snake, use ultrasound, gas, or barbs to drive it away; If the target is a rodent, use ultrasound, wolf howls / cat meows, gas, red and blue flashing lights, or barbs to drive it away.

8. The method for preventing small animals from climbing and tripping power lines according to claim 1, characterized in that, It also includes a preprocessing step, which includes: time alignment of multi-source data, removal of outliers, and standardization.

9. A protection system for preventing small animals from climbing and tripping power lines, characterized in that, A method for preventing small animals from climbing and tripping a power line according to any one of claims 1-8 includes: The system comprises an energy module, a sensing module, a protection and control module, and an actuator. The actuator includes a barb assembly, an ultrasonic generator, a sound simulator, a strobe LED light, and an odor generator. The protection and control module is electrically connected to the energy module, the sensing module, the ultrasonic generator, the sound simulator, the strobe LED light, and the odor generator, respectively. The sensing module includes a visual sensing unit, an auditory sensing unit, an olfactory sensing unit, and a tactile sensing unit; The protection and control module includes a preprocessing unit, a multi-source fusion unit, a target type identification unit, and a drive-away strategy generation unit.

10. A protection system for preventing small animals from climbing and tripping power lines according to claim 9, characterized in that, The energy module includes a solar photovoltaic module and an energy storage battery module. When the power provided by the solar photovoltaic module cannot meet the load demand, the power supply is switched to the energy storage battery module.

Images