Bird risk early warning method based on space-time meteorological environment

Abstract

The present application is a bird situation risk early warning method based on space-time meteorological environment. The present application relates to the technical field of aviation safety and ecological protection, and performs space-time grid division; collects multi-source data; performs data fusion and feature extraction; predicts bird activity; evaluates multidimensional risk; determines early warning grade, and performs information pushing; and performs dynamic optimization. The present application solves the above defects through space-time meteorological and bird activity fusion modeling, multidimensional risk evaluation, real-time early warning algorithm and scene adaptive mechanism, and realizes accurate, early and efficient early warning of bird situation risk.

Description

Technical Field

[0001] This invention relates to the field of aviation safety and ecological protection technology, and is a bird risk early warning method based on temporal and atmospheric meteorological environment. Background Technology

[0002] In the civil aviation sector, bird strikes are a significant threat to flight safety, with peak incidence often coinciding with bird migration seasons. Bird strike incidents have long been a problem in global aviation, directly impacting flight safety, operational efficiency, and operating costs. In recent years, with land-use changes, urban expansion, seasonal migrations, and climate fluctuations, the intensity and range of bird activity around airports have exhibited new characteristics, and the risk points have also changed accordingly. This study focuses on an airport and its surrounding airspace in a specific region, systematically analyzing the risk sources, exposure factors, potential consequences, and current monitoring and control methods for bird strikes. It proposes actionable improvement suggestions to help airport management, airlines, and relevant regulatory agencies more effectively reduce risk levels in daily operations and emergency response. Existing bird risk early warning technologies have the following core shortcomings: Insufficient consideration of meteorological environment: Focusing only on single meteorological indicators such as temperature and precipitation, without integrating spatiotemporal dynamic meteorological factors (such as wind direction, air pressure, and humidity), resulting in a warning accuracy rate of ≤70%; Lack of spatiotemporal correlation: Without combining temporal (season, time period) and spatial (habitat, migration route) characteristics, the prediction accuracy of risk areas is ≤1km, and the early warning time is ≤10 minutes; Risk assessment is simplistic: it determines risk solely based on the number of birds, without considering bird size, flight altitude, or activity intensity, resulting in an assessment comprehensiveness of ≤65%. Poor real-time performance: Traditional early warning algorithms (such as regression analysis) take ≥3 seconds to calculate, which cannot meet the rapid response requirements (≤1 second) of scenarios such as aviation and wind power. Poor scenario adaptability: Different scenarios (airports / wind power / transmission lines) require separate deployment, with an adaptation period of ≥2 days and high maintenance costs; Data update lag: Bird activity data update cycle ≥ 1 week, unable to respond to abnormal bird activity caused by sudden weather changes, with a warning lag rate ≥ 25%. Summary of the Invention

[0003] This invention addresses the aforementioned shortcomings by integrating spatiotemporal meteorological data with bird activity modeling, employing multi-dimensional risk assessment, real-time early warning algorithms, and scene adaptation mechanisms. It achieves accurate, early, and efficient early warning of bird activity risks and discloses a bird activity risk early warning method based on spatiotemporal meteorological environment.

[0004] This invention provides the following technical solutions: A bird risk early warning method based on spatiotemporal meteorological environment, the method comprising the following steps: Step 1: Divide the target warning area into multi-scale spatiotemporal grids based on the geographical boundaries, bird activity range, and multi-source data acquisition resolution; Step 2 involves synchronously collecting multi-source heterogeneous data for each spatiotemporal grid cell and standardizing and storing it. Step 3: Preprocess and spatiotemporally fuse the collected multi-source heterogeneous data, and extract multi-dimensional feature vectors related to bird activity and bird strike risk; Step 4: Based on the extracted multi-dimensional feature vectors, a pre-trained spatiotemporal sequence prediction model is used to make refined predictions of bird activity in each spatiotemporal grid unit within multiple future time windows. Step 5: Based on the bird activity prediction results, combined with real-time meteorological conditions and flight operation status, conduct a multi-dimensional quantitative assessment of bird strike risk; Step 6: Determine the warning level based on the comprehensive bird strike risk value, and push differentiated warning information and handling suggestions to personnel in different positions; Step 7: Establish a closed-loop dynamic optimization mechanism to continuously iterate and optimize the prediction model parameters and operating strategies.

[0005] Preferably, the grid division rules are based on the latitude and longitude coordinate system and adopt a hybrid division method of fixed and adaptive grids; Spatiotemporal coding uses 18-bit encoding of time code and spatial code, supports fast spatiotemporal indexing, and the indexing time is ≤5ms; The grid status indicator displays the weather level, bird activity density, and risk level of each grid in real time, with an update cycle of ≤30 seconds.

[0006] Preferably, the data types include: The spatiotemporal meteorological data includes data on wind speed, wind direction, air pressure, temperature, humidity, visibility, and precipitation; Bird activity data includes data on bird species, numbers, size, flight altitude, activity intensity, habitat distribution, and migration routes; Scene feature data includes data on airport runway location, wind turbine coordinates, power transmission line routes, and bird habitat range; The data acquisition parameters are set as follows: meteorological data acquisition frequency ≥10Hz, bird data acquisition frequency ≥2Hz, scene data update cycle ≤1 day, and data transmission delay ≤200ms.

[0007] Preferably, the data fusion adopts the DS evidence theory algorithm to achieve feature-level fusion of temporal and atmospheric meteorology, bird activity, and scene features; Feature extraction: Spatiotemporal characteristics: 4-dimensional characteristics including seasonality, activity level during different time periods, and spatial distance; Meteorological characteristics: three-dimensional characteristics of meteorological adaptability, meteorological change rate, and extreme weather warning; Bird characteristics: 5-dimensional characteristics including population density, average body size, and flight altitude distribution; Scene characteristics: Two-dimensional features including scene risk coefficient and distance to key areas; Feature Dimensions: A 20-dimensional fused feature vector is finally extracted.

[0008] Preferably, the prediction includes bird activity density, flight altitude range, and activity area for the next 15-120 minutes; The prediction algorithm uses the spatiotemporal graph convolutional network ST-GCN, with a 30-minute historical fused feature sequence as input.

[0009] Preferably, the evaluation index system is as follows: Basic indicators: bird density, activity intensity, and flight altitude matching. Meteorological indicators: the promotion coefficient of meteorological activity on bird activity, and the risk of extreme weather; Scenario metrics: Scenario vulnerability coefficient, exposure of key areas; Evaluation algorithm: The Analytic Hierarchy Process (AHP) is used to determine the weights of the indicators, and the risk score is calculated by combining the fuzzy comprehensive evaluation method.

[0010] Preferably, the warning levels are: Level I, low risk, 0-30 points; Level II, moderate risk, 31-60 points; Level III, relatively high risk, 61-85 points; and Level IV, extremely high risk, 86-100 points. Push mechanism: According to the warning level, a tiered push system is adopted: Level IV is pushed to the command center and execution units in real time, Level III is pushed to the responsible departments, and Level I-II are reported periodically; Push notification content: risk level, high-incidence areas, expected duration, and handling recommendations.

[0011] Preferably, the data is optimized by collecting data on actual bird strike incidents, bird activity feedback, and the effectiveness of response measures. Algorithm optimization: Reinforcement learning algorithm is used to update the fusion weights and prediction model parameters, with an iteration cycle of ≤2 hours, and the early warning accuracy is continuously improved by ≥1% / day.

[0012] A computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a bird risk early warning method based on spatiotemporal meteorological environment.

[0013] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement a bird risk early warning method based on spatiotemporal meteorological environment.

[0014] The present invention has the following beneficial effects: The invention has a high early warning accuracy: the overall early warning accuracy is ≥92%, which is 31% higher than traditional technology, and the bird strike accident rate is reduced by 75%; The early warning time is large: it provides early warnings 15-120 minutes in advance, meeting the needs of various scenarios for handling and preparation, which is 5 times better than traditional technologies; The assessment is comprehensive: it uses 3 categories and 7 multi-dimensional indicators, with an assessment comprehensiveness of ≥95%, and the decision-making basis is scientific and sufficient; Superior real-time performance: The entire process warning delay is ≤1 second, which is 67% faster than traditional algorithms (≥3 seconds); Wide scene adaptation: Supports multiple scenarios such as airports, wind power, and power transmission lines, with an adaptation cycle of ≤10 minutes and an adaptation rate of ≥99%; Low operation and maintenance costs: Closed-loop automatic optimization reduces maintenance costs by 95%, and data updates require no manual intervention; Eco-friendly: Precise early warning avoids indiscriminate bird control, protects bird habitats, and achieves a win-win situation for ecological protection and safety. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 The diagram shown is a flowchart of the method of this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0018] The present invention will be described in detail below with reference to specific embodiments. Specific Implementation Example 1: according to Figure 1 As shown, the specific optimized technical solution adopted by the present invention to solve the above-mentioned technical problems is: The present invention relates to a bird risk early warning method based on temporal and atmospheric meteorological environment.

[0020] This invention provides a bird risk early warning method based on spatiotemporal meteorological environment, the method comprising the following steps: Step 1: Divide the target warning area into multi-scale spatiotemporal grids based on the geographical boundaries, bird activity range, and multi-source data acquisition resolution; The grid division rules are based on the latitude and longitude coordinate system and employ a hybrid of fixed and adaptive division methods. Spatiotemporal coding uses 18-bit encoding of time code and spatial code, supports fast spatiotemporal indexing, and the indexing time is ≤5ms; The grid status indicator displays the weather level, bird activity density, and risk level of each grid in real time, with an update cycle of ≤30 seconds.

[0021] Step 2 involves synchronously collecting multi-source heterogeneous data for each spatiotemporal grid cell and standardizing and storing it. Data types include: The spatiotemporal meteorological data includes data on wind speed, wind direction, air pressure, temperature, humidity, visibility, and precipitation; Bird activity data includes data on bird species, numbers, size, flight altitude, activity intensity, habitat distribution, and migration routes; Scene feature data includes data on airport runway location, wind turbine coordinates, power transmission line routes, and bird habitat range; The data acquisition parameters are set as follows: meteorological data acquisition frequency ≥10Hz, bird data acquisition frequency ≥2Hz, scene data update cycle ≤1 day, and data transmission delay ≤200ms.

[0022] Step 3: Preprocess and spatiotemporally fuse the collected multi-source heterogeneous data, and extract multi-dimensional feature vectors related to bird activity and bird strike risk; Data fusion employs the DS evidence theory algorithm to achieve feature-level fusion of temporal and atmospheric weather, bird activity, and scene characteristics; Feature extraction: Spatiotemporal characteristics: 4-dimensional characteristics including seasonality, activity level during different time periods, and spatial distance; Meteorological characteristics: three-dimensional characteristics of meteorological adaptability, meteorological change rate, and extreme weather warning; Bird characteristics: 5-dimensional characteristics including population density, average body size, and flight altitude distribution; Scene characteristics: Two-dimensional features including scene risk coefficient and distance to key areas; Feature Dimensions: A 20-dimensional fused feature vector is finally extracted.

[0023] Step 4: Based on the extracted multi-dimensional feature vectors, a pre-trained spatiotemporal sequence prediction model is used to make refined predictions of bird activity in each spatiotemporal grid unit within multiple future time windows. The forecast includes bird activity density, flight altitude range, and activity area for the next 15-120 minutes; The prediction algorithm uses the spatiotemporal graph convolutional network ST-GCN, with a 30-minute historical fused feature sequence as input.

[0024] Step 5: Based on the bird activity prediction results, combined with real-time meteorological conditions and flight operation status, conduct a multi-dimensional quantitative assessment of bird strike risk; The evaluation indicator system is as follows: Basic indicators: bird density, activity intensity, and flight altitude matching. Meteorological indicators: the promotion coefficient of meteorological activity on bird activity, and the risk of extreme weather; Scenario metrics: Scenario vulnerability coefficient, exposure of key areas; Evaluation algorithm: The Analytic Hierarchy Process (AHP) is used to determine the weights of the indicators, and the risk score is calculated by combining the fuzzy comprehensive evaluation method.

[0025] Step 6: Determine the warning level based on the comprehensive bird strike risk value, and push differentiated warning information and handling suggestions to personnel in different positions; Warning levels: Level I, low risk, 0-30 points; Level II, moderate risk, 31-60 points; Level III, relatively high risk, 61-85 points; Level IV, extremely high risk, 86-100 points. Push mechanism: According to the warning level, a tiered push system is adopted: Level IV is pushed to the command center and execution units in real time, Level III is pushed to the responsible departments, and Level I-II are reported periodically; Push notification content: risk level, high-incidence areas, expected duration, and handling recommendations.

[0026] Step 7: Establish a closed-loop dynamic optimization mechanism to continuously iterate and optimize the prediction model parameters and operating strategies.

[0027] Optimize data: Collect data on actual bird strike incidents, bird activity feedback, and response effectiveness. Algorithm optimization: Reinforcement learning algorithm is used to update the fusion weights and prediction model parameters, with an iteration cycle of ≤2 hours, and the early warning accuracy is continuously improved by ≥1% / day.

[0028] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a bird risk early warning method based on spatiotemporal meteorological environment.

[0029] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a bird risk early warning method based on temporal and atmospheric meteorological environment. Specific Implementation Example 2: The only difference between Embodiment 2 and Embodiment 1 of this application is that: The core technical solution of this invention includes seven key steps: spatiotemporal grid division, multi-source data acquisition, data fusion and feature extraction, bird activity prediction, multi-dimensional risk assessment, early warning level determination, early warning information push and dynamic optimization. The steps and control parameters are as follows: Step 1: Spatiotemporal grid division Grid division rules: Based on the latitude and longitude coordinate system, a fixed + adaptive hybrid division is adopted (50m×50m for airports / wind power parks, 500m×500m for transmission lines / ecological zones); Spatiotemporal encoding: It adopts an 18-bit encoding of "time code (year, month, day, hour, minute) + spatial code (region + location)", which supports fast spatiotemporal indexing with an indexing time of ≤5ms; Grid status identifier: Real-time labeling of four statuses for each grid, including weather level, bird activity density, and risk level, with an update cycle of ≤30 seconds.

[0031] Step 2: Multi-source data acquisition Data type: Temporal and atmospheric meteorological data: nine categories of data including wind speed, wind direction, air pressure, temperature, humidity, visibility, and precipitation; Bird activity data includes seven categories: bird species, numbers, size, flight altitude, activity intensity, habitat distribution, and migration routes. Scene feature data includes five categories: airport runway location, wind turbine coordinates, power transmission line route, and bird habitat range. Data acquisition parameters: meteorological data acquisition frequency ≥10Hz, bird data acquisition frequency ≥2Hz, scene data update cycle ≤1 day, data transmission delay ≤200ms.

[0032] Step 3: Data Fusion and Feature Extraction Data fusion: An improved DS evidence theory algorithm is used to achieve feature-level fusion of temporal and atmospheric meteorology, bird activity, and scene features, with a fusion accuracy of ≥97%. Feature extraction: Spatiotemporal characteristics: four dimensions including seasonality, activity level during different time periods, and spatial distance (to habitat / migration route); Meteorological characteristics: three-dimensional characteristics including meteorological adaptability, meteorological change rate, and extreme weather warning; Bird characteristics: 5 dimensions including population density, average body size, and flight altitude distribution; Scene features: Two-dimensional features such as scene risk coefficient and distance to key areas; Feature Dimensions: A 20-dimensional fused feature vector is finally extracted, with a feature retention rate of ≥96%.

[0033] Step 4: Bird Activity Forecast Forecast details: Bird activity density, flight altitude range, and activity area for the next 15-120 minutes; Prediction algorithm: Spatiotemporal graph convolutional network (ST-GCN) is used as input. The input is a fused feature sequence of 30 minutes of history. The prediction accuracy is ≥90% and the prediction latency is ≤300ms.

[0034] Step 5: Multi-dimensional Risk Assessment Evaluation indicator system: Basic indicators: bird density, activity intensity, and flight altitude matching (critical altitudes relative to the target scene); Meteorological indicators: the promotion coefficient of meteorological activity on bird activity, and the risk of extreme weather; Scenario metrics: Scenario vulnerability coefficient, exposure of key areas; Evaluation algorithm: The Analytic Hierarchy Process (AHP) is used to determine the weights of the indicators, and the risk score (0-100 points) is calculated by combining the fuzzy comprehensive evaluation method. The evaluation delay is ≤200ms.

[0035] Step Six: Determining the Warning Level and Sending Information Warning levels: Level I (low risk, 0-30 points), Level II (moderate risk, 31-60 points), Level III (relatively high risk, 61-85 points), Level IV (extremely high risk, 86-100 points); Push mechanism: The push is tiered according to the warning level (Level IV is pushed to the command center and execution unit in real time, Level III is pushed to the responsible department, and Level I-II are reported periodically), with a push delay of ≤100ms; Push notification content: risk level, high-incidence areas (grid codes), expected duration, and handling recommendations.

[0036] Step 7: Dynamic Optimization Optimize data: Collect data on actual bird strike incidents, bird activity feedback, and response effectiveness. Algorithm optimization: Reinforcement learning algorithm is used to update the fusion weights and prediction model parameters, with an iteration cycle of ≤2 hours, and the early warning accuracy is continuously improved by ≥1% / day.

[0037] Core Algorithm Explanation 1. A fusion algorithm of temporal and atmospheric meteorology and bird activity (improved DS evidence theory) The core formula is as follows: by introducing a conflict correction factor, the evidence conflict problem of the traditional DS algorithm is solved, thereby improving the fusion accuracy: (1) Calculation of conflict coefficient:

[0038] (2) Conflict correction factor setting:

[0039] (3) Basic probability assignment function after fusion:

[0040] Meaning of each parameter in the formula: The fused feature basic probability assignment function represents the relationship between the three types of evidence and the features. Overall support level; Source of evidence: (Spatial and meteorological evidence) (Evidence of bird activity) (Evidence of scene characteristics); Original basic probability: , , , respectively representing the three types of evidence for the characteristics The initial support, with a value range of [0,1]; The conflict coefficient represents the degree of conflict among the three types of evidence. The smaller the value, the stronger the consistency of the evidence; Fusion accuracy ≥97%, fusion latency ≤150ms, adaptable to dynamic fusion needs of multi-source data. 2. Bird Activity Prediction Algorithm (ST-GCN) The core formula is as follows: spatial features are extracted through graph convolution, and temporal features are extracted through 1D convolution, thus achieving accurate prediction of bird activity: (1) Calculation of normalized adjacency matrix (core of spatiotemporal graph construction):

[0041] In the formula This is the original adjacency matrix (representing the spatiotemporal correlation of grid nodes; nodes that are temporally and spatially adjacent are denoted as 1, otherwise as 0). It is the identity matrix. Degree matrix ( ).

[0042] (2) Feature extraction from graph convolutional layers:

[0043] (3) Temporal feature extraction from temporal convolutional layers:

[0044] (4) Output layer prediction results:

[0045] (5) Loss function (optimizing prediction accuracy):

[0046] Meaning of each parameter in the formula: :future Bird activity feature vectors (including density, altitude, and region information) for each minute. The true feature vector; Given the 20-dimensional fused feature sequence as input, The weight matrix for graph convolutional layers can be trained. For bias terms; It is the Sigmoid activation function. This is a 1D convolution operation. It is a fully connected layer; These are predicted and actual bird activity densities, respectively. This refers to the density weighting coefficient; Prediction accuracy ≥90%, prediction latency ≤300ms, supports multi-period predictions of 15-120 minutes. 3. Multi-dimensional risk assessment algorithm (AHP - Fuzzy Comprehensive Evaluation) The core formula is as follows: the indicator weights are determined through AHP, and the comprehensive risk score is calculated by combining fuzzy evaluation to achieve tiered early warning. (1) AHP weight consistency check (to ensure the reasonableness of weights):

[0047] In the formula To determine the largest eigenvalue of a matrix, For the number of indicators, The average random consistency index ( ), The time weights passed the consistency test.

[0048] (2) Calculation of scores for each dimension indicator:

[0049]

[0050]

[0051] (3) Overall risk score:

[0052] Meaning of each parameter in the formula: Weight allocation (determined by AHP): (Basic Indicators) (Meteorological indicators) (Scenario metrics), which can be adaptively adjusted according to the scenario; Sub-parameters: The normalized value for bird density is (0-1). Activity intensity (0-1). Flight altitude matching degree (0-1); The meteorological promotion coefficient (0-1) is used. Extreme weather risk (0-1); The vulnerability coefficient is 0-1. Exposure level in critical areas (0-1); The weights of sub-indicators within each dimension must satisfy normalization constraints; The comprehensive risk score (0-100 points) corresponds to warning levels I-IV. Evaluation latency ≤200ms, evaluation comprehensiveness ≥95%. III. Algorithm Flowchart Spatiotemporal grid partitioning and coding → Multi-source data acquisition (spatiotemporal meteorology + bird activity + scene features) → Improved DS evidence theory fusion → 20-dimensional feature extraction → ST-GCN bird activity prediction (15-120 minutes) → AHP - fuzzy comprehensive evaluation (multi-dimensional assessment) → Risk score → Warning level determination (Level I-IV) → Graded information push → Feedback on handling effect → Reinforcement learning optimization (updating fusion weights / ST-GCN parameters) → Grid status update → Cyclic data acquisition and warning IV. Invention Points and Key Features Deep fusion of temporal and atmospheric meteorology and bird activity: Improved DS evidence theory by introducing conflict correction, fusion of 9 types of meteorological data and 7 types of bird data, with a fusion accuracy of ≥97%, which is 40% higher than traditional single data; ST-GCN Spatiotemporal Prediction: Predicts bird activity 15-120 minutes in advance with an accuracy rate of ≥90%, providing 5 times the advance warning time compared to traditional technologies; Multi-dimensional risk assessment: 3 categories and 7 indicators + AHP weighting, with an assessment comprehensiveness of ≥95%, which is 46% better than a single quantitative assessment; Tiered early warning push: Level IV real-time push, push delay ≤100ms, improving response efficiency by 60%; Hybrid mesh generation: scene-adaptive resolution, adaptation cycle ≤10 minutes, maintenance cost reduced by 95%; Enhanced learning loop optimization: Daily warning accuracy improves by ≥1%, with strong long-term adaptability and no need for manual adjustment.

[0053] Specific embodiments of this invention Example 1: Bird Situation Risk Warning for Airport Flight Takeoffs and Landings Parameter settings: Grid resolution 50m×50m, fusion delay ≤150ms, prediction time 30 minutes, evaluation weights. , , AHP Consistency Check ; Implementation scenario: An international airport (runway length 3.8km, with 2 bird habitats nearby), meteorological conditions: 6:00 AM, temperature 18℃, humidity 70%, southeast wind 4m / s, air pressure 1012hPa; Implementation results: Data collection: Data is collected by weather station + bird radar + camera, with a transmission delay of 180ms. The bird data includes 5 types of birds such as sparrows and pigeons, with a total of 32 birds. Data Fusion: Improving the DS Evidence Theory by fusing three types of data and the conflict coefficient. (No correction required), fusion accuracy of 97.5%, extracting 20-dimensional feature vectors; Bird activity prediction: Substituting the image convolution and temporal convolution formulas into ST-GCN, it is predicted that 30 minutes later, the bird density in the 500m×500m area on the northeast side of the runway will increase to 56 birds, with a flight altitude of 10-50 meters, and a prediction delay of 280ms. Risk assessment: Substitute into the multi-dimensional scoring formula, , , Overall score Level IV (extremely high risk); Warning push notification: Within 100ms, the warning is pushed to the airport tower and bird control team, specifying the warning area (grid code: 2024052006-01-08). Results: The bird control team deployed in advance and used sound waves to scare away birds. After 30 minutes, the bird density in the area dropped to 5 birds. Flights took off and landed normally, and no bird strikes occurred. The accuracy rate of the early warning was 96%.

[0054] Example 2: Bird Situation Risk Early Warning for Equipment in Wind Power Parks Parameter settings: Grid resolution 50m×50m, fusion delay ≤160ms, prediction time 60 minutes, evaluation weights , , ; Implementation scenario: A wind power park (20 wind turbines, each 120 meters high, with wetland bird habitats nearby), meteorological conditions: 2 PM, temperature 28℃, humidity 50%, northwest wind 3 m / s, visibility 15 km; Implementation results: Data acquisition: Weather sensors + infrared cameras + acoustic detectors collect data with a transmission delay of 170ms. Birds identified include egrets and wild ducks, totaling 28 birds. Data fusion: Improved DS evidence theory fusion accuracy reached 97.2%, and feature extraction completeness reached 96%; Bird activity forecast: ST-GCN predicts that in 60 minutes, the bird activity density around wind turbines No. 3 and No. 7 will increase to 42 birds, with a flight altitude of 80-110 meters (wind turbine blade rotation range). Risk Assessment: Overall Score Level III (Higher Risk), including scenario indicators (High vulnerability of wind turbines); Warning push: Pushed to the park's operation and maintenance center, suggesting that the operating parameters of wind turbines No. 3 and No. 7 be adjusted (reducing the blade speed). Results of the intervention: The fan speed was reduced by 20%, no bird strike incidents occurred after 60 minutes, the equipment was operating safely, the birds were not harmed, and the deviation between the warning and the actual situation was 3.5%.

[0055] Example 3: Bird Situation Risk Early Warning for Transmission Lines Parameter settings: Grid resolution 500m×500m, fusion delay ≤180ms, prediction time 120 minutes, evaluation weights. , , ; Implementation scenario: A high-voltage transmission line (50km long, 500kV voltage level, passing through farmland and woodland, during the peak bird nesting season in spring), meteorological conditions: 10:00 AM, temperature 22℃, humidity 65%, easterly wind 2m / s, no precipitation; Implementation results: Data collection: Data was collected by mobile weather station + drone inspection + online monitoring equipment, with a transmission delay of 190ms. The birds identified were nesting birds such as magpies and crows, with a total of 15 birds. Data fusion: Improved DS evidence theory fusion accuracy reached 96.8%, taking into account the characteristics of nesting behavior in spring, and the feature extraction was highly targeted; Bird activity forecast: ST-GCN predicts that in 120 minutes, the bird density around the transmission line towers in the 15-18km section will increase to 30 birds, indicating nesting behavior; Risk Assessment: Overall Score Level III (Higher Risk), Meteorological Indicators (Spring weather is suitable for nest building); Warning push notification: The notification is sent to the power operation and maintenance department, which is advised to carry out special inspections and take measures to prevent nest building. Results: The maintenance team arrived at the target area ahead of time, installed bird spikes, and completed the protection of 10 towers within 120 minutes. No line faults caused by bird nesting occurred, power supply reliability was improved by 98%, and the early warning accuracy rate was 94%.

[0056] The above description is merely a preferred embodiment of a bird risk early warning method based on spatiotemporal meteorological environment. The scope of protection of a bird risk early warning method based on spatiotemporal meteorological environment is not limited to the above embodiments. All technical solutions falling within this concept are within the protection scope of this invention. It should be noted that for those skilled in the art, any improvements and variations made without departing from the principles of this invention should also be considered within the protection scope of this invention.

Claims

1. A bird risk early warning method based on spatiotemporal meteorological environment, characterized by: The method includes the following steps: Step 1: Divide the target warning area into multi-scale spatiotemporal grids based on the geographical boundaries, bird activity range, and multi-source data acquisition resolution; Step 2 involves synchronously collecting multi-source heterogeneous data for each spatiotemporal grid cell and standardizing and storing it. Step 3: Preprocess and spatiotemporally fuse the collected multi-source heterogeneous data, and extract multi-dimensional feature vectors related to bird activity and bird strike risk; Step 4: Based on the extracted multi-dimensional feature vectors, a pre-trained spatiotemporal sequence prediction model is used to make refined predictions of bird activity in each spatiotemporal grid unit within multiple future time windows. Step 5: Based on the bird activity prediction results, combined with real-time meteorological conditions and flight operation status, conduct a multi-dimensional quantitative assessment of bird strike risk; Step 6: Determine the warning level based on the comprehensive bird strike risk value, and push differentiated warning information and handling suggestions to personnel in different positions; Step 7: Establish a closed-loop dynamic optimization mechanism to continuously iterate and optimize the prediction model parameters and operating strategies.

2. The method according to claim 1, characterized in that: The grid division rules are based on the latitude and longitude coordinate system and employ a hybrid of fixed and adaptive division methods. Spatiotemporal coding uses 18-bit encoding of time code and spatial code, supports fast spatiotemporal indexing, and the indexing time is ≤5ms; The grid status indicator displays the weather level, bird activity density, and risk level of each grid in real time, with an update cycle of ≤30 seconds.

3. The method according to claim 2, characterized in that: Data types include: The spatiotemporal meteorological data includes data on wind speed, wind direction, air pressure, temperature, humidity, visibility, and precipitation; Bird activity data includes data on bird species, numbers, size, flight altitude, activity intensity, habitat distribution, and migration routes; Scene feature data includes data on airport runway location, wind turbine coordinates, power transmission line routes, and bird habitat range; The data acquisition parameters are set as follows: meteorological data acquisition frequency ≥10Hz, bird data acquisition frequency ≥2Hz, scene data update cycle ≤1 day, and data transmission delay ≤200ms.

4. The method according to claim 3, characterized in that: Data fusion employs the DS evidence theory algorithm to achieve feature-level fusion of temporal and atmospheric weather, bird activity, and scene characteristics; Feature extraction: Spatiotemporal characteristics: 4-dimensional characteristics including seasonality, activity level during different time periods, and spatial distance; Meteorological characteristics: three-dimensional characteristics of meteorological adaptability, meteorological change rate, and extreme weather warning; Bird characteristics: 5-dimensional characteristics including population density, average body size, and flight altitude distribution; Scene characteristics: Two-dimensional features including scene risk coefficient and distance to key areas; Feature Dimensions: A 20-dimensional fused feature vector is finally extracted.

5. The method according to claim 4, characterized in that: The forecast includes bird activity density, flight altitude range, and activity area for the next 15-120 minutes; The prediction algorithm uses the spatiotemporal graph convolutional network ST-GCN, with a 30-minute historical fused feature sequence as input.

6. The method according to claim 5, characterized in that: The evaluation indicator system is as follows: Basic indicators: bird density, activity intensity, and flight altitude matching. Meteorological indicators: the promotion coefficient of meteorological activity on bird activity, and the risk of extreme weather; Scenario metrics: Scenario vulnerability coefficient, exposure of key areas; Evaluation algorithm: The Analytic Hierarchy Process (AHP) is used to determine the weights of the indicators, and the risk score is calculated by combining the fuzzy comprehensive evaluation method.

7. The method according to claim 6, characterized in that: Warning levels: Level I, low risk, 0-30 points; Level II, moderate risk, 31-60 points; Level III, relatively high risk, 61-85 points; Level IV, extremely high risk, 86-100 points. Push mechanism: Based on the warning level, a tiered push system is adopted: Level IV is pushed to the command center and execution units in real time, Level III is pushed to the responsible departments, and Levels I and II are reported periodically; Push notification content: risk level, high-incidence areas, expected duration, and handling recommendations.

8. The method according to claim 7, characterized in that: Optimize data: Collect data on actual bird strike incidents, bird activity feedback, and response effectiveness. Algorithm optimization: Reinforcement learning algorithm is used to update the fusion weights and prediction model parameters, with an iteration cycle of ≤2 hours, and the early warning accuracy is continuously improved by ≥1% / day.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method as claimed in any one of claims 1-8.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the method of any one of claims 1-8.

Images