A low-altitude air route-based micro weather station intelligent site selection method

Abstract

The application discloses a kind of micro meteorological station intelligent site selection method and system based on low-altitude air route, it is related to meteorological station site selection field, main scheme is: the numbered area that needs to establish micro meteorological base station is numbered division, and the data of each numbered area in numbered area is obtained;The data of each numbered area is preprocessed, and evaluation index set is obtained by data analysis;And integrated into comprehensive vector by the way of channel stacking, then the comprehensive vector is analyzed, and the site selection probability of each numbered area is obtained;The comprehensive vector, site selection probability and evaluation index set of each numbered area are comprehensively analyzed by GBDT model, and the site selection score of each numbered area can be obtained;And the site selection place of micro meteorological station is selected;The safety guarantee function of micro meteorological station to low-altitude air route is fully considered, the problem that micro meteorological station site selection is disconnected with low-altitude air route demand in prior art is solved, and strong meteorological support is provided for the development of low-altitude economy.

Description

Technical Field

[0001] This invention relates to the field of meteorological station site selection technology, specifically to an intelligent site selection method for micro-meteorological stations based on low-altitude air routes. Background Technology

[0002] Existing methods for selecting the location of micro-weather stations are typically based on experience or simple data analysis. For example, in the selection of locations for power micro-weather stations, the location distance is often determined by analyzing the relationship between the partial correlation coefficient of micro-weather data and the horizontal distance to the station.

[0003] However, with the rapid development of the low-altitude economy, the application of low-altitude air routes is becoming increasingly widespread. As a crucial means of low-altitude meteorological monitoring, the location of micro-weather stations has a vital impact on the safety and efficiency of low-altitude air routes. Micro-weather stations can monitor meteorological conditions along low-altitude air routes in real time, such as wind speed, wind direction, and visibility, providing accurate meteorological information for aircraft and ensuring flight safety. However, in some complex areas, the monitoring data from micro-weather stations is insufficient to meet the needs of refined meteorological services for low-altitude air routes. Therefore, the development of the low-altitude economy is creating an increasing demand for low-altitude meteorological services, and existing micro-weather station location methods lack consideration for adapting to the development of the low-altitude economy. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent site selection method for micro-weather stations based on low-altitude air routes, which at least solves the problem in existing technologies where the site selection process does not refer to specific low-altitude air route data, resulting in the weather station being unable to provide sufficient support for low-altitude air routes.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent site selection method for micro-weather stations based on low-altitude air routes, comprising:

[0008] Step 1: Divide the numbered areas where micro-meteorological base stations need to be established into numbered areas, and obtain meteorological condition data, topographic data, low-altitude flight path data, and surrounding environmental data for each numbered area;

[0009] Step 2: Preprocess the meteorological conditions, topography, low-altitude flight path, and surrounding environment data for each numbered area, and obtain a set of evaluation indices through data analysis;

[0010] Step 3: Integrate the preprocessed data into a comprehensive vector by channel stacking, and then analyze the comprehensive vector to obtain the location probability of each numbered area;

[0011] Step 4: By comprehensively analyzing the integrated vector, location probability, and evaluation index set of each numbered region using the GBDT model, the location score of each numbered region can be obtained;

[0012] Step 5: Based on the data analysis of Step 4, select the site locations for the micro-weather stations from several numbered areas.

[0013] In the preferred scheme of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the evaluation index set includes the degree of impact on low-altitude air route operation, regional topography index, low-altitude air route attractiveness value, and environmental impact index.

[0014] Meteorological data, topographic data, low-altitude airway data, and surrounding environmental data are normalized. A multiple linear regression model is used to analyze the meteorological data to determine the degree of impact on low-altitude airway operations in each numbered area. A topographic index model is used to analyze the topographic data to obtain the regional topographic index for each numbered area. A gravity model is used to analyze the low-altitude airway data to obtain the low-altitude airway attraction value. An environmental impact index model is used to analyze the surrounding environmental data to obtain the environmental impact index.

[0015] In the preferred embodiment of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the formula for calculating the degree of impact on low-altitude air route operation is as follows:

[0016] ;

[0017] Where Y represents the degree of impact on low-altitude air routes; X1, X2, X3, X4 and X5 are wind speed, temperature, humidity, air pressure and precipitation values, respectively. 0 represents the intercept term. 1. 2. 3. 4 and 5 represents the regression coefficients for wind speed, temperature, humidity, air pressure, and precipitation, respectively, and ξ represents the error term.

[0018] In the preferred embodiment of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the formula for calculating the regional topographic index is as follows:

[0019] ;

[0020] Where TI represents the regional topographic index, H is the altitude, S is the slope, A is the aspect, b is the weighting coefficient of H, c is the weighting coefficient of S, and f is the weighting coefficient of A.

[0021] In the preferred embodiment of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the attractive force value of the low-altitude air route is calculated using a gravity model, based on the following formula:

[0022] ;

[0023] Where T represents the low-altitude airway attractiveness value; O represents the number of low-altitude airways passing through this numbered area; Q represents the number of flight missions passing through this numbered area; and D represents the number of low-altitude airways passing through this numbered area. near This indicates the distance between this numbered area and the nearest low-altitude airway; k is a constant, and τ is the impedance coefficient.

[0024] In the preferred embodiment of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the formula for calculating the environmental impact index is:

[0025] ;

[0026] Where EI represents the Environmental Impact Index, B represents the average building height, V represents the vegetation coverage, d is the weighting coefficient of B, and e is the weighting coefficient of V.

[0027] In the preferred scheme of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the preprocessed meteorological condition data, topographic data, low-altitude air route data and surrounding environment data are organized into one-dimensional arrays by the convolutional layers of the convolutional neural network (CNN). Then, the one-dimensional arrays are stacked according to the channel dimension to generate a comprehensive vector.

[0028] In the preferred scheme of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the method for obtaining the site selection probability of each numbered area by analyzing the comprehensive vector is as follows:

[0029] The composite vector is analyzed using a Convolutional Neural Network (CNN). In a CNN, the output layer is designed with three neurons, enabling it to output three probability values. The composite vector is then input into a fully connected layer, which maps it back to the output layer to obtain the output score vector kz. kx ;

[0030] The output score vector kz is obtained through the Softmax function. kx Converting to a probability distribution, the formula is:

[0031] ;

[0032] ;

[0033] ;

[0034] Where π is a constant; kzkx It is the output score of the kx-th neuron; It is the probability of the output of the first neuron. It is the probability of the output of the second neuron. It is the probability of the output of the third neuron.

[0035] In the preferred scheme of the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes, the method for obtaining the site selection score by analyzing the comprehensive vector, site selection probability, and evaluation index set of each numbered region is as follows:

[0036] The comprehensive vector, location probability, and evaluation index set of each numbered region are concatenated in order into a vector ST[XT, [,Y,TI,T,EI], then ST[XT, Input [Y, TI, T, EI] into the GBDT model to obtain the location score. The calculation formula is as follows:

[0037] ;

[0038] Wherein, DF(ST) is the location score for the numbered region. h is the scaling factor for the j-th tree model; j (ST) is the output value of the j-th tree model.

[0039] (III) Beneficial Effects

[0040] This invention provides a method for intelligent site selection of micro-weather stations based on low-altitude air routes, which has the following beneficial effects:

[0041] (1) By preprocessing meteorological data, topographic data, low-altitude airway data, and surrounding environmental data of the numbered area and integrating them into a comprehensive vector, the problem of scattered and unsystematic site selection data in existing technologies is solved. Traditional site selection methods often only consider a single factor or simply superimpose a few factors, failing to comprehensively and deeply analyze the comprehensive conditions of site selection. This scheme integrates multiple data through channel stacking to construct a multi-dimensional comprehensive vector, making the site selection analysis more comprehensive and scientific. This not only improves the accuracy of site selection but also better adapts to the complex low-altitude environment, fully considering the interaction of various influencing factors, and providing stronger data support for subsequent site selection decisions.

[0042] (2) By introducing the GBDT model to comprehensively analyze the integrated vector, location probability, and evaluation index set of each numbered area, the problems of insufficient accuracy and reliability in existing location evaluation technologies are solved. Traditional location evaluation methods may be based on simple rules or empirical formulas, which are difficult to handle complex nonlinear relationships and large data volumes. However, the GBDT model, as a powerful machine learning algorithm, can automatically learn complex patterns and rules in the data and conduct comprehensive and in-depth mining and analysis of various location indicators. The location score obtained through the GBDT model can more accurately reflect the actual location value of each numbered area, effectively avoid the interference of human factors and subjective judgments, and improve the scientificity and reliability of location decisions.

[0043] (3) Based on the site selection calculation results, the location of the micro-weather station is selected from several numbered areas, solving the problem that the site selection decision lacks quantitative basis and is difficult to make objective comparisons in the existing technology. Traditional methods often rely on simple empirical index ranking when selecting site locations, lacking a systematic and quantitative evaluation system. However, this scheme provides a clear quantitative standard for site selection decision by constructing an evaluation index set and calculating the site selection score. This makes the comparison between different numbered areas more intuitive and objective, and can quickly and accurately screen out the most suitable site location, improving the efficiency and quality of the site selection work, ensuring that the site selection of the micro-weather station can meet the meteorological monitoring needs of low-altitude routes to the greatest extent, and ensuring the safety and efficiency of low-altitude flights.

[0044] (3) The entire intelligent site selection method for micro-weather stations based on low-altitude air routes fully considers the safety guarantee role of micro-weather stations for low-altitude air routes and solves the problem of the disconnect between micro-weather station site selection and low-altitude air route requirements in existing technologies. Traditional site selection methods often do not closely integrate with the actual operation of low-altitude air routes, resulting in the monitoring data of micro-weather stations being unable to effectively serve low-altitude flights. However, this scheme ensures that the site selection of micro-weather stations can accurately serve the meteorological support needs of low-altitude air routes by using low-altitude air route data as one of the important site selection criteria, thereby improving the safety and operational efficiency of low-altitude air routes. At the same time, this intelligent site selection method can also be dynamically adjusted and optimized according to changes in low-altitude air routes and the operation of micro-weather stations, which has strong adaptability and foresight, and provides strong meteorological support for the development of the low-altitude economy. Attached Figure Description

[0045] Figure 1 This is a schematic diagram illustrating the steps of an intelligent site selection method for micro-weather stations based on low-altitude air routes according to the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1:

[0048] Please see Figure 1 This invention provides a method for intelligent site selection of micro-weather stations based on low-altitude air routes, comprising:

[0049] Step 1: Divide the numbered areas where micro-meteorological base stations need to be established into numbered areas, and obtain meteorological condition data, topographic data, low-altitude flight path data, and surrounding environment data for each numbered area.

[0050] Specifically, the area where micro-meteorological base stations need to be established is divided into small square or rectangular grid areas at latitude and longitude intervals. For example, a small area can be divided every 0.01 degrees in the longitude direction and every 0.01 degrees in the latitude direction, so that each area point appears as a regular grid on the map, which facilitates geographic positioning and data management. The number of each small area can correspond to a matching latitude and longitude range, such as area s1 corresponding to "N39.99-40.00, E116.30-116.31".

[0051] It should be noted that meteorological data can be obtained from national meteorological departments, other meteorological agencies, or meteorological satellite data and weather radar, and may include at least wind speed, temperature, humidity, air pressure, and precipitation; topographic data can be obtained from geographic information system databases or field measurements, and may include at least altitude, slope, and aspect; low-altitude airway data can be obtained through cooperation agreements with aviation management departments and aviation data service providers, and may include at least the number of airways and traffic flow; by importing the location data and low-altitude airway data of each numbered area into GIS software, the spatial analysis function of GIS can be used to calculate the expected distance between each weather station to be set up and the low-altitude airway, which can be used as the distance between the weather station and the nearest low-altitude airway; surrounding environmental data can be obtained from urban planning departments or field surveys, and may include at least building height, vegetation coverage, and water area.

[0052] This step involves numbering and dividing the areas where micro-meteorological base stations need to be established, which solves the problem of unclear and unsystematic site division in traditional site selection methods. Traditional methods often rely on manual experience or simple geographical division, lacking scientific standards and methods for area division. By numbering and dividing, the boundaries and scope of each area can be clearly defined, facilitating subsequent refined data collection and analysis. Acquiring meteorological condition data, topographic data, low-altitude flight path data, and surrounding environmental data for each numbered area can solve the problem of incomplete and inaccurate data acquisition in existing technologies. Traditional site selection methods usually only consider a single factor or a simple superposition of a few factors, failing to comprehensively and deeply analyze the comprehensive conditions for site selection. This solution, by acquiring multi-dimensional data, can comprehensively and deeply analyze the site selection conditions of each numbered area, improving the scientific nature and accuracy of site selection.

[0053] Step 2: Preprocess the meteorological conditions, topography, low-altitude flight path, and surrounding environment data for each numbered area, and obtain a set of evaluation indices through data analysis.

[0054] Specifically, preprocessing includes at least data normalization and alignment. By normalizing meteorological data, topographic data, low-altitude flight path data, and surrounding environmental data, the values ​​of each data type can be normalized to the range of 0 to 1. For example, after normalizing wind speed, temperature, humidity, air pressure, and precipitation in meteorological data, the following indicators were selected: wind speed (m / s), temperature (°C), humidity (%), air pressure (hPa), and precipitation (mm). Assuming that in a certain numbered area, the values ​​of these indicators after normalization (e.g., mapping the data to the [0,1] interval) are: wind speed 0.6, temperature 0.4, humidity 0.5, air pressure 0.7, precipitation 0.2, etc., the meteorological data, topographic data, low-altitude flight path data, and surrounding environmental data of each numbered area are preprocessed, including data normalization and alignment. By normalizing the data, values ​​of different data types are mapped to the same range, making the data comparable and processable. This improves the data's compatibility and scalability, providing a foundation for subsequent comprehensive analysis. It avoids the problem in traditional site selection methods where diverse data types and different units of measurement lead to difficulties in data processing and prevent direct comprehensive analysis.

[0055] Furthermore, the assessment index set includes: the degree of impact on low-altitude air route operations, regional topography index, low-altitude air route attractiveness value, and environmental impact index.

[0056] Specifically, the degree of impact on low-altitude air route operations can be obtained by analyzing the relationship between meteorological factors and low-altitude air route operations using a multiple linear regression model. The calculation method used is as follows:

[0057] ;

[0058] Where Y represents the degree of impact on low-altitude air routes; X1, X2, X3, X4, and X5 represent wind speed, temperature, humidity, air pressure, and precipitation, respectively. 0 represents the intercept term. 1. 2. 3. 4 and 5 represents the regression coefficients for wind speed, temperature, humidity, air pressure, and precipitation, respectively, and ξ represents the error term. By collecting historical meteorological data and low-altitude flight route operation data, the regression coefficients are estimated using methods such as the least squares method, thereby obtaining the quantitative relationship between meteorological elements and low-altitude flight route operation. 0 can take the value 0.5. 1 can take the value 0.3. 2 can take the value 0.2. 3 can take the value 0.1. 4 can take the value 0.15. 5 can take the value 0.25.

[0059] Multiple linear regression models can quantify the impact of meteorological factors on low-altitude air route operations, providing a scientific basis for evaluation, improving the scientificity and accuracy of site selection, and avoiding the problem that traditional site selection methods cannot quantify the impact of meteorological factors on low-altitude air route operations, resulting in a lack of scientific basis for site selection decisions.

[0060] Specifically, the regional topographic index can be used to quantify the impact of topography and landforms using a topographic index model. The formula is as follows:

[0061] ;

[0062] In this system, TI represents the regional topographic index, H is the altitude, S is the slope, A is the aspect, b is the weighting coefficient for H (which can be 0.5), c is the weighting coefficient for S (which can be 0.3), and f is the weighting coefficient for A (which can be 0.2). By calculating and analyzing the topographic index, the impact of different terrain features on the site selection of micro-weather stations can be assessed, providing a topographical basis for site selection.

[0063] The topographic index model can comprehensively quantify the impact of topography on the site selection of micro-weather stations, provide scientific basis for topography, improve the adaptability and reliability of site selection, and avoid situations where the traditional site selection method does not fully consider the impact of topography, which may lead to the site selection failing to fully utilize the advantages of topography or avoid the disadvantages of topography.

[0064] By analyzing low-altitude flight path data using a gravity model, the attractive force of low-altitude flight paths on different regions can be obtained. The formula used is as follows:

[0065] ;

[0066] Where T represents the low-altitude airway attractiveness value; O represents the number of low-altitude airways passing through this numbered area; Q represents the number of flight missions passing through this numbered area; and D represents the number of low-altitude airways passing through this numbered area. near This indicates the distance between this numbered area and the nearest low-altitude airway; k is a constant, and τ is the impedance coefficient; the constant k is used to adjust the overall scale of the flow and can be a constant, such as 0.1. The impedance coefficient τ indicates the degree of influence of distance on the attraction value, which can be determined and adjusted through fitting analysis of historical data. If τ=2 is selected, it reflects a greater influence.

[0067] Gravity models can quantify the attractiveness of low-altitude air routes to different regions, providing a scientific basis for the demand for low-altitude air routes, improving the pertinence and effectiveness of site selection, and solving the limitation problem that traditional site selection methods do not adequately consider the attractiveness of low-altitude air routes, which may lead to site selection that cannot effectively serve the actual needs of low-altitude air routes.

[0068] Specifically, the environmental impact index can be used to quantify the impact on the surrounding environment using an environmental impact index model. The formula is as follows:

[0069] ;

[0070] Where EI represents the Environmental Impact Index, B represents the average building height, and V represents the vegetation coverage; d is the weighting coefficient of B, which can be 0.6; and e is the weighting coefficient of V, which can be 0.4.

[0071] The environmental impact index model can comprehensively quantify the impact of the surrounding environment on the site selection of micro-weather stations, provide scientific basis for environmental aspects, improve the adaptability and sustainability of site selection, and solve the problem that traditional site selection methods do not take into account the impact of the surrounding environment in a comprehensive manner, which may lead to the site selection not being able to fully integrate into the surrounding environment or avoid environmental interference.

[0072] Step 3: Integrate the preprocessed data into a comprehensive vector by channel stacking, and then analyze the comprehensive vector to obtain the location probability of each numbered area.

[0073] Specifically, the normalized and aligned meteorological condition data, topographic data, low-altitude flight path data, and surrounding environmental data are each organized into one-dimensional arrays through convolutional layers. For example, each indicator of the meteorological condition data can be organized into a one-dimensional array as [0.6, 0.4, 0.5, 0.7, 0.2]. Similarly, all data sets are converted into four one-dimensional arrays. Then, the four one-dimensional arrays are stacked according to the channel dimension, which can be achieved using the NumPy library in Python to form a comprehensive vector XT.

[0074] In the convolutional neural network (CNN) of this scheme, the output layer needs to be designed with three neurons so that it can output three probability values: the first neuron corresponds to low suitability, the second neuron to medium suitability, and the third neuron to high suitability. This is achieved using the Softmax function. After obtaining the comprehensive vector, a pooling layer is used to reduce the dimensionality of the comprehensive vector. The comprehensive vector is then input into a fully connected layer, which maps the comprehensive vector to the score of the output layer. The output score vector is then:

[0075] ;

[0076] Where kz is the output score vector, containing three elements , and W is the weight matrix of the fully connected layer, and its dimension is the number of neurons. The sum of the lengths of each subarray in a one-dimensional array. For example, if the number of neurons is 3 and the sum of the lengths of each subarray in a one-dimensional array is 20, then the dimension of W is 3. 20, ρ is the bias term, whose dimension is the same as the number of neurons. The dimension and value of the weight matrix determine how the input vector is mapped to the score vector of the output layer, thus affecting the final location probability calculation.

[0077] The output score vector kz is obtained through the Softmax function. kx The formula for converting the location probability P(kx) into different categories is as follows:

[0078] ;

[0079] ;

[0080] ;

[0081] Where π is a constant, it can take the value 3.14; kz kx It is the output score of the kx-th neuron; It is the probability of the output of the first neuron. It is the probability of the output of the second neuron. It is the probability of the output of the third neuron.

[0082] In this step, convolutional layers extract features from the input data, capturing important patterns and features. Pooling layers reduce the dimensionality of the composite vector, lowering computational complexity while retaining important information. This addresses the problem of traditional methods struggling to handle high-dimensional data, resulting in high computational complexity, low efficiency, and difficulty in automatically extracting important features, requiring manual intervention. By using convolutional and pooling layers, high-dimensional data can be processed efficiently, reducing computational complexity, automatically extracting important features, minimizing manual intervention, and improving the model's generalization ability. The composite vector output from the pooling layer is mapped to the score vector of the output layer using the weight matrix W and the bias term ρ. A linear transformation is performed, and the output layer is designed with three neurons, corresponding to low-appropriateness, medium-appropriateness, and high-appropriateness probability values, respectively. This addresses the problem that traditional site selection methods struggle to handle multi-classification problems, cannot accurately assess the suitability of different site selection schemes, and are difficult to map the comprehensive vector to probability values, resulting in a lack of scientific basis for site selection decisions. Through a fully connected layer and the Softmax function, the comprehensive vector can be mapped to three probability values, corresponding to low-appropriateness, medium-appropriateness, and high-appropriateness, respectively, providing a scientific basis for site selection evaluation. The Softmax function normalizes the output score vector to probability values, ensuring the rationality and interpretability of the probability values.

[0083] Step 4: By comprehensively analyzing the integrated vector, location probability, and evaluation index set of each numbered region through the GBDT model, the location score of each numbered region can be obtained.

[0084] Specifically, for each numbered region, the comprehensive vector, the location probabilities of different categories, and the set of evaluation indices are concatenated into a single vector in sequence, such as ST[XT, [Y, TI, T, EI], then input the vector into the GBDT model to obtain the location score DF(ST), calculated as follows:

[0085] ;

[0086] in, h is the scaling factor for the j-th tree model, used to control the contribution of each model tree; j (ST) is the output of the j-th tree model. The tree model settings can be controlled by some parameters, such as tree depth, number of leaf nodes, learning rate, etc. These parameters can be automatically adjusted through methods such as cross-validation to ensure the model's performance and generalization ability. The location score is obtained through multiple iterations to construct multiple tree models h. j(ST), and the outputs of these tree models are weighted and summed to obtain the GBDT model. In this way, the GBDT model can comprehensively consider multiple features, gradually optimize the location score, and help the agent better evaluate and select the optimal location strategy.

[0087] In this step, by concatenating different data groups into a single vector to form a unified input vector, the processability and analyzability of the data are improved. This addresses the problems in traditional site selection methods, such as the scattered display of different types of data, the lack of a unified representation, the potential for different dimensions in different data groups, and the lack of a unified scoring standard, which makes it difficult to compare evaluation results. Through the comprehensive analysis of the GBDT model, the suitability of each numbered region can be accurately evaluated, providing a scientific basis for site selection scoring. The GBDT model comprehensively analyzes the input vector to obtain the site selection score DF(ST). Through multiple iterations, multiple tree models hj(ST) are constructed, and the outputs of these tree models are weighted and summed to obtain the final site selection score. The GBDT model can comprehensively consider multiple features and gradually optimize the site selection score, improving the scientificity and accuracy of site selection. By integrating different evaluation indicators into a unified scoring model, the comparability and scientificity of the scoring results are ensured. Through multiple iterations and the construction of tree models, the GBDT model can handle complex data relationships, improve the model's generalization ability, and solve the problem that traditional methods struggle to handle complex data relationships, resulting in poor model generalization ability.

[0088] Step 5: Based on the calculation results of Step 4, select the site for the micro-weather station from several numbered areas.

[0089] Specifically, the calculation results from step four are traversed and filtered, and the numbered area corresponding to the maximum value is selected as the site for the micro-weather station.

[0090] Example 2:

[0091] Step four also includes: defining the comprehensive vector of each numbered region as the state of that region; calculating the location probability and location score together, and using the calculation result as a reward value to analyze the location value of each region through the Q-learning algorithm model;

[0092] Specifically, the formula for calculating the combined location probability and location score can be achieved by normalizing the location probability and location score, and then calculating the comprehensive analysis value through a weighted average. The formula can be:

[0093] ;

[0094] Where JL is the comprehensive analysis value, serving as the reward value for the Q-learning algorithm, and β1 is... The weighting coefficient can take a value of 0.1, and β2 is... The weighting coefficient can be 0.3, and β3 is... The weighting coefficient of β4 can be 0.3, and β4 is the weighting coefficient of DF(ST), which can also be 0.3.

[0095] It should be noted that the location value iterative update formula of the Q-learning algorithm is:

[0096] ;

[0097] in, This represents the location value of the current numbered area. Initially, the location value can be randomly initialized. As the Q-learning algorithm runs, the location value is gradually updated through the location value update formula. Specifically, after each action is taken, the current location value is updated based on the reward value and the expected value of the next location. The location value update formula is used for optimization, reflecting the long-term value of choosing the current action in the current state. Indicates the maximum location value of the next numbered area, and indicates the state in the next step. Next, iterate through all possible actions. Calculate the location value of each action, and take the maximum value. This value represents the location value of the optimal action in the next state. It is the current state feature of the current numbered region; This indicates the selected location of the micro-weather station under the current conditions; This represents the reward value obtained after taking the current action. =JL; μ represents the current learning step size, which controls the degree of integration of new information. It is usually a value between 0 and 1, and can be selected as 0.6 here; ω represents the discount factor, which represents the weight of future rewards and controls the degree of importance that current learning attaches to future rewards. It is usually a value between 0 and 1, and can be selected as 0.5; si represents the sequence number of the numbered area.

[0098] In the Q-learning algorithm model, at the beginning, a location value table is established. The location value table is initialized to zero or random values. The location value table records the location value of each state-action pair. The model is in the current state. Next action Execution of action It will give an immediate reward value rt+1 and the next state. Update the current state - the location value of the action pair. Then, continue to interact with the environment, repeat the above steps, and continuously update the location value table. After multiple iterations, when the location value in the location value table no longer changes significantly after multiple iterations, such as when the change value of the same numbered area is less than the set change threshold, it can be considered that the location value has converged, and the algorithm obtains the final result, that is, each numbered area has a stable location value, which serves as the location value of each numbered area.

[0099] In this scheme, the comprehensive vector of each numbered region is defined as the region's state. The location probability and location score are calculated together, and the location value of each region is analyzed using a Q-learning algorithm model. This method solves problems in traditional location selection methods such as data dispersion, difficulty in calculating reward values, inaccurate location value assessment, slow model convergence, difficulty in initializing location value, unscientific location value updates, difficulty in comprehensive evaluation, and unreasonable weight settings. By applying the Q-learning algorithm, the location value table can be gradually optimized, the location value of each region can be accurately assessed, and the scientificity and accuracy of location selection can be improved.

[0100] The site selection score and site selection value are combined to obtain a comprehensive evaluation value;

[0101] Specifically, the location score and location value of each numbered area are calculated comprehensively, and the calculation formula can be:

[0102] ;

[0103] in, This represents the comprehensive evaluation value of the numbered region si; This represents the location score for the fusion number region si; 1 represents The weighting coefficient can be set to 0.4. 2 indicates The weighting coefficient can be set to 0.6.

[0104] By comprehensively calculating the site selection score and site selection value, a scientific method for calculating the comprehensive evaluation value is provided, ensuring the rationality and interpretability of the comprehensive evaluation value.

[0105] Example 3:

[0106] The present invention also discloses an intelligent site selection system for micro-weather stations based on low-altitude air routes, which is used to implement the above-mentioned intelligent site selection method for micro-weather stations based on low-altitude air routes.

[0107] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0108] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0109] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent site selection of micro-weather stations based on low-altitude air routes, characterized in that, include: Step 1: Divide the numbered areas where micro-meteorological base stations need to be established into numbered areas, and obtain meteorological condition data, topographic data, low-altitude flight path data, and surrounding environmental data for each numbered area; Step 2: Preprocess the meteorological conditions, topography, low-altitude airway data, and surrounding environment data for each numbered area, and obtain an evaluation index set through data analysis. The evaluation index set includes the degree of impact on low-altitude airway operations, regional topography index, low-altitude airway attractiveness value, and environmental impact index. Normalize the meteorological conditions, topography, low-altitude airway, and surrounding environment data. Analyze the meteorological conditions data using a multiple linear regression model to obtain the degree of impact on low-altitude airway operations for each numbered area. By analyzing topographic data using a topographic index model, the regional topographic index for each numbered area can be obtained. By analyzing low-altitude airway data using a gravity model, the attractive force value of low-altitude airways can be obtained. An environmental impact index can be obtained by analyzing surrounding environmental data using an environmental impact index model. Step 3: Integrate the preprocessed data into a comprehensive vector by channel stacking, and then analyze the comprehensive vector to obtain the location probability of each numbered area; Step 4: By comprehensively analyzing the integrated vector, location probability, and evaluation index set of each numbered region using the GBDT model, a location score can be obtained for each numbered region. The method for obtaining the location score is as follows: The comprehensive vector, location probability, and evaluation index set of each numbered region are concatenated in order into a vector ST[XT, [,Y,TI,T,EI], then ST[XT, Input [Y, TI, T, EI] into the GBDT model to obtain the location score. The calculation formula is as follows: ； Wherein, DF(ST) is the location score for the numbered region. h is the scaling factor for the j-th tree model; j (ST) is the output value of the j-th tree model; XT is the composite vector; Y represents the probability of site selection; TI represents the degree of impact on low-altitude air route operations; T represents the regional topography index; EI represents the environmental impact index. Step 5: Based on the data analysis of Step 4, select the site locations for the micro-weather stations from several numbered areas.

2. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 1, characterized in that, The formula for calculating the impact on low-altitude air route operations is: ； Where X1, X2, X3, X4 and X5 are wind speed, temperature, humidity, air pressure and precipitation, respectively; α0 is the intercept term; α1, α2, α3, α4 and α5 are the regression coefficients of wind speed, temperature, humidity, air pressure and precipitation, respectively; and ξ is the error term.

3. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 2, characterized in that, The formula for calculating the regional topography index is: ； Where H is altitude, S is slope, A is aspect, b is the weighting coefficient of H, c is the weighting coefficient of S, and f is the weighting coefficient of A.

4. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 3, characterized in that, The formula used to calculate the attractive force of low-altitude routes through a gravity model is as follows: ； Where O represents the number of low-altitude air routes passing through this numbered area, Q represents the number of flight missions passing through this numbered area, and D represents the number of low-altitude air routes passing through this numbered area. near This indicates the distance between this numbered area and the nearest low-altitude airway; k is a constant, and τ is the impedance coefficient.

5. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 4, characterized in that, The formula for calculating the environmental impact index is: ； Where B represents the average building height, V represents the vegetation coverage, d is the weighting coefficient of B, and e is the weighting coefficient of V.

6. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 5, characterized in that, The convolutional layers of a convolutional neural network (CNN) organize the preprocessed meteorological data, topographic data, low-altitude flight path data, and surrounding environmental data into one-dimensional arrays. Then, the one-dimensional arrays are stacked according to the channel dimension to generate a comprehensive vector.

7. The intelligent site selection method for micro-weather stations based on low-altitude air routes according to claim 6, characterized in that, The method for obtaining the location probability of each numbered region by analyzing the comprehensive vector is as follows: The composite vector is analyzed using a Convolutional Neural Network (CNN). In a CNN, the output layer is designed with three neurons, enabling it to output three probability values. The composite vector is then input into a fully connected layer, which maps it back to the output layer to obtain the output score vector kz. kx ; The output score vector kz is obtained through the Softmax function. kx Converting to a probability distribution, the formula is: ； ； ； Where π is a constant; kz kx It is the output score of the kx-th neuron; It is the probability of the output of the first neuron. It is the probability of the output of the second neuron. It is the probability of the output of the third neuron.

Images