Anti-tolerance bird repelling method based on online evaluation
By constructing a bird deterrence strategy library and a three-dimensional indicator system, and combining reinforcement learning and analytic hierarchy process, the bird deterrence strategy was optimized online, solving the problems of bird tolerance and strategy adaptability, and improving the immediate success rate and long-term effectiveness of bird deterrence.
Patent Information
- Application Number
- CN202511319474.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-16
- Publication Date
- 2026-01-02
AI Technical Summary
Existing bird control technologies face challenges such as bird tolerance and a lack of intelligent dynamic adaptability, resulting in a decline in bird control effectiveness over time and making it difficult to assess and adjust bird control strategies in real time in complex and ever-changing scenarios.
A bird-repelling strategy library is constructed, employing a three-dimensional index system of response level, action delay, and action distance. Combined with improved Q-learning reinforcement learning and hierarchical analysis, the bird-repelling strategies are optimized and adaptively adjusted online.
By assessing bird deterrence effectiveness in real time and intelligently switching strategies, it effectively prevents bird tolerance, improves the immediate success rate and long-term sustainability of bird deterrence, and is suitable for critical locations such as substations and airports.
Smart Images
Figure CN121242008A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of bird repelling, and particularly relates to a method for preventing bird tolerance based on online evaluation. BACKGROUND
[0002] At present, for the bird intrusion problem of places such as substations and airports, common bird repelling methods mainly include ultrasonic wave emission, laser irradiation, reflective device, chemical dispersing bird repellent and biological simulation of enemy sound and the like. In order to improve the repelling effect, more and more mixed bird repelling strategies are used in practice, that is, the above-mentioned methods are used simultaneously or alternately in order to enhance the repelling effect on birds through composite sensory stimulation.
[0003] However, the existing bird repelling technology, whether single means or mixed strategy, has certain limitations. First, the bird tolerance problem: long-term repeated use of specific, fixed mode of bird repelling means, especially single or unchanged combination, is easy to cause the target bird group to produce behavioral adaptability or habituation, so that the bird repelling effect is obviously reduced or even invalid over time. Secondly, the existing methods generally lack intelligent dynamic adaptation ability, and it is difficult to evaluate the actual effect of the current strategy in real time and effectively in the face of different types of birds and other complex and variable actual scenes, and timely and automatically adjust or switch to a more optimal bird repelling strategy combination. The current strategy selection often relies on preset schemes or manual experience, and it is difficult to achieve closed-loop optimization driven by scene perception and effect feedback, so that the bird repelling efficiency is not high and the effect is difficult to last. SUMMARY
[0004] The purpose of the present application is to provide a method for preventing bird tolerance based on online evaluation, which builds a bird repelling strategy library, executes different bird repelling strategies according to the types of birds, and builds a bird repelling evaluation system to evaluate the bird repelling effect. After executing the bird repelling, the strategy is online optimized and adjusted according to the bird repelling effect, effectively preventing the birds from developing tolerance.
[0005] In order to achieve the above purpose, the solution of the present application is:
[0006] A method for preventing bird tolerance based on online evaluation, comprising the following steps:
[0007] Step 1, establishing a comprehensive strategy library suitable for the expulsion of different birds;
[0008] Step 2, building a bird repelling effect comprehensive evaluation system including response level, action time delay and action distance three-dimensional indicators, for evaluating the effectiveness of the bird repelling strategy;
[0009] Step 3, executing the bird repelling strategy in the comprehensive strategy library, and evaluating the effectiveness of the bird repelling strategy according to the bird repelling effect comprehensive evaluation system;
[0010] Step 4: Based on the evaluation results, determine whether bird deterrence has been completed. If bird deterrence is completed, the process ends. If bird deterrence is not completed, execute the online update algorithm for the bird deterrence strategy, update the bird deterrence strategy library, and return to Step 3 to execute bird deterrence again. Repeat this process until the bird deterrence task is completed.
[0011] In step 1 above, the comprehensive strategy library includes different bird species that need to be driven away, as well as bird-repelling methods formed by single or combined methods; wherein, single bird-repelling methods include ultrasound, laser, predator calls, visual intimidation and chemical substances.
[0012] In step 2 above, the response level includes three levels: "startled," "restless," and "no response." "Startled" means that the birds being driven away clearly fly away from their original location under the current bird-driving strategy, indicating that the bird-driving is successful. "Restless" means that the birds being driven away hover and jump around in their original location under the current bird-driving strategy, indicating that the bird-driving is unsuccessful. "No response" means that the birds being driven away do not show any significant behavioral changes under the current bird-driving strategy, indicating that the bird-driving is unsuccessful.
[0013] The delay in action is the time interval from the application of bird deterrence measures to the birds' response;
[0014] The effective distance is the maximum distance at which the bird deterrent device can effectively drive away birds under the current bird deterrent strategy.
[0015] In step 2 above, three-dimensional indicators of bird deterrence strategies are obtained, and then the three-dimensional indicators are standardized to obtain the entropy value of each indicator. Based on the entropy value of each indicator, an objective weight is obtained to evaluate the effectiveness of the bird deterrence strategy.
[0016] In step 3 above, a bird deterrent device is used to identify birds, a bird deterrent strategy currently set for that bird is selected from the comprehensive strategy library, bird deterrent work is carried out based on the bird deterrent strategy, and the bird deterrent effect is judged according to the comprehensive evaluation system of bird deterrent effect.
[0017] In step 4 above, the online update algorithm for the bird-repelling strategy is executed, including:
[0018] An improved Q-learning reinforcement learning method was used to adjust the weights of the comprehensive strategy library, and the weight allocation was continuously optimized according to the actual situation of bird deterrence, so as to form a comprehensive bird deterrence strategy for different bird species and flock sizes.
[0019] The comprehensive weight is calculated by combining subjective and objective weights using the analytic hierarchy process, and the existing strategies are evaluated based on the superior-inferiority distance method to select the most suitable bird-repelling strategy.
[0020] Wherein, the improved Q-learning reinforcement learning method is used to adjust the weight of the comprehensive strategy library, and the weight distribution is continuously optimized according to the actual bird driving situation, forming a comprehensive bird driving strategy for different bird species and bird group size, including,
[0021] Step a1, initialize the Q-learning algorithm parameters, including the state set S of the bird driving strategy and the corresponding behavior set A, the discount factor γ; wherein the state set S contains bird information, and the behavior set A is different mixed bird driving strategy;
[0022] Step a2, select a group of states s∈S and actions a∈A, execute the bird driving strategy to obtain the next time state s';
[0023] Step a3, according to the feedback of the bird driving effect of the bird driving executor, the dynamic learning rate α(r,β) is calculated, and the related calculation expression is Wherein β is the growth coefficient, r is the reward mean, σ r Is the reward standard deviation, and ε is a small constant to prevent division by zero;
[0024] Step a4, get the next behavior a' and update the bird driving effect state value;
[0025] Step a5, judge whether the algorithm meets the convergence condition or the iteration number has reached the upper limit;
[0026] After training, the bird driving strategy is selected according to the state.
[0027] Wherein, the comprehensive weight is calculated by combining the analytic hierarchy process with subjective and objective weight, and the existing strategy is evaluated based on the distance method of superior and inferior solutions, so as to select the most suitable bird driving strategy, including,
[0028] A hierarchical structure including scheme layer, decision layer and execution layer is established;
[0029] The calibration method is used to construct the judgment matrix X, which compares the importance of the bird driving strategy parameters in the same level horizontally, and verifies the matrix consistency;
[0030] According to the judgment matrix X, the comprehensive weight of the bird driving strategy parameters is obtained;
[0031] In the distance method of superior and inferior solutions, the Q value corresponding to the bird driving strategy parameters is a maximum type index; assuming that there are n bird driving strategies, each bird driving strategy has m parameters, and the decision matrix X=[x ij ] n×m The corresponding normalized decision matrix is Y=[y ij ] n×m The calculation formula is:
[0032]
[0033] Suppose that the Q-value of the j-th parameter of the i-th bird-repelling strategy is close to the Q-value of the optimal strategy parameter. The degree of closeness to the Q value of the worst policy parameter is Then there is,
[0034]
[0035] In the formula, and These are the maximum and minimum values of the j-th column in the standardized decision matrix Y, i.e., the optimal solution and the worst solution;
[0036] Using parameter C i Let represent the degree of similarity between the i-th bird-repelling strategy and the optimal strategy, calculated using the following formula:
[0037]
[0038] C i The larger the value, the better the bird-repelling strategy; the bird-repelling strategies are ranked by proximity to obtain the optimal strategy.
[0039] Among them, verifying the consistency of the judgment matrix X includes,
[0040] The consistency index CR of the judgment matrix X is calculated according to the following formula.
[0041]
[0042] Where, λ max (X) is the largest eigenvalue of the judgment matrix X, n is the order of the judgment matrix X, and RI is the random consistency index.
[0043] If the consistency index CR is greater than the set threshold, the judgment matrix X is considered to be inconsistent and the judgment matrix X is reconstructed.
[0044] Among them, based on the judgment matrix X, the comprehensive weights of the bird-repelling strategy parameters are obtained, including:
[0045] The eigenvector corresponding to the largest eigenvalue is normalized to obtain the subjective weight, and the objective weight of each bird-repelling strategy is determined based on the response effect of birds to different bird-repelling strategies.
[0046] By allocating subjective and objective weights using specific proportional coefficients, the comprehensive weights of the bird deterrence strategy parameters are obtained.
[0047] The significant advantages of this invention, achieved by adopting the above scheme, are as follows: By constructing a comprehensive bird-repelling strategy library and a three-dimensional bird-repelling effectiveness evaluation system based on response level, action delay, and action distance, and integrating an improved reinforcement learning algorithm for online strategy exploration and weight optimization, combined with the analytic hierarchy process (AHP) and the superior-inferiority distance method for comprehensive strategy evaluation and optimization, a closed-loop feedback and dynamically adjusted adaptive online decision-making mechanism is formed. This invention can evaluate the effectiveness of bird-repelling strategies in real time. When signs of bird tolerance are detected, it intelligently explores, filters, and switches to more effective alternative or combined strategies, thereby effectively suppressing bird tolerance. This improves the immediate success rate and long-term sustainability of bird repelling, making it particularly suitable for critical locations such as substations and airports that require efficient and sustained bird repelling. Attached Figure Description
[0048] Figure 1 This is an overall flowchart of the present invention;
[0049] Figure 2 This is a schematic diagram of the comprehensive evaluation structure of bird deterrence strategies in the case of sound and light bird deterrence.
[0050] Figure 3 This is a diagram illustrating the closed-loop execution mechanism of the bird deterrence algorithm. Detailed Implementation
[0051] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.
[0052] This invention provides a specific implementation of a bird tolerance prevention and control method based on online evaluation, addressing the bird control needs of substations. (In conjunction with...) Figure 1 The implementation of this invention mainly includes the following steps:
[0053] Step S1: Based on different bird species, establish a comprehensive bird control strategy library applicable to common substation birds such as sparrows, magpies, and crows, and bird control methods, including single or combined bird control methods such as predator calls, visual intimidation, chemical substances, ultrasound, and lasers.
[0054] Step S2: Construct a comprehensive evaluation system for bird deterrence effectiveness that includes three-dimensional indicators: response level, effect delay, and effect distance, to evaluate the effectiveness of bird deterrence strategies;
[0055] Step S3: Implement the bird deterrence strategy and evaluate its effectiveness according to the evaluation system;
[0056] Step S4: Based on the evaluation results, determine whether bird deterrence has been completed; if bird deterrence is completed, the algorithm ends; if bird deterrence is not completed, execute the online update algorithm for the bird deterrence strategy, update the bird deterrence strategy library, and return to step S3 to execute bird deterrence again, repeating the process until the bird deterrence task is completed.
[0057] In step S1 above, a comprehensive strategy library applicable to the repulsion of common birds in substations, such as sparrows, magpies, and crows, is included, containing ultrasound, laser, and acoustic-optical hybrid (combinations of different types of ultrasound and laser), as shown in Table 1.
[0058] Table 1. Substation Sound and Light Bird Repelling Strategies Database
[0059]
[0060] The species and number of birds, the methods of deterrence, weather conditions, and other factors all affect the effectiveness of bird control, necessitating different effective bird control strategies. Figure 2 The more bird-repelling methods a bird-repelling device is equipped with, the more mixed bird-repelling strategies can be selected, making it difficult to choose the optimal bird-repelling strategy for different scenarios. The species and number of birds, weather conditions, and other factors all affect the bird-repelling effect, corresponding to different optimal bird-repelling strategies.
[0061] In step S2 above, a bird deterrence effectiveness evaluation system is established based on three-dimensional indicators—response level, action delay, and action distance—and the entropy weight method. This system is used to evaluate the bird deterrence effect of mixed bird deterrence methods on birds. The evaluation of the bird deterrence response level is based on the birds' reaction to the bird deterrence methods, including three states: "startled flight," "restlessness," and "no response." "Startled flight" indicates that the birds have clearly flown away from their original location under the current bird deterrence strategy, indicating successful bird deterrence. "Restlessness" indicates that the birds are lingering and jumping in place, and "no response" indicates that the birds show no significant behavioral change, both indicating that the bird deterrence strategy is unsuccessful. The bird deterrence action delay is the time interval from the application of the bird deterrence method to the birds' response; the action distance is the maximum distance at which the bird deterrence device can effectively deter birds under a certain bird deterrence strategy. In this example, multi-factor controlled experiments were designed to investigate the effectiveness of ultrasonic bird deterrence and laser bird deterrence. The experimental data were then processed using the entropy weight method to construct a three-dimensional evaluation matrix that includes response level, action delay, and action distance. An evaluation system for different bird deterrence strategies was established to evaluate the effectiveness of different mixed bird deterrence strategies.
[0062] In step S2, the entropy weight method used in the process of establishing the evaluation system employs the extreme value method to standardize the three-dimensional indicators:
[0063]
[0064] The entropy value of the j-th item indicator is calculated using the following formula:
[0065]
[0066] Final output objective weights Among them, response level (j=1) is a very large indicator, action delay (j=2) is a very small indicator, and action distance (j=3) is a very large indicator.
[0067] In step S3, a comprehensive evaluation of the bird deterrence effect is conducted based on the bird deterrence effect evaluation system established in S2; and the "bird response status" is used as the basis for whether the bird deterrence is successful, with "startled flight" indicating successful bird deterrence, and "restlessness" and "no response" indicating failed bird deterrence.
[0068] In step S4, if the birds take flight, the bird deterrence task is complete and the algorithm ends; if the birds only exhibit "agitation" or "no response," the online update algorithm for the bird deterrence strategy is initiated. This algorithm mainly consists of two parts: an improved Q-learning reinforcement learning algorithm and an AHP-TOPSIS algorithm.
[0069] Combination Figure 3 This paper further explains the online update algorithm for bird deterrence strategies. The algorithm first adjusts the weights of the strategy library based on an improved Q-learning algorithm, and then continuously optimizes the weight allocation according to the actual bird deterrence situation, thereby forming a comprehensive bird deterrence strategy for different bird species and flock sizes. The Q-learning algorithm is a reinforcement learning algorithm with the characteristics of autonomous learning and online optimization, capable of driving the model's adaptive evolution through real-time bird deterrence feedback. To further accelerate the convergence speed of the algorithm, the learning rate is dynamically adjusted according to the bird deterrence effect. The specific execution process of the algorithm is as follows:
[0070] 1) Initialize the Q-learning algorithm parameters, including the state set S of the bird deterrence strategy and the corresponding behavior set A, discount factor γ, etc.; where the state set S contains bird information such as bird species, number, and distance, which is mainly obtained by the sensors of the bird deterrence device; the behavior set A consists of different hybrid bird deterrence strategies.
[0071] 2) Select a set of states s∈S and actions a∈A, execute the bird-repelling strategy, and obtain the state s' at the next moment;
[0072] 3) Based on the feedback from the bird deterrent actuator regarding the bird deterrent effect, calculate the dynamic learning rate α(r,β). The relevant calculation expression is as follows: Where β is the growth coefficient. For the mean reward, σ r To reward the standard deviation, ε is a small constant used to prevent division by zero, σ r Together with ε, they serve as a normalization factor to avoid an excessively large learning rate.
[0073] 4) Obtain the next action a' and update the bird deterrence effect status value;
[0074] 5) Determine whether the algorithm meets the convergence condition or whether the number of iterations has reached the upper limit.
[0075] After training is complete, the Q-learning algorithm can select the bird-repelling strategy based on the state.
[0076] In the online update algorithm for bird-repelling strategies, historical strategies output by the Q-learning algorithm are recorded, and the weights of strategies in the database are adjusted using the AHP-TOPSIS method to achieve optimization and update of bird-repelling strategies. A comprehensive weight is calculated using the Analytic Hierarchy Process (AHP) combined with subjective and objective weights, and the bird-repelling effect is comprehensively analyzed based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). First, a hierarchical structure is established based on the improved Q-learning algorithm. After establishing the hierarchical structure, a judgment matrix X is constructed using a calibration method, which is generated by horizontally comparing the importance of bird-repelling strategy parameters at the same level. The elements x in matrix X... ij The scale represents the relative importance of the i-th parameter compared to the j-th parameter in the bird deterrence strategy. The specific scale and meaning are shown in Table 2.
[0077] Table 2 Calibration Scale Settings
[0078]
[0079] At this point, the consistency index CR of the judgment matrix X needs to be calculated. If its value is greater than 0.1, it indicates that the generated judgment matrix does not meet the consistency requirements and needs to be reconstructed. The formula for calculating CR is as follows:
[0080]
[0081] In the formula, λ max (X) represents the largest eigenvalue of the judgment matrix X, n represents the order of the judgment matrix X, and RI represents the random consistency index. The correspondence between RI and n is shown in Table 3.
[0082] Table 3. Correspondence between random consistency index and matrix order.
[0083]
[0084] After verifying matrix consistency, the eigenvector corresponding to the largest eigenvalue is normalized to obtain subjective weights, which represent the relative importance of different parameters. Simultaneously, the objective weights of each bird-repelling strategy are determined based on the birds' response to different bird-repelling measures. These subjective and objective weights are then allocated using specific proportional coefficients to ultimately generate a comprehensive weighting of the bird-repelling strategy parameters.
[0085] In TOPSIS analysis, the Q-values corresponding to the bird deterrence strategy parameters are all maximally large indices. Assuming there are n bird deterrence strategies, each with m parameters, a decision matrix X = [x ij ] n×m The corresponding standardized decision matrix is Y = [y ij ] n×m The calculation formula is as follows:
[0086]
[0087] Suppose that the Q-value of the j-th parameter of the i-th bird-repelling strategy is close to the Q-value of the optimal strategy parameter. The degree of closeness to the Q value of the worst policy parameter is The two can be represented as:
[0088]
[0089] In the formula, and These are the maximum and minimum values in the j-th column of the standardized decision matrix Y, representing the optimal and worst solutions, respectively. Their calculation formulas are as follows:
[0090] Z + =(max{z 11 ,z 21 ,...,z n1},max{z 12 ,z 22 ,...,z n2},...,max{z 1m ,z 2m ,...,z nm})(5)
[0091] Z - =(min{z 11 ,z 21 ,...,z n1},min{z 12 ,z 22 ,...,z n2},...,min{z 1m ,z 2m ,...,z nm})(6)
[0092] Using parameter C i The value represents the degree of similarity between the i-th bird-repelling strategy and the optimal strategy, and is calculated using the following formula:
[0093]
[0094] C iThe larger the value, the better the bird-repelling strategy. Ranking bird-repelling strategies by proximity can help obtain the optimal strategy. Furthermore, the execution frequency of specific bird-repelling strategies can be arranged based on the ranking results.
[0095] If the bird-repelling task is not completed, the online update algorithm for the bird-repelling strategy will be continuously executed in a loop to achieve adaptive adjustment of the bird-repelling strategy. The specific process of this online optimization is as follows: Figure 3 As shown, this online evaluation cycle not only learns the current optimal bird-repelling strategy, but also adjusts the weight of the bird-repelling strategy in a timely manner based on feedback. Once birds become adaptive, the strategy can be changed promptly, thereby achieving a long-term anti-tolerance bird-repelling effect under the hybrid bird-repelling method.
[0096] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0099] Although embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make further changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the embodiments as well as all changes and modifications falling within the scope of the invention.
[0100] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A bird tolerance prevention and control method based on online evaluation, characterized in that... Includes the following steps: Step 1: Establish a comprehensive strategy library applicable to the repulsion of different bird species; Step 2: Construct a comprehensive evaluation system for bird deterrence effectiveness, which includes three-dimensional indicators such as response level, action delay, and action distance, to evaluate the effectiveness of bird deterrence strategies. Step 3: Execute the bird-repelling strategies in the comprehensive strategy library, and evaluate the effectiveness of the bird-repelling strategies according to the comprehensive evaluation system for bird-repelling effects; Step 4: Based on the evaluation results, determine whether the bird deterrence has been completed; If bird control is successfully completed, the process ends; if bird control is not successfully completed, the online update algorithm for the bird control strategy is executed to update the bird control strategy library and return to step 3 to execute bird control again, repeating this process until the bird control task is completed.
2. The bird tolerance prevention and control method based on online evaluation as described in claim 1, characterized in that: In step 1, the comprehensive strategy library includes different bird species that need to be driven away, as well as bird repelling methods formed by single or combined methods; wherein, single bird repelling methods include ultrasound, laser, predator calls, visual intimidation and chemical substances.
3. The bird tolerance prevention and control method based on online evaluation as described in claim 1, characterized in that: In step 2, the response level includes three levels: "startled," "restless," and "no response." "Startled" means that the birds being driven away clearly fly away from their original location under the current bird-driving strategy, indicating that the bird-driving is successful. "Restless" means that the birds being driven away hover and jump around in their original location under the current bird-driving strategy, indicating that the bird-driving is unsuccessful. "No response" means that the birds being driven away do not show any significant behavioral changes under the current bird-driving strategy, indicating that the bird-driving is unsuccessful. The delay in action is the time interval from the application of bird deterrence measures to the birds' response; The effective distance is the maximum distance at which the bird deterrent device can effectively drive away birds under the current bird deterrent strategy.
4. The bird tolerance prevention and control method based on online evaluation as described in claim 1, characterized in that: In step 2, three-dimensional indicators of the bird-repelling strategy are obtained, and then the three-dimensional indicators are standardized to obtain the entropy value of each indicator. Based on the entropy value of each indicator, an objective weight is obtained to evaluate the effectiveness of the bird-repelling strategy.
5. The bird tolerance prevention and control method based on online evaluation as described in claim 1, characterized in that: In step 3, a bird deterrent device is used to identify birds, a bird deterrent strategy for the bird is selected from the comprehensive strategy library, bird deterrent work is carried out based on the bird deterrent strategy, and the bird deterrent effect is judged according to the comprehensive evaluation system of bird deterrent effect.
6. The bird tolerance prevention and control method based on online evaluation as described in claim 1, characterized in that: In step 4, the online update algorithm for the bird-repelling strategy is executed, including: An improved Q-learning reinforcement learning method was used to adjust the weights of the comprehensive strategy library, and the weight allocation was continuously optimized according to the actual situation of bird control, so as to form a comprehensive bird control strategy for different bird species and flock sizes. The comprehensive weight is calculated by combining subjective and objective weights using the analytic hierarchy process, and the existing strategies are evaluated based on the superior-inferiority distance method to select the most suitable bird-repelling strategy.
7. The bird tolerance prevention and control method based on online evaluation as described in claim 6, characterized in that: An improved Q-learning reinforcement learning method was used to adjust the weights of the comprehensive strategy library, and the weight allocation was continuously optimized based on the actual bird control situation to form a comprehensive bird control strategy for different bird species and flock sizes, including: Step a1: Initialize the Q-learning algorithm parameters, including the state set S of the bird deterrence strategy and the corresponding behavior set A and discount factor γ; wherein, the state set S contains bird information, and the behavior set A consists of different mixed bird deterrence strategies. Step a2: Select a set of states s∈S and actions a∈A, execute the bird-repelling strategy, and obtain the state s' at the next moment; Step a3: Based on the feedback from the bird deterrent actuator regarding the bird deterrent effect, calculate the dynamic learning rate α(r,β). The relevant calculation expression is as follows: Where β is the growth coefficient. For the mean reward, σ r To reward the standard deviation, ε is a small constant used to prevent division by zero; Step a4: Obtain the next action a' and update the bird deterrence effect status value; Step a5: Determine whether the algorithm meets the convergence condition or whether the number of iterations has reached the upper limit; After training is completed, choose a bird deterrent strategy based on the bird's condition.
8. The bird tolerance prevention and control method based on online evaluation as described in claim 6, characterized in that: The comprehensive weight is calculated by combining subjective and objective weights using the analytic hierarchy process (AHP), and existing strategies are evaluated based on the superior-inferiority distance method to select the most suitable bird-repelling strategy. include, Establish a hierarchical structure that includes the solution layer, decision-making layer, and execution layer; A judgment matrix X was constructed using a calibration method. The importance of bird deterrence strategy parameters at the same level was compared horizontally, and the consistency of the matrix was verified. Based on the judgment matrix X, the comprehensive weights of the bird deterrence strategy parameters are obtained; In the superior-inferiority distance method, the Q-values corresponding to the bird-repelling strategy parameters are all maximally large indices. Assuming there are n bird-repelling strategies, each with m parameters, a decision matrix X = [x...] is constructed. ij ] n×m The corresponding standardized decision matrix is Y = [y ij ] n×m The calculation formula is as follows: Suppose that the Q-value of the j-th parameter of the i-th bird-repelling strategy is close to the Q-value of the optimal strategy parameter. The degree of closeness to the Q value of the worst policy parameter is Then there is, In the formula, and These are the maximum and minimum values of the j-th column in the standardized decision matrix Y, i.e., the optimal solution and the worst solution; Using parameter C i This indicates the degree of similarity between the i-th bird-repelling strategy and the optimal strategy, calculated using the following formula: C i The larger the value, the better the bird-repelling strategy; the bird-repelling strategies are ranked by proximity to obtain the optimal strategy.
9. The bird tolerance prevention and control method based on online evaluation as described in claim 8, characterized in that: Verifying the consistency of the judgment matrix X includes, The consistency index CR of the judgment matrix X is calculated according to the following formula. Where, λ max (X) is the largest eigenvalue of the judgment matrix X, n is the order of the judgment matrix X, and RI is the random consistency index. If the consistency index CR is greater than the set threshold, the judgment matrix X is considered to be inconsistent and the judgment matrix X is reconstructed.
10. The bird tolerance prevention and control method based on online evaluation as described in claim 9, characterized in that: Based on the judgment matrix X, the comprehensive weights of the bird-repelling strategy parameters are obtained, including: The eigenvector corresponding to the largest eigenvalue is normalized to obtain the subjective weight, and the objective weight of each bird-repelling strategy is determined based on the response effect of birds to different bird-repelling strategies. By allocating subjective and objective weights using specific proportional coefficients, the comprehensive weights of the bird deterrence strategy parameters are obtained.