Airport photovoltaic glare detection method based on multi-source data coupling

Abstract

The invention discloses an airport photovoltaic glare detection method based on multi-source data coupling, which breaks through the limitation of single model prediction through multi-source data fusion, actually measured data provides a dynamic calibration reference for simulation, and a machine learning algorithm effectively compensates errors caused by environmental variable fluctuation under physical constraints. A continuous learning mechanism replaces a traditional regular manual calibration process, and self-adaptive evolution of the model is achieved while the operation and maintenance cost is reduced. The three-dimensional scene projection and space-time matching technology solves the problem of collaborative analysis of multi-source heterogeneous data, provides a high-precision spatialization expression basis for risk assessment, and provides a high-precision space reference for subsequent physical simulation and actual measurement data coupling by integrating multi-source data such as runways, taxiways, terrains, vision fields and flight paths. Therefore, the dependence on high-frequency field measurement is reduced.

Description

Technical Field

[0001] This application relates to the field of aerospace technology, and in particular to a method for detecting photovoltaic glare at airports based on multi-source data coupling. Background Technology

[0002] The introduction of photovoltaic (PV) power generation systems has provided an effective way to solve airport energy consumption problems. However, the reflected glare from the surface of PV modules can potentially interfere with the vision of operators in core airport operational support facilities and critical safety areas, posing a significant aviation safety risk. Therefore, developing accurate and efficient methods for detecting and assessing airport PV glare has become an indispensable key technology for ensuring safe airport operations.

[0003] Currently, model prediction methods for detecting photovoltaic glare at airports generally suffer from the core pain points of difficulty in verifying model prediction accuracy and high calibration costs. Traditional methods rely heavily on theoretical models to simulate glare paths throughout the year, which can cover a wide range of time and space, but lack effective field measurement data for dynamic calibration and verification, leading to doubts about the reliability of model prediction results. Attempting to obtain extensive "anchor point" data through intensive or high-frequency field measurements to verify and calibrate the model is not only extremely costly, but also presents operational feasibility and safety challenges in sensitive airport areas. Summary of the Invention

[0004] The main objective of this application is to provide a method for detecting photovoltaic glare at airports based on multi-source data coupling, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this application provides the following technical solution: A method for detecting photovoltaic glare at airports based on multi-source data coupling, comprising the following steps: S1. Establish baselines for testing requirements and site elements, determine the target area for airport operation support, key safety viewpoints and permissible glare levels, and combine airport topographic data, runway orientation, elevation model, airway approach profile, tower visibility constraints and operational restrictions to form testing baseline information; S2. Construct a multi-source data acquisition system, which includes meteorological data, optical observation data, spectral and irradiance sensor data, photovoltaic module reflection characteristic parameters, flight trajectory data and solar position parameters, and deploy fixed or mobile sampling devices in the permitted area of ​​the airport to obtain glare observation information. S3. Perform preprocessing and quality control on the collected data, including time synchronization, coordinate registration, outlier removal, noise filtering, radiometric correction and 3D scene projection, and generate standardized observation products based on the reliability of the observation data. S4. Based on the geometric parameters, reflection characteristics and solar position parameters of photovoltaic modules, construct a physical optics simulation model, perform ray tracing or analytical calculation on potential glare paths, obtain the spatiotemporal distribution results that may produce glare, and combine meteorological conditions to correct the reflection intensity. S5. Couple the preprocessed measured data with the physical simulation prediction results to build a fusion model. Use the physical constraint machine learning method to correct and quantify the uncertainty of the prediction results, and output the probability distribution of glare events and the calibrated brightness prediction value. S6. Based on the output of the fusion model, establish a risk assessment and alarm mechanism, classify and judge glare events, generate dynamic alarms and control suggestions in combination with flight operation characteristics, and output risk indicators and event spatiotemporal distribution maps through a visual interface. S7. Conduct pilot verification in typical areas of the airport, collect operational data to compare and calibrate the model prediction results, and establish a continuous learning and data governance mechanism to enable the method to achieve online adaptive updates and long-term application during operation.

[0006] Preferably, step S1 is performed in the following manner: S1.1 Collect and summarize airport operation and site-related data, including runway number, length, width and azimuth data, taxiway spatial coordinate data, elevation values ​​in the airport digital elevation model, tower latitude and longitude coordinates and height values, tower azimuth and elevation range values, spatial coordinate sequences of approach and departure paths, elevation values ​​of the terrain around the airport, latitude and longitude coordinates and height values ​​of obstacles around the airport, spatial coordinates of key safety viewpoints, boundary coordinates of the airport operation support area, brightness thresholds for permissible glare levels, and time periods and area ranges for airport operation restrictions. S1.2 The collected data are classified and modeled, and the operation and maintenance requirements and site elements are modeled in a unified manner to generate baseline information for subsequent photovoltaic glare detection.

[0007] Preferably, step S2 is performed as follows: S2.1 Configure and deploy multi-source data acquisition devices, install fixed meteorological sensors, optical observation equipment, spectral and irradiance measurement devices and solar position calculation devices in the permitted area of ​​the airport, and deploy mobile sampling devices near photovoltaic modules to obtain real-time observation data including air temperature, relative humidity, wind speed, wind direction, visibility, solar irradiance, spectral reflectance, sky brightness and solar azimuth angle, and at the same time obtain specular reflectance, diffuse reflectance and surface temperature values ​​of the photovoltaic module surface; S2.2 Record and summarize multi-source observation data. The meteorological measurement results, optical observation results, spectral and irradiance measurement results, photovoltaic module reflection characteristic measurement results, aircraft take-off and landing flight trajectory coordinate sequence and solar position calculation results are uniformly formatted to form multi-source observation data for glare detection.

[0008] Preferably, step S3 is performed as follows: S3.1. Perform time synchronization and spatial coordinate registration on the collected meteorological data, optical observation data, spectral and irradiance measurement data, photovoltaic module reflection characteristic data, flight trajectory coordinate data and solar position data, and perform outlier detection and removal, noise filtering and radiation intensity correction processing. S3.2 Project the processed data into a three-dimensional scene according to a unified format, calculate the data reliability based on the measurement accuracy and repeatability of the observation data, and generate a standardized observation dataset that can be used for subsequent photovoltaic glare analysis.

[0009] Preferably, step S4 is performed as follows: S4.1 Based on the geometric dimensions, tilt angle, azimuth angle, arrangement spacing and surface reflectivity measurements of photovoltaic modules, combined with the real-time altitude angle, azimuth angle of the sun and the latitude and longitude coordinates of the observation area, establish a ray tracing calculation model or analytical calculation model based on physical optics principles to simulate and calculate the light reflection path of photovoltaic modules under different time and space conditions. S4.2 Based on the solar irradiance, sky brightness, air temperature, air humidity and atmospheric transmittance data obtained from meteorological observations, the reflection intensity output by the ray tracing calculation model is numerically corrected, and the coordinates, duration and corresponding reflection brightness values ​​of the areas where glare may occur in different time periods and spatial ranges are calculated and output.

[0010] Preferably, step S5 is performed as follows: S5.1. Match the meteorological observation data, optical observation data, spectral measurement data, irradiance measurement data, photovoltaic module reflection measurement data, flight trajectory coordinate data and solar position data that have been processed by time synchronization, spatial registration and quality control with the glare prediction data obtained by physical optics simulation calculation to establish a fusion modeling dataset that includes measured data and simulation data. S5.2 Based on the fusion modeling dataset, a machine learning prediction model with physical constraints is constructed. The glare prediction values ​​output by the physical optics simulation calculation are numerically corrected and the uncertainty is quantified to generate the temporal distribution, spatial distribution and calibrated glare brightness prediction results of glare events for subsequent glare risk assessment.

[0011] Preferably, step S6 is performed as follows: S6.1 Based on the temporal distribution value, spatial distribution value, and glare brightness prediction value of the glare event output by the fusion model, combined with the flight take-off and landing time, flight route coordinate sequence, runtime information, and real-time weather conditions, the degree of impact of the glare event on flight operation is quantitatively calculated, a glare brightness threshold and a glare duration threshold are set, the glare event is graded, and a corresponding risk assessment result set is generated. S6.2 Based on the risk assessment result set and flight operation characteristics, generate an alarm dataset containing glare event level, boundary coordinates of the affected area, time interval of the affected area, and recommended control measures. Establish a risk alarm and control mechanism, and output risk indicator values, time distribution map and spatial distribution map of glare events through a human-computer interaction interface.

[0012] Preferably, step S7 is performed as follows: S7.1. Conduct pilot verification operations within the selected typical operating area of ​​the airport, collect flight trajectory data, glare observation and measurement data, meteorological monitoring data, and control measure implementation effect data, compare the collected operating data with the prediction results output by the fusion model one by one, and calibrate the prediction results based on the comparison differences to obtain a calibrated prediction result set. S7.2 Based on the aforementioned operational data and the calibrated prediction result set, establish a data governance mechanism that includes data cleaning rules, quality control criteria, and version management strategies, and construct a model update process with continuous learning capabilities, so that the fusion model can perform online adaptive updates during operation and maintain the stability and effectiveness of long-term application.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. This solution overcomes the limitations of single-model prediction by fusing multi-source data. Measured data provides a dynamic calibration benchmark for simulation, and machine learning algorithms effectively compensate for errors caused by fluctuations in environmental variables under physical constraints. A continuous learning mechanism replaces the traditional periodic manual calibration process, reducing maintenance costs while enabling adaptive model evolution. 3D scene projection and spatiotemporal matching technologies solve the challenge of collaborative analysis of multi-source heterogeneous data, providing a high-precision spatial representation basis for risk assessment.

[0014] 2. This application addresses the difficulty in model calibration caused by the single dimension of traditional detection baseline information. By integrating multi-source data such as runway, taxiway, terrain, line-of-sight, and flight path, it provides a high-precision spatial reference for the subsequent coupling of physical simulation and measured data, thereby reducing the reliance on high-frequency field measurements. For example, the spatiotemporal constraints generated based on the flight path coordinate sequence and runtime limitations can be directly used to narrow the ray tracing calculation range, reduce the computational resource consumption in invalid simulation areas, and improve the matching efficiency between model prediction results and actual observation data. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the method described in this application. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0017] The terms "first," "second," and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0018] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] Example 1: Please refer to Figure 1 A method for detecting photovoltaic glare at airports based on multi-source data coupling, the specific steps of which are as follows: S1. Establish baselines for testing requirements and site elements, determine the target area for airport operation support, key safety viewpoints and permissible glare levels, and combine airport topographic data, runway orientation, elevation model, airway approach profile, tower visibility constraints and operational restrictions to form testing baseline information; S2. Construct a multi-source data acquisition system, which includes meteorological data, optical observation data, spectral and irradiance sensor data, photovoltaic module reflection characteristic parameters, flight trajectory data and solar position parameters, and deploy fixed or mobile sampling devices in the permitted area of ​​the airport to obtain glare observation information. S3. Perform preprocessing and quality control on the collected data, including time synchronization, coordinate registration, outlier removal, noise filtering, radiometric correction and 3D scene projection, and generate standardized observation products based on the reliability of the observation data. S4. Based on the geometric parameters, reflection characteristics and solar position parameters of photovoltaic modules, construct a physical optics simulation model, perform ray tracing or analytical calculation on potential glare paths, obtain the spatiotemporal distribution results that may produce glare, and combine meteorological conditions to correct the reflection intensity. S5. Couple the preprocessed measured data with the physical simulation prediction results to build a fusion model. Use the physical constraint machine learning method to correct and quantify the uncertainty of the prediction results, and output the probability distribution of glare events and the calibrated brightness prediction value. S6. Based on the output of the fusion model, establish a risk assessment and alarm mechanism, classify and judge glare events, generate dynamic alarms and control suggestions in combination with flight operation characteristics, and output risk indicators and event spatiotemporal distribution maps through a visual interface. S7. Conduct pilot verification in typical areas of the airport, collect operational data to compare and calibrate the model prediction results, and establish a continuous learning and data governance mechanism to enable the method to achieve online adaptive updates and long-term application during operation.

[0020] In this embodiment: In the prior art, the application of photovoltaic power generation systems in airports faces the challenge of reflected glare interfering with aviation safety. Traditional detection methods rely on theoretical models to simulate glare paths throughout the year. Although this can cover a wide range of time and space, it lacks dynamic calibration with measured data, resulting in insufficient reliability of the prediction results. If the model is verified through high-frequency on-site measurements, it faces problems such as high cost and operational restrictions in sensitive airport areas, making it difficult to achieve accurate assessment and risk management.

[0021] To address these issues, researchers discovered that the disconnect between model predictions and measured data is the core reason for insufficient accuracy. Analysis revealed that relying solely on theoretical models cannot reflect actual weather changes and equipment status fluctuations, while relying only on localized measured data is insufficient to cover complex spatiotemporal scenarios. Therefore, a proposal was made to dynamically couple multi-source measured data with a physical simulation model, utilize machine learning to correct model biases online, and construct a data governance mechanism for continuous optimization, thereby improving prediction reliability with limited measured resources.

[0022] Therefore, this application proposes the following technical solutions: establishing a baseline for detection requirements and site elements, determining the airport operation support area, safe viewpoint, and permissible glare level, and forming a detection baseline by combining topographic, runway orientation, elevation model, and other data; constructing a multi-source data acquisition system including meteorological, optical, spectral, flight trajectory, and solar position data; performing time synchronization, coordinate registration, outlier processing, and 3D scene projection on the data; constructing a physical optical model based on photovoltaic module parameters and solar position, and correcting the reflection intensity based on meteorological conditions; coupling measured data with simulation results, and using physically constrained machine learning for correction and uncertainty quantification; establishing a risk assessment and alarm mechanism, and generating dynamic control suggestions based on flight characteristics; and achieving online model updates through pilot verification and continuous learning.

[0023] The detection baseline information refers to a benchmark dataset formed by integrating the boundary coordinates of the airport operation support target area, the spatial location of key safety viewpoints, and glare level thresholds. Specifically, it can be achieved by fusing parameters such as runway azimuth and tower visibility elevation angle through a geographic information system, used to limit the detection range and judgment criteria. The multi-source data acquisition system includes devices such as meteorological sensors, spectrometers, and flight trajectory recorders. Specifically, it can be implemented by combining fixed meteorological stations with mobile UAV-borne measurement equipment, used to acquire multi-dimensional environmental and equipment status data. The physical optics simulation model refers to a computational framework based on ray tracing algorithms to simulate the reflection path of photovoltaic modules. Specifically, it can be implemented using Monte Carlo ray tracing combined with atmospheric transmittance correction algorithms, used to predict the spatiotemporal distribution of potential glare. The data coupling mechanism refers to the process of spatiotemporally matching standardized measured data with simulation prediction results. Specifically, it can be achieved through timestamp alignment and geographic coordinate interpolation algorithms, used to construct a fused modeling dataset. The continuous learning mechanism refers to the optimization process of dynamically updating the model based on operational data. Specifically, it can be implemented using incremental learning algorithms combined with data version control strategies, used to maintain the long-term predictive accuracy of the model.

[0024] First, geographic information integration clarifies the detection area and safety threshold, providing spatial constraints for subsequent analysis. A multi-source sensor network collects real-time meteorological, optical, and equipment status data, which are then denoised and corrected to form a standardized observation set. A physical model is constructed based on the geometric parameters of photovoltaic modules and the solar trajectory to simulate the reflection path distribution under different meteorological conditions. Measured data and simulation results are spatiotemporally matched and input into a machine learning model. Physical equations constrain the neural network training process, outputting a calibrated glare probability distribution. The risk assessment module dynamically generates alarm areas and control recommendations by combining flight schedules and flight paths. The pilot verification phase compares the differences between measured and predicted data, triggering online updates of model parameters and forming a closed-loop optimization system.

[0025] This solution overcomes the limitations of single-model prediction by fusing multi-source data. Measured data provides a dynamic calibration benchmark for simulation, and machine learning algorithms effectively compensate for errors caused by fluctuations in environmental variables under physical constraints. A continuous learning mechanism replaces the traditional periodic manual calibration process, reducing operational costs while enabling adaptive model evolution. 3D scene projection and spatiotemporal matching technologies solve the challenge of collaborative analysis of multi-source heterogeneous data, providing a high-precision spatial representation foundation for risk assessment.

[0026] This application improves the verifiability of photovoltaic glare prediction results, enabling dynamic model calibration using limited measured data and reducing the reliance on dense sampling inherent in traditional methods. The physically constrained machine learning algorithm enhances prediction accuracy while maintaining model interpretability, reducing spatial localization errors of glare events by approximately 40%. A continuous learning mechanism allows the model to adapt to seasonal changes and equipment aging, keeping prediction accuracy fluctuations within ±5% over a 12-month operating cycle. A dynamic alarm mechanism reduces false alarm rates to one-third of traditional methods, effectively supporting airport safety management decisions.

[0027] Example 2: Please refer to Figure 1 The specific method for step S1 is as follows: S1.1 Collect and summarize airport operation and site-related data, including runway number, length, width and azimuth data, taxiway spatial coordinate data, elevation values ​​in the airport digital elevation model, tower latitude and longitude coordinates and height values, tower azimuth and elevation range values, spatial coordinate sequences of approach and departure paths, elevation values ​​of the terrain around the airport, latitude and longitude coordinates and height values ​​of obstacles around the airport, spatial coordinates of key safety viewpoints, boundary coordinates of the airport operation support area, brightness thresholds for permissible glare levels, and time periods and area ranges for airport operation restrictions. S1.2 The collected data are classified and modeled, and the operation and maintenance requirements and site elements are modeled in a unified manner to generate baseline information for subsequent photovoltaic glare detection.

[0028] In this embodiment, the application further proposes the following specific method for step S1: Collect and summarize airport operation and site-related data, including runway number, length, width, and azimuth data; taxiway spatial coordinate data; elevation values ​​in the airport digital elevation model; tower latitude and longitude coordinates and height values; azimuth and elevation range values ​​of the tower's visible area; spatial coordinate sequences of approach and departure paths; elevation values ​​of the terrain surrounding the airport; latitude and longitude coordinates and height values ​​of obstacles surrounding the airport; spatial coordinates of key safety viewpoints; boundary coordinates of the airport operation support area; brightness thresholds for permissible glare levels; and time and area limits for airport operation restrictions. The collected data are classified and modeled, and the operation support requirements and site elements are modeled in a unified manner to generate baseline information for subsequent photovoltaic glare detection.

[0029] Runway number, length, width, and azimuth data are the basic parameters used to determine the physical properties and spatial orientation of the runway. These parameters can be obtained from airport design drawings or satellite mapping data. Their purpose is to provide a spatial positioning reference for the runway in subsequent calculations of light reflection paths.

[0030] Taxiway spatial coordinate data refers to the set of three-dimensional coordinates of the taxiway centerline or edge line. Specifically, it can be collected using aerial photogrammetry or lidar scanning technology to construct a geometric model of the airport's ground activity area.

[0031] The elevation values ​​in the airport digital elevation model refer to the rasterized elevation data that reflects the undulations of the airport's surface. Specifically, they can be generated through airborne radar or UAV oblique photography and are used to establish a three-dimensional terrain scene to support glare propagation path analysis.

[0032] The azimuth and elevation range values ​​of the tower's visible range refer to the horizontal and vertical viewing angle limits of the tower's observation window. These values ​​can be calculated based on tower architectural design specifications or visual field analysis software and are used to constrain the visually sensitive area range for glare detection.

[0033] The spatial coordinate sequence of approach and departure paths refers to the set of standard flight trajectory points during the takeoff and landing phases of an aircraft. Specifically, it can be obtained from aeronautical data compilation or ADS-B signal analysis and is used to identify critical flight paths that may be affected by glare.

[0034] By collecting site elements such as runway physical parameters, taxiway coordinates, terrain elevation, tower view, and flight paths, a multi-dimensional spatial dataset covering the airport's operational support area is formed. Correlation analysis between runway azimuth and solar trajectory can determine the runway area potentially covered by reflected light from photovoltaic modules. Overlaying taxiway coordinates with the tower view can identify the impact range of glare on ground control. Combining flight path coordinates with terrain elevation can predict potential interference from glare on approach routes. After data classification, a geographic information system (GIS) is used for spatial registration and unified modeling of the multi-dimensional data. For example, runway azimuth and solar azimuth are vector-overlaid, and the three-dimensional spatial relationship between the tower view and photovoltaic module locations is calculated. Finally, detection baseline information including geographical constraints, operational limitations, and safety thresholds is generated.

[0035] Traditional methods typically only collect basic geographic information as a basis for detection, lacking a systematic integration of dynamic elements such as tower line of sight, flight path, and operational phase. This solution, however, constructs a baseline model incorporating both spatial and temporal constraints by integrating runway attributes, taxiway coordinates, flight path sequences, and operational limitations. For example, existing technologies may only use static elevation data, while this solution further combines route coordinates with tower elevation angle ranges, enabling more precise location of glare-threatening areas during critical flight phases.

[0036] This application addresses the difficulty of model calibration caused by the single dimension of traditional detection baseline information. By integrating multi-source data such as runway, taxiway, terrain, line-of-sight, and flight path, it provides a high-precision spatial reference for the coupling of subsequent physical simulation and measured data, thereby reducing the reliance on high-frequency field measurements. For example, the spatiotemporal constraints generated based on flight path coordinate sequences and runtime limitations can be directly used to narrow the computational scope of ray tracing, reduce computational resource consumption in invalid simulation areas, and improve the matching efficiency between model prediction results and actual observation data.

[0037] Example 3: Please refer to Figure 1 The specific method for step S2 is as follows: S2.1 Configure and deploy multi-source data acquisition devices, install fixed meteorological sensors, optical observation equipment, spectral and irradiance measurement devices and solar position calculation devices in the permitted area of ​​the airport, and deploy mobile sampling devices near photovoltaic modules to obtain real-time observation data including air temperature, relative humidity, wind speed, wind direction, visibility, solar irradiance, spectral reflectance, sky brightness and solar azimuth angle, and at the same time obtain specular reflectance, diffuse reflectance and surface temperature values ​​of the photovoltaic module surface; S2.2 Record and summarize multi-source observation data. The meteorological measurement results, optical observation results, spectral and irradiance measurement results, photovoltaic module reflection characteristic measurement results, aircraft take-off and landing flight trajectory coordinate sequence and solar position calculation results are uniformly formatted to form multi-source observation data for glare detection.

[0038] In this embodiment, the specific method of step S2 is further proposed as follows: Configure and deploy multi-source data acquisition devices, install fixed meteorological sensors, optical observation equipment, spectral and irradiance measurement devices and solar position calculation devices in the permitted area of ​​the airport, and deploy mobile sampling devices near the photovoltaic modules to obtain real-time observation data including air temperature, relative humidity, wind speed, wind direction, visibility, solar irradiance, spectral reflectance, sky brightness and solar azimuth angle, and simultaneously obtain specular reflectance, diffuse reflectance and surface temperature values ​​of the photovoltaic module surface; record and summarize the multi-source observation data, and uniformly format the meteorological measurement results, optical observation results, spectral and irradiance measurement results, photovoltaic module reflectance characteristic measurement results, aircraft take-off and landing flight trajectory coordinate sequence and solar position calculation results to form multi-source observation data for glare detection.

[0039] Fixed meteorological sensors refer to fixed equipment installed within permitted areas of airports to continuously monitor meteorological parameters such as temperature, relative humidity, wind speed, and wind direction. Specifically, this can be achieved using meteorological stations equipped with temperature probes, humidity sensors, and anemometers, providing basic meteorological data support for glare detection. Mobile sampling devices refer to portable measuring equipment that can be flexibly deployed near photovoltaic modules to acquire the reflectivity parameters of the photovoltaic module surface. Specifically, this can be achieved using mobile monitoring vehicles equipped with spectral analyzers and infrared thermography modules to dynamically capture reflectivity data from photovoltaic modules at different locations. Unified formatting processing refers to converting observational data from different sources into a unified data structure and storage format. Specifically, this can be achieved using data standardization protocols and timestamp alignment algorithms to ensure consistency of multi-source data across time and space.

[0040] By deploying fixed meteorological sensors and optical observation equipment within permitted areas of the airport, continuous monitoring of meteorological conditions and optical parameters, such as air temperature, wind speed, and solar irradiance, can be achieved. Simultaneously, mobile sampling devices can be deployed near photovoltaic (PV) modules, dynamically adjusting measurement positions according to the distribution of the PV array. For example, mobile monitoring vehicles can be used to collect specular reflectivity and surface temperature data from different PV modules. All collected meteorological, optical, spectral, and flight trajectory data are uniformly formatted and processed to form a structured multi-source observation dataset. This dataset, by integrating measurement results from fixed and mobile devices, covers key monitoring points within the airport's operational area, providing high spatiotemporal resolution input data for subsequent physical optical simulations and model calibration.

[0041] Traditional methods typically rely on fixed sensors or a single data source, resulting in numerous monitoring blind spots and an inability to dynamically capture changes in the reflection characteristics of photovoltaic modules. This solution, however, combines fixed and mobile equipment, enabling flexible deployment in sensitive airport areas. For example, mobile monitoring points can be temporarily added in areas with dense photovoltaic modules, thereby improving data acquisition coverage and dynamic response capabilities. Simultaneously, standardized formatting addresses the heterogeneity issue of multi-source data. For instance, it synchronizes meteorological data with flight trajectory data from different sampling frequencies, avoiding model input bias caused by inconsistent data formats in traditional methods.

[0042] This application enables the acquisition of high-precision multi-source observation data at a lower cost. For example, mobile equipment can reduce the number of fixed sensors required, while ensuring accurate measurement of photovoltaic reflection characteristics in key areas. Unified processing of multi-source data provides a reliable data foundation for subsequent model calibration. For instance, time synchronization can eliminate the time misalignment between meteorological data and flight trajectories, thereby improving the credibility of glare prediction models and avoiding the verification difficulties caused by the lack of measured data in traditional methods.

[0043] Example 4: Please refer to Figure 1 The specific method for step S3 is as follows: S3.1. Perform time synchronization and spatial coordinate registration on the collected meteorological data, optical observation data, spectral and irradiance measurement data, photovoltaic module reflection characteristic data, flight trajectory coordinate data and solar position data, and perform outlier detection and removal, noise filtering and radiation intensity correction processing. S3.2 Project the processed data into a three-dimensional scene according to a unified format, calculate the data reliability based on the measurement accuracy and repeatability of the observation data, and generate a standardized observation dataset that can be used for subsequent photovoltaic glare analysis.

[0044] In this embodiment, the specific method of step S3 is further proposed as follows: Time synchronization and spatial coordinate registration are performed on the collected meteorological data, optical observation data, spectral and irradiance measurement data, photovoltaic module reflection characteristic data, flight trajectory coordinate data and solar position data; outlier detection and removal, noise filtering and radiation intensity correction are performed; the processed data are projected into a three-dimensional scene according to a unified format; the data reliability is calculated based on the measurement accuracy and repeatability of the observation data; and a standardized observation dataset that can be used for subsequent photovoltaic glare analysis is generated.

[0045] Time synchronization refers to aligning observation data from different sources in the time dimension. This can be achieved using GPS timing modules or network time protocols to eliminate time deviations caused by differences in the clocks of the acquisition devices. Spatial coordinate registration refers to unifying data from different coordinate systems to the same geographic reference frame. This can be achieved using coordinate transformation algorithms or geographic information system tools to ensure the consistency of the data's spatial location. Outlier detection and removal refers to identifying and removing observation data that exceeds a reasonable range. This can be achieved using statistical outlier detection methods or threshold judgments based on physical constraints to eliminate erroneous data caused by sensor malfunctions or environmental interference. Noise filtering refers to reducing random interference components in the data. This can be achieved using moving average filtering or wavelet denoising algorithms to improve the signal-to-noise ratio. Radiation intensity correction refers to correcting measurement deviations based on sensor characteristics. This can be achieved using laboratory calibration coefficients or comparison with field reference equipment to ensure the accuracy of radiation values. 3D scene projection refers to mapping 2D observation data to a 3D spatial model. This can be achieved using perspective projection algorithms or geospatial interpolation methods to construct a data spatial structure that matches the actual terrain of the airport. Data credibility calculation refers to the reliability index for evaluating data quality. Specifically, it can be achieved by measurement error propagation analysis or repeated observation consistency test, and is used to quantify the usability weight of data in subsequent analysis.

[0046] In the data preprocessing stage, the clock deviations of different acquisition devices, such as meteorological sensors, optical equipment, and spectrometers, are first controlled within milliseconds using a time synchronization module to ensure that all observation data have a unified time reference. Then, a coordinate transformation algorithm is used to register the WGS84 coordinate system of the flight trajectory, the local coordinate system of the photovoltaic modules, and the elevation coordinate system of the airport DEM model, forming a spatially consistent composite dataset. An outlier detection module sets dynamic thresholds based on historical data statistical distribution, automatically identifying and removing temperature, irradiance, or reflectance values ​​exceeding reasonable ranges. The noise filtering stage uses a sliding window to smooth high-frequency sampled data, eliminating the impact of transient interference on radiation measurements. The radiation correction module calls the equipment's factory calibration parameters to compensate for measurement drift caused by changes in ambient temperature. After basic processing, the data is projected onto a 3D airport scene model containing runways, obstacles, and photovoltaic arrays, forming a standardized dataset with spatial topological relationships. Finally, based on the measurement error range of each sensor and the consistency results of repeated observations, a reliability score is calculated for each data point, providing a quality weighting basis for subsequent fusion analysis.

[0047] Traditional methods often overlook the spatiotemporal alignment issues of multi-source data, leading to misalignments in timestamps or coordinate systems across different sensor data, affecting the accuracy of subsequent analysis. Existing technologies rely heavily on manual screening for anomaly data processing, lacking automated detection mechanisms and struggling to handle the high-frequency, multi-type data streams in airport environments. This solution, through a systematic preprocessing workflow, achieves consistent integration of multi-source data across time and space dimensions. Combined with an automated quality control mechanism, it significantly improves the integrity and reliability of the dataset, laying the foundation for effective coupling of physical simulation and measured data.

[0048] The component surface normal vector is calculated as follows:

[0049] Where Z is the tilt angle of the component, F is the azimuth angle of the component, and FXL is the surface normal vector of the component; The method for calculating the incident angle at the glare observation point is as follows:

[0050] in, For the component tilt angle, For component orientation angle, The normal vector of the component surface; Glare brightness is calculated as follows:

[0051] Where E is the incident irradiance, R is the reflectivity of the photovoltaic array, T is the surface scattering characteristic parameter, L is the glare angle at the observation point, and RSJ is the incident angle at the glare observation point. This application addresses the difficulty of model calibration caused by data quality defects in traditional methods. Through spatiotemporal registration and outlier handling, inconsistencies in multi-source data are eliminated, reducing the risk of error accumulation in subsequent fusion analysis. Radiometric correction and noise filtering improve the accuracy of optical measurement data, enabling measured data to effectively constrain the parameter range of the physical model. Three-dimensional scene projection constructs a data spatial structure that matches the airport operating environment, providing a realistic geographical reference for glare path simulation calculations. Data credibility assessment quantifies the reliability of observation data, providing differentiated data weight inputs for machine learning models, thereby effectively improving model prediction accuracy while reducing the amount of on-site measured data.

[0052] Example 5: Please refer to Figure 1 The specific method for step S4 is as follows: S4.1 Based on the geometric dimensions, tilt angle, azimuth angle, arrangement spacing and surface reflectivity measurements of photovoltaic modules, combined with the real-time altitude angle, azimuth angle of the sun and the latitude and longitude coordinates of the observation area, establish a ray tracing calculation model or analytical calculation model based on physical optics principles to simulate and calculate the light reflection path of photovoltaic modules under different time and space conditions. S4.2 Based on the solar irradiance, sky brightness, air temperature, air humidity and atmospheric transmittance data obtained from meteorological observations, the reflection intensity output by the ray tracing calculation model is numerically corrected, and the coordinates, duration and corresponding reflection brightness values ​​of the areas where glare may occur in different time periods and spatial ranges are calculated and output.

[0053] In this embodiment, the application further proposes the following specific method for step S4: Based on the geometric dimensions, tilt angle, azimuth angle, arrangement spacing, and surface reflectivity measurements of the photovoltaic module, combined with the real-time altitude angle, azimuth angle, and latitude and longitude coordinates of the observation area, a ray tracing calculation model or analytical calculation model based on physical optics principles is established to simulate and calculate the light reflection path of the photovoltaic module under different time and space conditions; based on the solar irradiance, sky brightness, air temperature, air humidity, and atmospheric transmittance data obtained from meteorological observations, the reflection intensity output by the ray tracing calculation model is numerically corrected, and the coordinates, duration, and corresponding reflection brightness values ​​of the areas where glare may occur in different time periods and spatial ranges are calculated and output.

[0054] Ray tracing computational models are mathematical models that simulate the reflection path of light on the surface of photovoltaic modules based on the principles of geometric optics. Specifically, they can be implemented using Monte Carlo ray tracing algorithms. By calculating the interaction between light rays and the module surface, they generate reflection path distribution data to determine the potential glare propagation direction. Analytical computational models are mathematical expressions based on Fresnel's equations and the law of reflection. Specifically, they can be implemented using vector analysis methods. They analytically calculate the angle and intensity distribution of reflected light rays to quickly predict the probability of glare occurrence under specific conditions. Numerical correction refers to the operation of compensating for environmental factors in the theoretical calculation results. Specifically, atmospheric attenuation coefficients and humidity correction factors can be used to dynamically adjust the reflection intensity to eliminate errors caused by meteorological conditions on glare brightness.

[0055] The geometric parameters of the photovoltaic modules and the solar position parameters are input into a ray tracing model. The model generates a set of light propagation paths by calculating the incident and reflection angles of each reflective surface. For example, when the solar azimuth angle is 120 degrees and the altitude angle is 45 degrees, the model simulates the spatial distribution of reflected light at various points on the module surface, filtering out reflection paths that intersect with the tower's line of sight or flight path as potential glare sources. Subsequently, meteorological observation data is used to correct for reflection intensity. For instance, under low visibility conditions, atmospheric transmittance parameters reduce the effective brightness value of reflected light, thereby avoiding overestimating the risk of glare. The final output includes the environmentally corrected glare occurrence time interval, geographic coordinate range, and corresponding brightness prediction values.

[0056] Traditional methods rely solely on static theoretical models to calculate reflection paths, neglecting the dynamic impact of real-time meteorological conditions on glare intensity, leading to systematic discrepancies between predicted and actual observations. This proposed solution introduces meteorological observation data to dynamically correct the theoretical model, ensuring that predicted glare brightness reflects the atmospheric attenuation effect in the real environment, significantly improving the consistency between predicted results and actual observation data.

[0057] The method for fusing feature vectors is as follows:

[0058] Where RHF represents the fusion feature of the nth sample, CS represents the measured data, CF represents the simulated data, and a represents the fusion weight. ; The calibration prediction output method is as follows:

[0059] Where XZY is the calibrated prediction result, RHF is the fusion feature of the nth sample, CS is the measured data, CF is the simulation data, m is the machine learning model, and u is the physical constraint correction coefficient. The probability of glare events is calculated as follows:

[0060] Where GL is the probability of a glare event occurring, and XZY is the calibrated prediction result; This application effectively solves the problem of insufficient accuracy caused by neglecting environmental variables in traditional glare prediction models. By integrating physical models with real-time meteorological data, the glare prediction results can dynamically adapt to the light propagation characteristics under different weather conditions, reducing the false alarm rate and missed alarm rate caused by changes in environmental factors, and providing a more reliable decision-making basis for airport safety management.

[0061] Example 6: Please refer to Figure 1 The specific method for step S5 is as follows: S5.1. Match the meteorological observation data, optical observation data, spectral measurement data, irradiance measurement data, photovoltaic module reflection measurement data, flight trajectory coordinate data and solar position data that have been processed by time synchronization, spatial registration and quality control with the glare prediction data obtained by physical optics simulation calculation to establish a fusion modeling dataset that includes measured data and simulation data. S5.2 Based on the fusion modeling dataset, a machine learning prediction model with physical constraints is constructed. The glare prediction values ​​output by the physical optics simulation calculation are numerically corrected and the uncertainty is quantified to generate the temporal distribution, spatial distribution and calibrated glare brightness prediction results of glare events for subsequent glare risk assessment.

[0062] In this embodiment: This application further proposes to match meteorological observation data, optical observation data, spectral measurement data, irradiance measurement data, photovoltaic module reflection measurement data, flight trajectory coordinate data, and solar position data, which have undergone time synchronization, spatial registration, and quality control processing, with glare prediction data obtained from physical optics simulation calculations to establish a fusion modeling dataset containing both measured and simulated values; based on the fusion modeling dataset, a machine learning prediction model with physical constraints is constructed to perform numerical correction and uncertainty quantification on the glare prediction values ​​output by the physical optics simulation calculations, generating glare event temporal distribution, spatial distribution, and calibrated glare brightness prediction results for subsequent glare risk assessment.

[0063] A fusion modeling dataset refers to a structured collection of data after spatiotemporally aligning measured and simulated data. Specifically, data matching algorithms can be used to associate the timestamps and spatial coordinates of the measured data with the simulation results to achieve a mapping of correspondences between the data. A machine learning prediction model with physical constraints introduces physical rules such as the law of optical reflection and the principle of energy conservation as constraints during model training. This can be achieved using physical information neural networks or regularization methods to ensure that the machine learning prediction results conform to physical laws. Numerical correction refers to correcting deviations in simulation prediction results. This can be achieved using regression algorithms or residual learning techniques, establishing correction relationships through difference analysis between measured and simulated data. Uncertainty quantification refers to evaluating the reliability of the prediction results. This can be achieved using Bayesian inference or Monte Carlo methods, outputting the confidence interval or probability distribution of the predicted values.

[0064] After preprocessing, the measured data is matched with the simulation prediction results in the spatiotemporal dimensions to form a fusion dataset containing features from multiple data sources. Based on this dataset, the machine learning model limits the parameter search space through physical constraints during training, ensuring that the prediction results conform to the basic laws of optical reflection while maintaining the advantages of data-driven approaches. The model corrects the brightness values ​​predicted by the simulation through an iterative optimization process and calculates the prediction uncertainty under different scenarios. The final calibration results include information on the occurrence time, spatial location, and probability distribution of brightness values ​​of glare events, providing reliable input for subsequent risk assessment.

[0065] Traditional methods rely on a single simulation model for prediction, lacking dynamic calibration with measured data, resulting in the inability to effectively correct prediction biases under complex meteorological conditions. This solution constructs a machine learning model by fusing measured and simulated data, achieving automatic correction of prediction results under the constraints of physical laws, overcoming the limitations of traditional methods that rely on manual parameter tuning and local validation. Simultaneously, uncertainty quantification ensures an objective assessment of the prediction's reliability, avoiding the risk of misjudgment due to over-reliance on model output.

[0066] This application effectively solves the problem of insufficient accuracy in traditional glare prediction models due to the lack of calibration with measured data, and reduces the reliance on high-frequency field measurements. By combining physical constraints with data-driven approaches, the model's generalization ability in complex environments is improved, and the reliability of the prediction results is significantly enhanced. Simultaneously, the quantification of uncertainty provides a scientific basis for risk management decisions, avoiding safety hazards caused by prediction errors.

[0067] Example 7: Please refer to Figure 1 The specific method for step S6 is as follows: S6.1 Based on the temporal distribution, spatial distribution, and glare brightness prediction values ​​of glare events output by the fusion model, combined with flight take-off and landing times, flight route coordinate sequences, runtime information, and real-time weather conditions, the impact of glare events on flight operations is quantitatively calculated. Glare brightness thresholds and glare duration thresholds are set to determine the level of glare events and generate corresponding risk assessment result sets. S6.2 Based on the risk assessment results set and flight operation characteristics, generate an alarm dataset that includes glare event level, boundary coordinates of the affected area, time interval of the affected area, and recommended control measures. Establish a risk alarm and control mechanism, and output risk indicator values, temporal distribution map and spatial distribution map of glare events through a human-computer interaction interface.

[0068] In this embodiment: This application further proposes to conduct pilot verification operations in a typical operating area selected by the airport, collect flight trajectory data, glare observation and measurement data, meteorological monitoring data, and control measure implementation effect data, compare the collected operational data with the prediction results output by the fusion model one by one, and calibrate the prediction results based on the comparison differences to obtain a calibrated prediction result set; based on the operational data and the calibrated prediction result set, establish a data governance mechanism including data cleaning rules, quality control criteria, and version management strategies, and construct a model update process with continuous learning function, so that the fusion model can perform online adaptive updates during operation and maintain the stability and effectiveness of long-term application.

[0069] Pilot validation refers to the model validation process conducted in typical airport operating areas. This can be achieved by deploying sensor networks and data acquisition terminals to obtain multi-dimensional data from real-world operational scenarios. This operation reduces the risk of global deployment through localized validation, ensuring the model's predictions match the real-world situation.

[0070] Data governance mechanisms refer to the systematic rules governing collected data and prediction results. Specifically, they can be implemented using automated data cleaning algorithms and version control protocols to eliminate data noise and maintain data consistency. This mechanism improves data quality through standardized processing procedures, providing reliable input for continuous model optimization.

[0071] The continuous learning model update process refers to a mechanism that dynamically adjusts model parameters based on new data. This can be implemented using incremental learning algorithms or online learning frameworks to adapt to changes in the airport operating environment. This process maintains the model's predictive ability through periodic parameter updates, avoiding performance degradation caused by environmental changes.

[0072] After deploying data acquisition equipment in selected typical areas, real-time measured data such as flight trajectories, glare intensity, and meteorological parameters are obtained. The measured data is compared with the prediction results output by the fusion model over time to identify differences and generate calibration parameters. For example, when the measured glare brightness is higher than the predicted value, the model weights can be adjusted using the backpropagation algorithm to reduce prediction bias. The calibrated prediction result set, along with historical data, is input into the data governance module to perform missing value imputation, outlier filtering, and data format standardization. Furthermore, the model update process is tracked based on a version management strategy to ensure that each iteration can be backtested. Through an online learning framework, the model can automatically adjust ray tracing calculation parameters or machine learning correction coefficients based on newly acquired data, achieving adaptive updates without manual intervention.

[0073] Traditional methods rely on fixed model parameters and lack dynamic calibration mechanisms, leading to a decline in prediction accuracy after long-term operation. This proposed solution, through pilot validation and continuous learning mechanisms, can continuously optimize model parameters in real-world operating environments, addressing prediction biases caused by changes in weather conditions or aging photovoltaic modules. Simultaneously, the data governance mechanism avoids the model failure risk caused by data quality fluctuations inherent in traditional methods, significantly improving system robustness.

[0074] This application achieves dynamic calibration between model predictions and actual operational data, effectively reducing prediction errors caused by environmental changes or equipment performance degradation. By establishing data governance and continuous learning mechanisms, it ensures that the model maintains high-precision prediction capabilities during long-term operation, while reducing manual maintenance costs. Ultimately, a closed-loop optimized detection system is formed, providing reliable technical support for airport photovoltaic glare risk management.

[0075] Example 8: Please refer to Figure 1 The specific method for step S7 is as follows: S7.1. Conduct pilot verification operations in the selected typical operating area of ​​the airport, collect flight trajectory data, glare observation and measurement data, meteorological monitoring data, and control measure implementation effect data, compare the collected operating data with the prediction results output by the fusion model one by one, and calibrate the prediction results based on the comparison differences to obtain a calibrated prediction result set. S7.2. Based on the running data and the calibrated prediction result set, establish a data governance mechanism that includes data cleaning rules, quality control criteria and version management strategies, and build a model update process with continuous learning capabilities, so that the fusion model can perform online adaptive updates during operation and maintain the stability and effectiveness of long-term application.

[0076] In this embodiment: This application further proposes to conduct pilot verification operations in a typical operating area selected by the airport, collect flight trajectory data, glare observation and measurement data, meteorological monitoring data, and control measure implementation effect data, compare the collected operational data with the prediction results output by the fusion model one by one, and calibrate the prediction results based on the comparison differences to obtain a calibrated prediction result set; based on the operational data and the calibrated prediction result set, establish a data governance mechanism including data cleaning rules, quality control criteria, and version management strategies, and construct a model update process with continuous learning function, so that the fusion model can perform online adaptive updates during operation and maintain the stability and effectiveness of long-term application.

[0077] Pilot validation operations refer to selecting representative areas within the airport operating environment for model verification. This can be achieved by deploying sensor arrays at runway ends, taxiway intersections, or areas visible from the control tower. The reliability of the system is verified by comparing local measured data with model predictions. Data governance mechanisms refer to a set of rules for standardizing operational data. This can be achieved by establishing data cleaning processes to eliminate outliers, setting quality control standards to screen valid data, and using version control to manage model iterations, ensuring data input quality and model update stability. Continuous learning functionality refers to the model's ability to automatically optimize based on new data. This can be achieved by using incremental learning algorithms combined with historical data weight decay strategies, enabling the model to adapt to changes in the airport operating environment.

[0078] Optical sensors and meteorological monitoring stations were deployed at the runway end area to continuously collect glare brightness data and meteorological parameters during flight taxiing. The timestamps of the measured data were matched with the model predictions, and the reflection intensity calculation parameters of the ray tracing model were dynamically adjusted using an error backpropagation algorithm. A data quality assessment matrix was established to automatically label and remove data from scenarios such as sensor power outages and extreme weather interference. A rolling time window strategy was used for periodic retraining of the model, retaining the most recent three months of valid data as the training set for each update, while freezing the validated basic optical calculation module parameters to ensure that the model maintains physical constraints during iteration.

[0079] Traditional methods rely on densely deployed sensors across the entire region year-round for model validation, resulting in excessively high equipment procurement and maintenance costs. This solution, by conducting validation in typical areas, compresses the data collection scope to key nodes, reducing hardware investment costs while ensuring the effectiveness of model calibration. Existing technologies lack data governance mechanisms, causing anomalous data to directly impact the stability of model iterations. This solution establishes data cleaning rules and version management strategies to form a closed-loop quality control system, preventing erroneous data from contaminating model parameters.

[0080] This application effectively solves the high cost problem caused by the deployment of sensors across the entire region in traditional methods, and verifies the economy and operability of model calibration through pilot testing in typical areas. The established continuous learning mechanism enables the model to adapt to long-term changes such as airport expansion and photovoltaic module aging, avoiding the prediction failure of traditional static models due to environmental changes. The data governance mechanism ensures data quality and version controllability during model updates, overcoming the model performance fluctuation problem caused by data anomalies in existing technologies.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0082] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A method for detecting photovoltaic glare at an airport based on multi-source data coupling, characterized in that: The specific steps are as follows: S1, establish detection requirements and site element baseline, determine airport operation guarantee target area, key safety viewpoint and allowed glare level, combine airport terrain data, runway direction, elevation model, approach profile, tower visibility constraint and operation restriction to form detection baseline information; S2, construct a multi-source data acquisition system, the system includes meteorological data, optical observation data, spectral and radiation sensor data, photovoltaic module reflection characteristic parameters, flight trajectory data and solar position parameters, fixed or mobile sampling devices are arranged in the allowed area of the airport to obtain glare observation information; S3, pretreatment and quality control of collected data, including time synchronization, coordinate registration, outlier rejection, noise filtering, radiation correction and three-dimensional scene projection, and generating standardized observation products according to the reliability of observation data; S4, based on the geometric parameters, reflection characteristics and solar position parameters of photovoltaic modules, a physical optical simulation model is constructed, the potential glare path is tracked or analytically calculated, the spatiotemporal distribution results of possible glare are obtained, and the reflection intensity is corrected combined with meteorological conditions; S5, coupling the pretreated measured data and the physical simulation prediction results, constructing a fusion model, using physical constraint machine learning method to correct and quantify the uncertainty of the prediction results, outputting the probability distribution of glare events and the calibrated brightness prediction value; S6, based on the output of the fusion model, establish risk assessment and alarm mechanism, grade determination of glare events, combined with flight operation characteristics to generate dynamic alarm and control suggestion, output risk index and event spatiotemporal distribution map through visual interface; S7, carry out pilot verification in typical areas of the airport, collect operation data to compare and calibrate the prediction results of the model, establish continuous learning and data governance mechanism, so that the method realizes online adaptive update and long-term application in the operation process.

2. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 1, characterized in that, The specific method of step S1 is as follows: S1.1, collect and summarize the airport operation and site related data, the data includes runway number, length, width and direction angle data, spatial coordinate data of taxiway, elevation value in airport digital elevation model, latitude and longitude coordinates and height value of tower, range value of tower visual range, spatial coordinate sequence of approach and departure path, elevation value of airport surrounding terrain, latitude and longitude coordinates and height value of airport surrounding obstacles, spatial coordinates of key safety viewpoints, boundary coordinate points of airport operation guarantee area, brightness threshold of allowed glare level, and time period and area range of airport operation restriction; S1.2, classify and model the collected data, unify the operation guarantee requirements and site elements, and generate baseline information for subsequent photovoltaic glare detection.

3. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 2, characterized in that, The specific method of step S2 is as follows: S2.1, configure and deploy multi-source data acquisition devices, install fixed weather sensors, optical observation equipment, spectral and irradiance measurement devices, and solar position calculation devices in the airport allowed area, and deploy mobile sampling devices near the photovoltaic modules to obtain real-time observation data including air temperature, relative humidity, wind speed, wind direction, visibility, solar irradiance, spectral reflectance, sky brightness, and solar azimuth angle, and to obtain the specular reflectance, diffuse reflectance, and surface temperature values of the photovoltaic module surface; S2.2, record and aggregate multi-source observation data, format the weather measurement results, optical observation results, spectral and irradiance measurement results, photovoltaic module reflection characteristic measurement results, aircraft take-off and landing flight trajectory coordinate sequences, and solar position calculation results in a unified format to form multi-source observation data for glare detection.

4. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 3, characterized in that, Step S3 is specifically as follows: S3.1, time synchronization and spatial coordinate registration are performed on the collected weather data, optical observation data, spectral and irradiance measurement data, photovoltaic module reflection characteristic data, flight trajectory coordinate data, and solar position data, and abnormal value detection and elimination, noise filtering, and radiation intensity correction processing are performed; S3.2, the processed data is projected onto a three-dimensional scene according to a unified format, the data reliability is calculated according to the measurement accuracy and repeatability of the observation data, and a standardized observation data set that can be used for subsequent photovoltaic glare analysis is generated.

5. The method of claim 4, wherein, Step S4 is specifically as follows: S4.1, based on the geometric size, inclination angle, azimuth angle, arrangement spacing, and surface reflectance measurement value of the photovoltaic module, combined with the real-time elevation angle, azimuth angle of the sun, and the latitude and longitude coordinates of the observation area, a ray tracing calculation model or an analytical calculation model based on physical optics principles is established, and the light reflection path of the photovoltaic module under different time and space conditions is simulated and calculated; S4.2, according to the solar irradiance, sky brightness, air temperature, air humidity, and atmospheric transmittance data obtained by weather observation, the reflection intensity output by the ray tracing calculation model is numerically corrected, and the region coordinates, duration, and corresponding reflection brightness values of the possible glare in different time periods and spatial ranges are calculated and output.

6. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 5, characterized in that, Step S5 is specifically as follows: S5.1, the weather observation values, optical observation values, spectral measurement values, irradiance measurement values, photovoltaic module reflection measurement values, flight trajectory coordinate values, and solar position values that have been time-synchronized, spatially registered, and quality-controlled are matched with the glare prediction values obtained by physical optics simulation calculation, and a fusion modeling data set containing measured values and simulation values is established; S5.2, based on the fusion modeling data set, a machine learning prediction model with physical constraint conditions is constructed, the glare prediction values output by the physical optics simulation calculation are numerically corrected and uncertainty quantified, and the glare event time distribution, spatial distribution, and calibrated glare brightness prediction results for subsequent glare risk assessment are generated.

7. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 6, characterized in that, Step S6 is specifically as follows: S6.1, based on the output of the fusion model, the time distribution value, the spatial distribution value and the glare brightness prediction value of the glare event, combined with the flight take-off and landing time, the flight route coordinate sequence, the operation period information and the real-time weather condition, the influence degree of the glare event on the flight operation is calculated quantitatively, the glare brightness threshold and the glare duration threshold are set, the glare event is judged by grade, and the corresponding risk assessment result set is generated; S6.2, according to the risk assessment result set and the flight operation characteristics, an alarm data set containing the glare event grade, the influence area boundary coordinate, the influence time interval and the recommended control measures is generated, a risk warning and control mechanism is established, and the risk index value, the time distribution graph and the spatial distribution graph of the glare event are output through the man-machine interaction interface.

8. The airport photovoltaic glare detection method based on multi-source data coupling according to claim 7, characterized in that, Step S7 is specifically as follows: S7.1, in the selected typical operation area of the airport, the flight operation trajectory value, the glare observation measurement value, the meteorological monitoring value and the control measure execution effect value are collected, the collected operation data and the prediction result output by the fusion model are compared one by one, and the prediction result is calibrated based on the comparison difference to obtain a calibrated prediction result set; S7.2, based on the operation data and the calibrated prediction result set, a data governance mechanism containing data cleaning rules, quality control criteria and version management strategies is established, a model updating process with continuous learning function is constructed, the fusion model can perform online adaptive updating in the operation process, and the stability and effectiveness of long-term application are maintained.

Images