Distributed ocean photovoltaic panel automatic anti-bird device and method

Abstract

The invention discloses a distributed ocean photovoltaic panel automatic anti-bird device and method. The method comprises the following steps: S1, networking and sensing acquisition; s2, identification and threshold adjustment; s3, linkage and expelling are carried out; s4, performing feedback and data recording; s5, performing optimization and operation and maintenance calibration; the problem of adaptation of special marine environments is solved in a targeted mode, environmental interference such as high salt fog and light intensity fluctuation is effectively resisted by deploying salt fog-resistant weather-resistant edge nodes, integrating a multi-module ad hoc network and adopting wavelet noise reduction processing, stable operation of equipment and data acquisition purity are guaranteed, and the blank that a conventional bird prevention scheme is insufficient in marine environment adaptability is filled up; the method overcomes the pain point of inaccurate bird recognition and intention judgment, dynamically adjusts a threshold value by means of deep learning, a behavior model and track time sequence analysis in combination with environmental parameters, accurately distinguishes passing birds from resident birds, accurately judges landing and excretion intentions, greatly reduces false triggering and missed triggering probabilities, and improves the reliability of recognition decision.

Description

Technical Field

[0001] This invention relates to the field of bird prevention for marine photovoltaic panels, and more specifically, to an automatic bird prevention device and method for distributed marine photovoltaic panels. Background Technology

[0002] With the rapid development of the global new energy industry, marine photovoltaics, as a clean and efficient form of energy utilization, has become an important development direction for the photovoltaic industry due to its advantages of not occupying land resources, stable sunlight conditions, and high power generation efficiency. However, marine photovoltaic arrays are exposed to the marine environment for a long time and are severely affected by seabird activities: seabirds landing on the surface of the photovoltaic panels and excreting will form stains, block sunlight and corrode the surface of the photovoltaic panels, resulting in a significant decrease in photoelectric conversion efficiency; trampling by some large seabirds may also cause physical damage to the photovoltaic panels, increasing equipment maintenance costs. Therefore, bird prevention measures have become a key link in ensuring the stable operation of marine photovoltaic projects.

[0003] Currently, existing bird control technologies mainly fall into two categories: physical bird control and electronic bird deterrence. Physical bird control methods, such as installing bird nets and bird spikes, suffer from problems such as bulky structures, difficult installation and maintenance, and are not suitable for large-scale deployment of distributed marine photovoltaic panels. They may also interfere with the normal migration of marine birds. Electronic bird deterrence methods, such as fixed-frequency sound waves and single strong light deterrence, offer some flexibility but have significant drawbacks. On the one hand, the marine environment is characterized by high salt spray, strong humidity fluctuations, and drastic changes in light intensity. Conventional sensors are easily affected by environmental interference, resulting in low accuracy in bird identification, and fixed detection thresholds cannot adapt to complex environmental changes. On the other hand, single, fixed-pattern deterrence methods can easily lead to bird habituation, and the deterrence effect will significantly decrease after long-term use. In addition, existing solutions lack distributed node coordination mechanisms, making it difficult for single-point deterrence to cover large-scale photovoltaic arrays. Furthermore, a closed-loop optimization system has not been established, making it impossible to dynamically adjust strategies based on changes in bird behavior and equipment operating status. At the same time, the problems of difficult equipment fault location and high maintenance costs in the marine environment have not been effectively solved.

[0004] Existing technologies fail to adequately consider adaptability to the unique marine environment, long-term bird protection, and ease of operation and maintenance, making it difficult to meet the actual bird protection needs of distributed marine photovoltaic panels. Therefore, there is an urgent need for a targeted, efficient, and stable automatic bird protection technology solution. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention aims to provide a distributed marine photovoltaic panel automatic bird deterrence device and method, which addresses the following issues: In the marine environment, high salt spray, fluctuating light intensity, and drastic temperature and humidity changes easily lead to large sensor detection errors and poor equipment stability. Conventional bird deterrence solutions lack environmentally adaptable design, failing to achieve accurate bird identification and stable operation. The invention also addresses the problem of inaccurate bird identification and intent judgment: Existing solutions often use fixed threshold judgments without dynamic adjustment based on environmental parameters, making it difficult to distinguish between passing birds and resident birds, and failing to accurately identify birds' landing and excretion intentions, easily resulting in false triggers or missed triggers. Furthermore, the invention addresses the problem of insufficient long-term deterrence effect: Single, fixed-pattern deterrence methods easily lead to bird habituation, and the lack of a distributed node coordination mechanism results in limited coverage of single-point deterrence, failing to form comprehensive protection. Finally, the invention addresses the problem of a lack of closed-loop optimization in the technical solution: Existing solutions lack a feedback iteration mechanism, failing to optimize strategies based on changes in bird behavior, deterrence effects, and equipment operation data, leading to a decline in bird deterrence efficiency after long-term use.

[0006] To solve the above problems, the present invention adopts the following technical solution.

[0007] An automatic bird-proofing method for distributed marine photovoltaic panels includes the following steps:

[0008] S1: Networking and Sensing Acquisition. Salt spray-resistant and weather-resistant edge nodes are deployed at preset locations on the photovoltaic array, integrating sensing, de-escalation, communication, and energy storage modules. They are connected to a remote management and control platform via long-distance communication. Each node forms a self-organizing network to create a monitoring network, loading bird databases, verification rules, and sensitive sound and light frequency band databases. Environmental and bird data are collected in real time, and effective feature data is retained after noise reduction and anti-interference processing.

[0009] S2: Identification and threshold adjustment. Bird species and numbers are identified through deep learning and basic database. Birds are distinguished from transient birds and resident birds by behavioral models. Landing and excretion intentions are determined by trajectory fitting and time-series posture analysis, combined with preset posture thresholds. The detection and judgment thresholds are dynamically adjusted in conjunction with environmental parameters, and multi-level verification is achieved by linking voiceprint comparison.

[0010] S3: Linkage and Removal. When S2 determines that there are resident birds and related intentions, the node triggers a removal command, which is synchronized to surrounding nodes via the communication unit to form a collaborative protection. Based on the sensitive frequency band database, interference from non-target organisms is avoided. Specific parameters are called in combination with bird species to perform randomized acoustic, optical, and electronic removal.

[0011] S4: Feedback and data recording; after the birds are driven away, continuously monitor the birds’ remaining presence and trajectory changes; collect response data, energy consumption parameters and equipment status; and upload the data to the network sharing node and remote management platform after classification.

[0012] S5: Optimization and Operation and Maintenance Calibration. The remote management and control platform analyzes the feedback data from S4 through deep learning, identifies the habitual characteristics of birds and updates the deportation parameter library, and simultaneously distributes the optimization strategies to each node; it also constructs an error compensation model based on environmental parameters to calibrate sensor accuracy, and achieves accurate fault location, intelligent alarm and remote operation and maintenance scheduling through long-distance communication.

[0013] As a further technical solution of the present invention, the noise reduction and anti-interference processing in S1 specifically includes the following sub-steps:

[0014] S11: Collect raw signals output by each sensing module of the edge node, including raw environmental parameter signals and raw bird monitoring signals, and set the signal sampling frequency and duration;

[0015] S12: Perform wavelet decomposition on the original signal, determine the number of decomposition levels and extract the wavelet coefficients of each level, and separate the effective components of the signal from the noise components.

[0016] S13: Threshold denoising is performed on the decomposed wavelet coefficients to retain the wavelet coefficients corresponding to the effective components and suppress the wavelet coefficients corresponding to the noise components. The denoised signal is then reconstructed by inverse wavelet transform. The denoising process of wavelet decomposition, threshold processing, and inverse transform is clearly defined, which effectively separates the effective components and noise of environmental and bird signals, improves data purity, and lays the foundation for accurate identification in the future.

[0017] As a further technical solution of the present invention, S13 is implemented using a wavelet soft threshold denoising function, specifically: the processed wavelet coefficients The rule for determining the value is that when the original coefficients after wavelet decomposition... The absolute value is greater than the soft threshold. hour, for The sign function value and absolute value minus The product of the differences; when The absolute value is less than or equal to hour, Set to 0; where the soft threshold The value is determined by the noise standard deviation. With the number of signal sampling points pass The corresponding operational relationship yields the noise standard deviation. The median estimation method is used for calculation, specifically: The value is derived from the original coefficients after wavelet decomposition. The median of the absolute values ​​is obtained by dividing by 0.6745; The sign function outputs 1 when the input is positive and -1 when the input is negative. The threshold is calculated using the wavelet soft threshold function and a clear formula to accurately suppress noise, retain effective signals, solve the problem of marine environmental signal interference, and ensure the reliability of feature data.

[0018] As a further technical solution of the present invention, the "dynamic adjustment detection and intent determination threshold" in S2 specifically includes the following sub-steps:

[0019] S21: Extract the environmental parameters of light intensity and salt spray concentration after S1 collection and noise reduction, and use them as input variables for threshold adjustment;

[0020] S22: Based on the preset environmental impact weight model, calculate the threshold compensation coefficients corresponding to each environmental parameter;

[0021] S23: Combining the basic threshold and the compensation coefficient, a real-time detection threshold and an intent determination threshold are generated and synchronously updated to the edge node recognition algorithm. Using light intensity and salt spray concentration as inputs, the compensation coefficient is calculated through the environmental influence weight model to achieve dynamic adjustment of the threshold, adapt to marine environmental fluctuations, and improve recognition accuracy.

[0022] As a further technical solution of the present invention, the dynamic threshold adjustment in S23 adopts a linear weighted model, the specific logic of which is: real-time threshold The value is determined by the preset base threshold. The result is obtained by multiplying by an environmental correction factor, which is derived from the light intensity compensation coefficient. Multiply by its weighting factor Salt spray concentration compensation coefficient Multiply by its weighting factor Then sum them up, and then add the correction factor. Composition; among which , These are the weighting coefficients for light intensity and salt spray concentration, respectively. The values ​​range from 0 to 1; It is the light intensity compensation coefficient, which is determined by the ratio of the actual light intensity to the standard light intensity; The salt spray concentration compensation coefficient is determined by the ratio of the actual salt spray concentration to the standard salt spray concentration. The correction coefficient, ranging from 0.9 to 1.1, is used to offset systematic errors in the model. A linear weighted model is used to quantify the environmental impact, and the rules for weight and coefficient values ​​are clearly defined to make the threshold adjustment more accurate and avoid misjudgments or omissions caused by changes in salt spray and light intensity.

[0023] As a further technical solution of the present invention, the "execution of acoustic-optical-electric randomized expulsion" in S3 specifically includes the following sub-steps:

[0024] S31: Based on the bird species identified in S2, retrieve the corresponding basic deterrent frequency and intensity from the sensitive sound and light frequency band database;

[0025] S32: Generate random perturbation based on time series to ensure that the perturbation fluctuates outside the frequency band sensitive to non-target organisms;

[0026] S33: The basic repulsion parameters are superimposed with random disturbances to output the final acoustic, optical, and electronic repulsion parameters and trigger the execution module. The basic repulsion parameters are retrieved according to the bird species, and random disturbances are superimposed to avoid bird habituation and interference from non-target organisms, thereby improving the targeting and effectiveness of the repulsion.

[0027] As a further technical solution of the present invention, the randomized expulsion frequency generation logic in S33 is as follows: Momentary frequency of expulsion The value is determined by the base dispersal frequency. This is obtained by superimposing a time-varying sinusoidal perturbation; the maximum amplitude of this sinusoidal perturbation is... angular frequency Used to control the frequency change period and initial phase. The value range is 0- The frequency change trajectory is generated by a random function, so that the frequency change trajectory is different for each drive-away. Differentiated frequency trajectories are generated by sinusoidal perturbation, which enhances the long-term effectiveness of the drive-away, while strictly limiting the range of perturbation, taking into account both bird deterrence and marine ecological protection.

[0028] As a further technical solution of the present invention, the "construction of error compensation model" in step S5 specifically includes the following sub-steps:

[0029] S51: Collect sensor calibration data daily, including actual measured values ​​and standard reference values, and calculate the corresponding error values;

[0030] S52: Using temperature and running time as independent variables and error as the dependent variable, fit the sensor drift curve to determine the drift pattern;

[0031] S53: Based on the drift curve, an error compensation model is constructed. Real-time environmental parameters and runtime are substituted into the model, and the sensor calibration value is output. The drift curve is fitted based on temperature and runtime to construct the error compensation model, accurately calibrate the sensor, offset the effects of environment and aging, and ensure the long-term stability of the detection data.

[0032] As a further technical solution of the present invention, the specific logic of the sensor drift error compensation model in S53 is as follows: the output value after sensor calibration Raw measurement values ​​from the sensor The total drift error term is obtained by subtracting the total drift error term; the total drift error term is a linear superposition of the temperature drift component and the time aging component, wherein the temperature drift component is the temperature drift coefficient. Multiply by the difference between the actual temperature and the standard calibration temperature. The time aging component is the time aging factor. Multiply by the continuous operating time of the sensor It quantifies temperature drift and time aging errors, calculates the total error through linear superposition, accurately corrects the sensor output, and solves the problem of equipment detection accuracy decay in marine environments.

[0033] An automatic bird deterrent device for distributed marine photovoltaic panels includes salt spray-resistant and weather-resistant edge nodes and a remote control platform. The edge nodes integrate sensing, bird deterrence, communication, and energy storage modules. Each edge node self-organizes to form a monitoring network and establishes a connection with the remote control platform through long-distance communication. The device also includes the following functional modules:

[0034] Distributed sensing and networking module: mounted on each edge node, used to deploy edge nodes at preset locations in the photovoltaic panel array, load bird database, verification rules and sensitive sound and light frequency band database, collect environmental and bird data in real time, perform noise reduction and anti-interference processing on the collected data and retain effective feature data, and at the same time realize self-organizing network of each node and long-distance communication connection with remote management and control platform;

[0035] Intelligent recognition and decision-making module: It is used to identify bird species and numbers through deep learning combined with a basic database, distinguish between passing birds and resident birds based on behavior models, determine the landing and excretion intentions of birds through trajectory fitting and time-series posture analysis, and combine preset posture thresholds. It dynamically adjusts the detection and judgment thresholds in combination with environmental parameters, and links voiceprint comparison to achieve multi-level verification.

[0036] Collaborative Repulsion Execution Module: Equipped on each edge node, it is used to trigger a repulsion command when the intelligent recognition and decision-making module determines that there are resident birds and related intentions, and synchronizes the command to the surrounding nodes through the communication unit to form collaborative protection. It avoids interference from non-target organisms based on the sensitive frequency band database, and calls specific parameters based on bird species to perform randomized acoustic, optical and electronic repulsion operations.

[0037] Effect feedback monitoring module: mounted on each edge node, used to continuously monitor bird remnants and trajectory changes after the expulsion operation, collect expulsion response data, equipment energy consumption parameters and equipment operating status, classify and process the data and upload it to the network sharing node and remote management and control platform;

[0038] Adaptive Operation and Maintenance Optimization Module: Equipped on the remote management and control platform, it uses deep learning to analyze feedback data uploaded by the effect feedback monitoring module, identify bird habitual characteristics and update the deportation parameter library, and synchronously distribute the data to each edge node to optimize the deportation strategy. Based on environmental parameters, it constructs an error compensation model to calibrate sensor accuracy and achieves accurate equipment fault location, intelligent alarm, and remote operation and maintenance scheduling through long-distance communication.

[0039] Compared with the prior art, the advantages of this invention are:

[0040] This invention specifically addresses the challenge of adapting to special marine environments. By deploying salt spray-resistant and weather-resistant edge nodes, integrating multi-module self-organizing networks, and employing wavelet noise reduction processing, it effectively resists environmental interference such as high salt spray and light intensity fluctuations, ensuring stable equipment operation and data acquisition purity, thus filling the gap in the insufficient adaptability of conventional bird-proofing solutions to marine environments.

[0041] Overcoming the pain point of inaccurate bird identification and intent determination, relying on deep learning, behavioral models and trajectory time series analysis, combined with dynamic adjustment of thresholds based on environmental parameters, it accurately distinguishes between passing birds and resident birds, accurately determines landing and excretion intentions, significantly reduces the probability of false triggers and missed triggers, and improves the reliability of identification decisions.

[0042] To address the issues of insufficient long-term effectiveness and limited coverage of bird deterrence, a comprehensive protective network is formed through node-based collaborative protection. Specific parameters are invoked based on bird species, and randomized acoustic, electrical, and electronic deterrence strategies are superimposed. This not only prevents birds from developing habits but also avoids interference from non-target organisms, significantly improving the targeting and long-term effectiveness of deterrence.

[0043] To address the lack of closed-loop optimization in existing solutions, this approach uses post-repellent data feedback and deep learning analysis on a remote management platform to dynamically update the repellent parameter library and strategies. This enables continuous iteration of "collection-judgment-execution-optimization," ensuring long-term stability of bird deterrence and preventing efficiency degradation.

[0044] In response to the challenges of operation and maintenance in marine environments, this project calibrates sensor accuracy by constructing an error compensation model and combines it with long-distance communication to achieve precise fault location, intelligent alarm, and remote operation and maintenance scheduling. This significantly reduces the operation and maintenance costs of distributed photovoltaic arrays, improves the overall operational efficiency of the project, and provides strong support for the stable power generation of marine photovoltaic projects. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0046] 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.

[0047] This invention provides an embodiment 1

[0048] Please see Figure 1 An automatic bird-proofing method for distributed marine photovoltaic panels includes the following steps:

[0049] S1: Networking and Sensing Acquisition. Salt spray-resistant and weather-resistant edge nodes are deployed at preset locations on the photovoltaic array, integrating sensing, de-escalation, communication, and energy storage modules. They are connected to a remote management and control platform via long-distance communication. Each node forms a self-organizing network to create a monitoring network, loading bird databases, verification rules, and sensitive sound and light frequency band databases. Environmental and bird data are collected in real time, and effective feature data is retained after noise reduction and anti-interference processing.

[0050] The noise reduction and anti-interference processing in S1 specifically includes the following sub-steps:

[0051] S11: Collect raw signals output by each sensing module of the edge node, including raw environmental parameter signals and raw bird monitoring signals, and set the signal sampling frequency and duration;

[0052] S12: Perform wavelet decomposition on the original signal, determine the number of decomposition levels and extract the wavelet coefficients of each level, and separate the effective components of the signal from the noise components.

[0053] S13: Threshold denoising is performed on the decomposed wavelet coefficients to retain the wavelet coefficients corresponding to the effective components and suppress the wavelet coefficients corresponding to the noise components. The denoised signal is then reconstructed by inverse wavelet transform. The denoising process of wavelet decomposition, threshold processing, and inverse transform is clarified, which effectively separates the effective components and noise of environmental and bird signals, improves data purity, and lays the foundation for subsequent accurate identification.

[0054] S13 is implemented using a wavelet soft thresholding noise reduction function. The specific logic is as follows: processed wavelet coefficients The rule for determining the value is that when the original coefficients after wavelet decomposition... The absolute value is greater than the soft threshold. hour, for The sign function value and absolute value minus The product of the differences; when The absolute value is less than or equal to hour, Set to 0; where the soft threshold The value is determined by the noise standard deviation. With the number of signal sampling points pass The corresponding operational relationship yields the noise standard deviation. The median estimation method is used for calculation, specifically: The value is derived from the original coefficients after wavelet decomposition. The median of the absolute values ​​is obtained by dividing by 0.6745; The sign function outputs 1 when the input is positive and -1 when the input is negative. The threshold is calculated using the wavelet soft threshold function and a clear formula to accurately suppress noise, retain effective signals, solve the problem of marine environmental signal interference, and ensure the reliability of feature data.

[0055] S2: Identification and threshold adjustment. Bird species and numbers are identified through deep learning and basic database. Birds are distinguished from transient birds and resident birds by behavioral models. Landing and excretion intentions are determined by trajectory fitting and time-series posture analysis, combined with preset posture thresholds. The detection and judgment thresholds are dynamically adjusted in conjunction with environmental parameters, and multi-level verification is achieved by linking voiceprint comparison.

[0056] The "Dynamic Adjustment Detection and Intent Determination Threshold" in S2 specifically includes the following sub-steps:

[0057] S21: Extract the environmental parameters of light intensity and salt spray concentration after S1 collection and noise reduction, and use them as input variables for threshold adjustment;

[0058] S22: Based on the preset environmental impact weight model, calculate the threshold compensation coefficients corresponding to each environmental parameter;

[0059] S23: Combining the basic threshold and the compensation coefficient, a real-time detection threshold and an intent determination threshold are generated and synchronously updated to the edge node recognition algorithm. Using light intensity and salt spray concentration as inputs, the compensation coefficient is calculated through the environmental impact weight model to achieve dynamic adjustment of the threshold, adapt to marine environmental fluctuations, and improve recognition accuracy.

[0060] In S23, the dynamic threshold adjustment adopts a linear weighted model, specifically: real-time threshold The value is determined by the preset base threshold. The result is obtained by multiplying by an environmental correction factor, which is derived from the light intensity compensation coefficient. Multiply by its weighting factor Salt spray concentration compensation coefficient Multiply by its weighting factor Then sum them up, and then add the correction factor. Composition; among which , These are the weighting coefficients for light intensity and salt spray concentration, respectively. The values ​​range from 0 to 1; It is the light intensity compensation coefficient, which is determined by the ratio of the actual light intensity to the standard light intensity; The salt spray concentration compensation coefficient is determined by the ratio of the actual salt spray concentration to the standard salt spray concentration. The correction coefficient, ranging from 0.9 to 1.1, is used to offset systematic errors in the model. A linear weighted model is used to quantify the environmental impact, and the rules for weight and coefficient values ​​are clearly defined to make the threshold adjustment more accurate and avoid misjudgments and missed judgments caused by salt spray and changes in light intensity.

[0061] S3: Linkage and Removal. When S2 determines that there are resident birds and related intentions, the node triggers a removal command, which is synchronized to surrounding nodes via the communication unit to form a collaborative protection. Based on the sensitive frequency band database, interference from non-target organisms is avoided. Specific parameters are called in combination with bird species to perform randomized acoustic, optical, and electronic removal.

[0062] S3's "Perform randomized acoustic-optical-electrical expulsion" specifically includes the following sub-steps:

[0063] S31: Based on the bird species identified in S2, retrieve the corresponding basic deterrent frequency and intensity from the sensitive sound and light frequency band database;

[0064] S32: Generate random perturbation based on time series to ensure that the perturbation fluctuates outside the frequency band sensitive to non-target organisms;

[0065] S33: The basic repulsion parameters are superimposed with random disturbances to output the final acoustic, optical and electronic repulsion parameters and trigger the execution module. The basic repulsion parameters are retrieved according to the bird species, and random disturbances are superimposed to avoid bird habituation and avoid interference from non-target organisms, thereby improving the targeting and effectiveness of the repulsion.

[0066] The randomized expulsion frequency generation logic in S33 is as follows: Momentary frequency of expulsion The value is determined by the base dispersal frequency. This is obtained by superimposing a time-varying sinusoidal perturbation; the maximum amplitude of this sinusoidal perturbation is... angular frequency Used to control the frequency change period and initial phase. The value range is 0- The frequency change trajectory is generated by a random function to ensure that each drive-away has a different trajectory. The differentiated frequency trajectory is generated by sinusoidal disturbance to enhance the long-term effectiveness of the drive-away. At the same time, the range of disturbance is strictly limited to balance bird deterrence and marine ecological protection.

[0067] S4: Feedback and data recording; after the birds are driven away, continuously monitor the birds’ remaining presence and trajectory changes; collect response data, energy consumption parameters and equipment status; and upload the data to the network sharing node and remote management platform after classification.

[0068] S5: Optimization and Operation / Maintenance Calibration. The remote management platform analyzes the feedback data from S4 through deep learning, identifies the habitual characteristics of birds and updates the dispersal parameter library, and simultaneously distributes optimization strategies to each node; it also builds an error compensation model based on environmental parameters to calibrate sensor accuracy, and achieves accurate fault location, intelligent alarm, and remote operation and maintenance scheduling through long-distance communication.

[0069] The "Constructing an Error Compensation Model" in S5 specifically includes the following sub-steps:

[0070] S51: Collect sensor calibration data daily, including actual measured values ​​and standard reference values, and calculate the corresponding error values;

[0071] S52: Using temperature and running time as independent variables and error as the dependent variable, fit the sensor drift curve to determine the drift pattern;

[0072] S53: Based on the drift curve, an error compensation model is constructed. Real-time environmental parameters and running time are substituted to output the sensor calibration value. The drift curve is fitted based on temperature and running time to construct the error compensation model, accurately calibrate the sensor, offset the effects of environment and aging, and ensure the long-term stability of detection data.

[0073] The specific logic of the sensor drift error compensation model in S53 is as follows: After sensor calibration, the output value... Raw measurement values ​​from the sensor The total drift error term is obtained by subtracting the total drift error term; the total drift error term is a linear superposition of the temperature drift component and the time aging component, wherein the temperature drift component is the temperature drift coefficient. Multiply by the difference between the actual temperature and the standard calibration temperature. The time aging component is the time aging factor. Multiply by the continuous operating time of the sensor The error caused by temperature drift and time aging is quantified, and the total error is calculated by linear superposition to accurately correct the sensor output and solve the problem of equipment detection accuracy decay in marine environments.

[0074] This invention provides an embodiment 2

[0075] S13 is implemented using a wavelet soft thresholding noise reduction function, with the specific formula as follows:

[0076]

[0077] in, These are the original coefficients after wavelet decomposition; These are the wavelet coefficients after thresholding. The soft threshold is calculated using the following formula: , This represents the number of signal sampling points. The noise standard deviation is calculated using the median estimation method, and the formula is as follows: ; This is a sign function; it outputs 1 when the input is positive and -1 when the input is negative.

[0078] In S23, the dynamic threshold adjustment adopts a linear weighted model, and the specific formula is as follows:

[0079]

[0080] in, Real-time threshold; The preset base threshold; , These are the weighting coefficients for light intensity and salt spray concentration, respectively. The values ​​range from 0 to 1; It is the light intensity compensation coefficient, which is determined by the ratio of the actual light intensity to the standard light intensity; The salt spray concentration compensation coefficient is determined by the ratio of the actual salt spray concentration to the standard salt spray concentration. This is a correction factor, with a value ranging from 0.9 to 1.1, used to offset systematic errors in the model;

[0081] The formula for generating the randomized expulsion frequency in S33 is:

[0082]

[0083] in, Let be the expulsion frequency at time t; Based on the drive-off frequency; The maximum frequency disturbance is represented by a value of [value to be filled in]. 10%-20%; Angular frequency is used to control the period of frequency change; It is a time variable; The initial phase, with a value range of 0-2π, is generated by a random function to ensure that the frequency change trajectory is different for each expulsion.

[0084] The formula for the sensor drift error compensation model in S53 is:

[0085]

[0086] in, The output value after sensor calibration; The original measurement value of the sensor; This is the temperature drift coefficient, expressed in mV / ℃, determined by experimental calibration. This is the difference between the actual temperature and the standard calibration temperature. The time aging factor, measured in mV / h, is used to characterize the aging error of the sensor during long-term operation. This refers to the continuous operating time of the sensor.

[0087] This invention provides an embodiment 3

[0088] An automatic bird deterrent device for distributed marine photovoltaic panels includes salt spray-resistant and weather-resistant edge nodes and a remote control platform. The edge nodes integrate sensing, bird deterrence, communication, and energy storage modules. Each edge node self-organizes to form a monitoring network and establishes a connection with the remote control platform through long-distance communication. The device also includes the following functional modules:

[0089] Distributed sensing and networking module: mounted on each edge node, used to deploy edge nodes at preset locations in the photovoltaic panel array, load bird database, verification rules and sensitive sound and light frequency band database, collect environmental and bird data in real time, perform noise reduction and anti-interference processing on the collected data and retain effective feature data, and at the same time realize self-organizing network of each node and long-distance communication connection with remote management and control platform;

[0090] Intelligent recognition and decision-making module: It is used to identify bird species and numbers through deep learning combined with a basic database, distinguish between passing birds and resident birds based on behavior models, determine the landing and excretion intentions of birds through trajectory fitting and time-series posture analysis, and combine preset posture thresholds. It dynamically adjusts the detection and judgment thresholds in combination with environmental parameters, and links voiceprint comparison to achieve multi-level verification.

[0091] Collaborative Repulsion Execution Module: Equipped on each edge node, it is used to trigger a repulsion command when the intelligent recognition and decision-making module determines that there are resident birds and related intentions, and synchronizes the command to the surrounding nodes through the communication unit to form collaborative protection. It avoids interference from non-target organisms based on the sensitive frequency band database, and calls specific parameters based on bird species to perform randomized acoustic, optical and electronic repulsion operations.

[0092] Effect feedback monitoring module: mounted on each edge node, used to continuously monitor bird remnants and trajectory changes after the expulsion operation, collect expulsion response data, equipment energy consumption parameters and equipment operating status, classify and process the data and upload it to the network sharing node and remote management and control platform;

[0093] Adaptive Operation and Maintenance Optimization Module: Equipped on the remote management and control platform, it is used to analyze the feedback data uploaded by the effect feedback monitoring module through deep learning, identify the habitual characteristics of birds and update the repulsion parameter library, and simultaneously distribute it to each edge node to optimize the repulsion strategy. Based on environmental parameters, it builds an error compensation model to calibrate the sensor accuracy, and realizes accurate location of equipment faults, intelligent alarm and remote operation and maintenance scheduling through long-distance communication.

[0094] In summary, this invention specifically addresses the challenge of adapting to special marine environments. By deploying salt spray-resistant and weather-resistant edge nodes, integrating multi-module self-organizing networks, and employing wavelet noise reduction processing, it effectively resists environmental interference such as high salt spray and light intensity fluctuations, ensuring stable equipment operation and data acquisition purity, thus filling the gap in the insufficient adaptability of conventional bird-proofing solutions to marine environments.

[0095] Overcoming the pain point of inaccurate bird identification and intent determination, relying on deep learning, behavioral models and trajectory time series analysis, combined with dynamic adjustment of thresholds based on environmental parameters, it accurately distinguishes between passing birds and resident birds, accurately determines landing and excretion intentions, significantly reduces the probability of false triggers and missed triggers, and improves the reliability of identification decisions.

[0096] To address the issues of insufficient long-term effectiveness and limited coverage of bird deterrence, a comprehensive protective network is formed through node-based collaborative protection. Specific parameters are invoked based on bird species, and randomized acoustic, electrical, and electronic deterrence strategies are superimposed. This not only prevents birds from developing habits but also avoids interference from non-target organisms, significantly improving the targeting and long-term effectiveness of deterrence.

[0097] To address the lack of closed-loop optimization in existing solutions, this approach uses post-repellent data feedback and deep learning analysis on a remote management platform to dynamically update the repellent parameter library and strategies. This enables continuous iteration of "collection-judgment-execution-optimization," ensuring long-term stability of bird deterrence and preventing efficiency degradation.

[0098] In response to the challenges of operation and maintenance in marine environments, this project calibrates sensor accuracy by constructing an error compensation model and combines it with long-distance communication to achieve precise fault location, intelligent alarm, and remote operation and maintenance scheduling. This significantly reduces the operation and maintenance costs of distributed photovoltaic arrays, improves the overall operational efficiency of the project, and provides strong support for the stable power generation of marine photovoltaic projects.

[0099] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. An automatic bird-proofing method for distributed marine photovoltaic panels, characterized in that: Includes the following steps: S1: Networking and Sensing Acquisition: Salt spray-resistant and weather-resistant edge nodes are deployed at preset locations on the photovoltaic panel array, integrating sensing, de-escalation, communication and energy storage modules. They are connected to a remote management and control platform via long-distance communication. Each node forms a self-organizing network to create a monitoring network, loading bird databases, verification rules and sensitive sound and light frequency band databases, and collecting environmental and bird data in real time. After noise reduction and anti-interference processing, effective feature data is retained. S2: Identification and Threshold Adjustment: Bird species and numbers are identified through deep learning and basic database. Birds passing by and resident birds are distinguished by behavioral models. Landing and excretion intentions are determined by trajectory fitting and time-series posture analysis, combined with preset posture thresholds. The detection and judgment thresholds are dynamically adjusted in conjunction with environmental parameters, and multi-level verification is achieved by linking voiceprint comparison. S3: Linkage and Repelling: When S2 determines that there are resident birds and related intentions, the node triggers a repelling command, which is synchronized to surrounding nodes through the communication unit to form a collaborative protection; based on the sensitive frequency band database, non-target biological interference is avoided, and specific parameters are called in combination with bird species to perform randomized acoustic, optical and electronic repelling; S4: Feedback and Data Recording: After the birds are driven away, continuously monitor the birds' remains and trajectory changes, collect response data, energy consumption parameters and equipment status, and upload the data to the network sharing node and remote management platform after classification. S5: Optimization and Operation and Maintenance Calibration: The remote management and control platform analyzes the feedback data from S4 through deep learning, identifies the habitual characteristics of birds and updates the dispersal parameter library, and simultaneously distributes the optimization strategies to each node. An error compensation model is constructed based on environmental parameters to calibrate sensor accuracy, and long-distance communication enables precise fault location, intelligent alarm, and remote operation and maintenance scheduling.

2. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 1, characterized in that: The noise reduction and anti-interference processing in S1 specifically includes the following sub-steps: S11: Collect raw signals output by each sensing module of the edge node, including raw environmental parameter signals and raw bird monitoring signals, and set the signal sampling frequency and duration; S12: Perform wavelet decomposition on the original signal, determine the number of decomposition levels and extract the wavelet coefficients of each level, and separate the effective components of the signal from the noise components. S13: Threshold denoising is performed on the decomposed wavelet coefficients to retain the wavelet coefficients corresponding to the effective components and suppress the wavelet coefficients corresponding to the noise components. The denoised signal is then reconstructed by inverse wavelet transform.

3. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 2, characterized in that: S13 is implemented using a wavelet soft thresholding noise reduction function. The specific logic is as follows: processed wavelet coefficients The rule for determining the value is that when the original coefficients after wavelet decomposition... The absolute value is greater than the soft threshold. hour, for The sign function value and absolute value minus The product of the differences; when The absolute value is less than or equal to hour, Set to 0; where the soft threshold The value is determined by the noise standard deviation. With the number of signal sampling points pass The corresponding operational relationship yields the noise standard deviation. The median estimation method is used for calculation, specifically: The value is derived from the original coefficients after wavelet decomposition. The median of the absolute values ​​is obtained by dividing by 0.6745; This is a sign function; it outputs 1 when the input is positive and -1 when the input is negative.

4. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 1, characterized in that: The dynamic adjustment detection and intent determination threshold in S2 specifically includes the following sub-steps: S21: Extract the environmental parameters of light intensity and salt spray concentration after S1 collection and noise reduction, and use them as input variables for threshold adjustment; S22: Based on the preset environmental impact weight model, calculate the threshold compensation coefficients corresponding to each environmental parameter; S23: Combine the basic threshold and the compensation coefficient to generate the real-time detection threshold and the intent determination threshold, and update them synchronously to the edge node recognition algorithm.

5. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 4, characterized in that: The dynamic threshold adjustment in S23 adopts a linear weighted model, specifically: real-time threshold The value is determined by the preset base threshold. Multiply by an environmental correction factor to obtain; This environmental correction factor is determined by the light intensity compensation coefficient. Multiply by its weighting factor Salt spray concentration compensation coefficient Multiply by its weighting factor Then sum them up, and then add the correction factor. Composition; among which , These are the weighting coefficients for light intensity and salt spray concentration, respectively. The values ​​range from 0 to 1; It is the light intensity compensation coefficient, which is determined by the ratio of the actual light intensity to the standard light intensity; The salt spray concentration compensation coefficient is determined by the ratio of the actual salt spray concentration to the standard salt spray concentration. This is a correction factor, with a value ranging from 0.9 to 1.1, used to offset systematic errors in the model.

6. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 1, characterized in that: The execution of the acoustic-optical-electric randomized expulsion in S3 specifically includes the following sub-steps: S31: Based on the bird species identified in S2, retrieve the corresponding basic deterrent frequency and intensity from the sensitive sound and light frequency band database; S32: Generate random perturbation based on time series to ensure that the perturbation fluctuates outside the non-target biological sensitive frequency band; S33: Superimpose the basic dispersal parameters with the random disturbance, output the final acoustic-optical-electric dispersal parameters, and trigger the execution module.

7. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 6, characterized in that: The randomized expulsion frequency generation logic in S33 is as follows: Momentary frequency of expulsion The value is determined by the base dispersal frequency. This is obtained by superimposing a time-varying sinusoidal perturbation; the maximum amplitude of this sinusoidal perturbation is... angular frequency Used to control the frequency change period and initial phase. The value range is 0- It is generated by a random function, thereby ensuring that the frequency of each expulsion changes in a different trajectory.

8. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 1, characterized in that: The construction of the error compensation model in S5 specifically includes the following sub-steps: S51: Collect sensor calibration data daily, including actual measured values ​​and standard reference values, and calculate the corresponding error values; S52: Using temperature and running time as independent variables and error as the dependent variable, fit the sensor drift curve to determine the drift pattern; S53: Construct an error compensation model based on the drift curve, input real-time environmental parameters and runtime, and output sensor calibration values.

9. The automatic bird-proofing method for distributed marine photovoltaic panels according to claim 8, characterized in that: The specific logic of the sensor drift error compensation model in S53 is as follows: the sensor output value after calibration Raw measurement values ​​from the sensor Subtract the total drift error term to obtain; The total drift error term is a linear superposition of the temperature drift component and the time aging component, where the temperature drift component is the temperature drift coefficient. Multiply by the difference between the actual temperature and the standard calibration temperature. The time aging component is the time aging factor. Multiply by the continuous operating time of the sensor .

10. The automatic bird-proof device for distributed marine photovoltaic panels according to claim 1, characterized in that: The device includes salt spray-resistant and weather-resistant edge nodes and a remote control platform. The edge nodes integrate sensing, de-energizing, communication, and energy storage modules. Each edge node self-organizes to form a monitoring network and establishes a connection with the remote control platform through long-distance communication. The device also includes the following functional modules: Distributed sensing and networking module: mounted on each edge node, used to deploy edge nodes at preset locations in the photovoltaic panel array, load bird database, verification rules and sensitive sound and light frequency band database, collect environmental and bird data in real time, perform noise reduction and anti-interference processing on the collected data and retain effective feature data, and at the same time realize self-organizing network of each node and long-distance communication connection with remote management and control platform; Intelligent recognition and decision-making module: It is used to identify bird species and numbers through deep learning combined with a basic database, distinguish between passing birds and resident birds based on behavior models, determine the landing and excretion intentions of birds through trajectory fitting and time-series posture analysis, and combine preset posture thresholds. It dynamically adjusts the detection and judgment thresholds in combination with environmental parameters, and links voiceprint comparison to achieve multi-level verification. Collaborative Repulsion Execution Module: Equipped on each edge node, it is used to trigger a repulsion command when the intelligent recognition and decision-making module determines that there are resident birds and related intentions, and synchronizes the command to the surrounding nodes through the communication unit to form collaborative protection. It avoids interference from non-target organisms based on the sensitive frequency band database, and calls specific parameters based on bird species to perform randomized acoustic, optical and electronic repulsion operations. Effect feedback monitoring module: mounted on each edge node, used to continuously monitor bird remnants and trajectory changes after the expulsion operation, collect expulsion response data, equipment energy consumption parameters and equipment operating status, classify and process the data and upload it to the network sharing node and remote management and control platform; Adaptive Operation and Maintenance Optimization Module: Equipped on the remote management and control platform, it uses deep learning to analyze feedback data uploaded by the effect feedback monitoring module, identify bird habitual characteristics and update the deportation parameter library, and synchronously distribute the data to each edge node to optimize the deportation strategy. Based on environmental parameters, it constructs an error compensation model to calibrate sensor accuracy and achieves accurate equipment fault location, intelligent alarm, and remote operation and maintenance scheduling through long-distance communication.

Images