A method for estimating urban atmospheric ventilation coefficient based on dual-wavelength lidar

Abstract

The present invention relates to a method for estimating the urban atmospheric ventilation coefficient based on a dual-wavelength laser radar, which belongs to the field of radar detection, including collecting data, calculating the first local volatility of the dual-wavelength laser radar detection data; drawing a change characteristic curve, obtaining the label data of the second local volatility, checking to obtain the label data of the third local volatility, obtaining the data after segmented structure optimization and inputting it into a solution model, and obtaining the corrected laser radar water vapor profile data; calculating the RI index at different heights of the collected data, constructing an original feature data set, and obtaining the estimated data of the urban atmospheric mixing height; constructing a data set and inputting it into an analysis model, and the obtained result is input into an extreme gradient rise model to obtain the corrected horizontal wind speed; and calculating to obtain the estimated value of the urban atmospheric ventilation coefficient. The present invention can improve the accuracy of the estimation result, and give full play to the advantages of various observation equipment and data, and provide support for urban environmental management and meteorological forecasting.

Description

Technical Field

[0001] The invention relates to the field of radar detection, and in particular to a method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar. Background Art

[0002] The atmospheric ventilation coefficient is a key indicator for measuring the air mobility, mixing degree and ventilation capacity in the atmosphere, and has important research significance in meteorology, environmental science and air quality management. The urban atmospheric ventilation coefficient directly affects the diffusion and transmission process of pollutants in the urban atmosphere. Studying the urban atmospheric ventilation coefficient can help evaluate air quality and assess the diffusion range and duration of pollution, and formulate corresponding prevention and control measures. In addition, accurate estimation of the atmospheric ventilation coefficient is of great significance in urban planning and environmental management. The urban heat island effect refers to the phenomenon that the temperature in urban areas is higher than that in surrounding rural areas; the high-density buildings, transportation and energy consumption in cities affect atmospheric ventilation and flow, resulting in poor urban air mobility and high pollutant concentrations. Studying the atmospheric ventilation coefficient can help understand the causes of the urban heat island effect and provide suggestions for improving urban air quality. By analyzing the atmospheric ventilation coefficient, high-pollution areas in cities can be identified, providing a basis for the planning of urban green infrastructure (such as green space, vegetation, permeable pavement, etc.), improving urban air mobility, and helping to improve the lives of residents.

[0003] At present, there are several conventional methods for obtaining atmospheric ventilation coefficients: (1) Field observation and monitoring, which mainly includes radiosondes, lidar and meteorological satellites. Radiosondes can provide meteorological parameter data at different altitudes with high measurement accuracy. However, radiosondes have some limitations: the observation site is often not located in the core area of ​​the city, the observation frequency is low, and the height resolution of the measurement is not detailed enough. Therefore, relying solely on radiosondes cannot effectively obtain the urban atmospheric ventilation coefficient with high temporal and spatial resolution. Lidar obtains the distribution of meteorological elements such as atmospheric wind field and water vapor through inversion technology, and has a high spatial resolution. However, the inversion data of lidar is less accurate, especially in the estimation of parameters such as wind speed, its accuracy is not as good as the directly measured radiosonde data. Although the detection data of meteorological satellites can cover a wide area, its temporal and spatial resolution and accuracy are usually not as good as those of ground remote sensing equipment. Especially in urban areas, satellite data often cannot provide sufficiently detailed information. (2) Theoretical analysis: This method is based on existing fluid dynamics, thermodynamics and meteorological theories, and estimates the atmospheric ventilation coefficient by analyzing the laws of atmospheric movement. Although this method can provide a certain estimate in theory, its accuracy still depends on the spatiotemporal resolution and accuracy of conventional meteorological observation elements. If these basic data are fully guaranteed, this method can obtain better estimation results under certain conditions. (3) Inversion method based on machine learning model: Through inversion technology, remote sensing data (such as radar, LIDAR or satellite data) is used to invert atmospheric wind field information, and then combined with meteorological elements such as wind speed, temperature, and humidity to estimate the atmospheric ventilation coefficient. Compared with traditional methods, machine learning methods can extract the intrinsic characteristics of multi-source data, conduct in-depth analysis and reveal the correlation with the inversion target variable, so they can usually obtain more accurate inversion results. However, the accuracy of the inversion results still depends on the quality and accuracy of remote sensing equipment and meteorological data.

[0004] Although the above methods have their own advantages, how to take into account the impact of city-specific factors (such as geographical environment, population density, factory distribution, temperature and humidity differences between urban and suburban areas, etc.) on the estimation results in the current estimation of urban atmospheric ventilation coefficient is a problem that needs to be solved at present. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for estimating the urban atmospheric ventilation coefficient based on a dual-wavelength laser radar, thereby solving the shortcomings of the prior art.

[0006] The object of the present invention is achieved by the following technical solution: a method for estimating the urban atmospheric ventilation coefficient based on a dual-wavelength laser radar, the estimation method comprising:

[0007] S1. Collect detection data and operating parameter data of dual-wavelength lidar deployed in urban areas, high-frequency radiosonde data close to the location where the lidar is deployed, urban geography and environment data, atmospheric temperature and humidity data near the surface of urban and suburban areas, and socioeconomic data of urban areas, and calculate the first local volatility of dual-wavelength lidar detection data by gradient formula;

[0008] S2. Draw a change characteristic curve according to the generated first local volatility, find the transition point between the flat area and the mutation area of ​​the change characteristic curve, and obtain the label data of the second local volatility in combination with the clustering model, check the label data of the second local volatility to generate the label data of the third local volatility of multiple segmented areas, and construct a segment boundary optimization model to obtain the data after the segment structure is optimized;

[0009] S3, input the data after the segmented structure optimization into the lidar water vapor profile solution model, obtain the corresponding water vapor profile data according to the two schemes set, and combine the data collected in S1 to obtain the water vapor profile data after time-space matching according to the time-space matching algorithm, and obtain the corrected lidar water vapor profile data based on the least squares method;

[0010] S4, calculate the RI index at different heights of the data collected in S1 by using the RI index calculation formula, calculate the corresponding data according to the judgment conditions, construct the original feature data set, obtain the urban impact characteristic factor and the meteorological impact characteristic factor according to the mutual information analysis model, and obtain the estimated data of the urban atmospheric mixing height according to the urban impact characteristic factor and the meteorological impact characteristic factor;

[0011] S5, input the data collected in S1 into the radar velocity azimuth integrated wind field inversion model to obtain the horizontal wind speeds at different distances and heights, construct input data sets and output data sets and input them into the analysis model, and input the obtained results into the extreme gradient ascent model to obtain the corrected horizontal wind speeds at different heights of the lidar;

[0012] S6. Based on the estimated data of the urban atmospheric mixing layer height and the horizontal wind speed at different heights of the calibrated lidar, the estimated value of the urban atmospheric ventilation coefficient is calculated according to the atmospheric ventilation coefficient calculation formula.

[0013] The dual-wavelength laser radar detection data includes the emission wavelength The water vapor Raman scattering echo intensity data, signal-to-noise ratio level data, and emission wavelength data at different distance libraries detected at the same time Detect radial velocity data;

[0014] The dual-wavelength laser radar operating parameter data includes the laser radar wavelength, beam scanning azimuth and elevation, scanning time and cycle, range library length, instrument height, longitude and latitude;

[0015] The high-frequency radiosonde data include water vapor mixing ratio data, temperature, humidity, air pressure, horizontal wind speed, horizontal wind direction, altitude, longitude and latitude data at different times and altitudes measured at relatively low time intervals;

[0016] The urban geography and environment data include urban topography data of slope, aspect and altitude, urban building data of building density, material and height, road and traffic data of traffic density and road coverage, and green space and water body distribution data of greening rate and water surface ratio;

[0017] The socio-economic data include population density data and factory distribution data.

[0018] The method of calculating the first local fluctuation rate of the dual-wavelength laser radar detection data by using the gradient formula includes:

[0019] According to the collected dual-wavelength lidar detection data, the wavelength is taken out The water vapor Raman scattering echo intensity data is recorded as , where i represents different heights;

[0020] Calculate using the gradient formula Gradient , and set the window width to , using the local volatility calculation formula, starting from the first data window Data calculation within The first local volatility , and then slide this window at intervals of M times the distance library length, and calculate The first local volatility of the data within each window in , until the tail of the entire data.

[0021] The step of drawing a change characteristic curve according to the generated first local volatility, finding the transition point between the flat region and the mutation region of the change characteristic curve, and obtaining the label data of the second local volatility in combination with the clustering model specifically includes the following contents:

[0022] According to the generated first local volatility , calculate the local volatility difference between each point and all other points, find the K points with the smallest local volatility difference, and arrange these K local volatility differences in ascending order, and draw a change characteristic curve with the index of the data point as the horizontal axis and the calculated local volatility difference as the vertical axis;

[0023] Find the transition point between the flat area and the mutation area of ​​the characteristic curve, set the index position of the transition point to the neighborhood radius ε, and set the minimum number of neighbors to N_m=2ε. According to the two parameters set for the neighborhood radius ε and the minimum number of neighbors N_m, the first local volatility The data is input into the density-based clustering model, and the clustering model is run to start clustering analysis. The data is divided into low volatility area and high volatility area, and the means of all data in the low volatility area and high volatility area are calculated respectively, which are recorded as and ;

[0024] according to and , the first local volatility Each data in the chart is marked as high, medium and low local volatility points according to the volatility division rules, forming the second local volatility The data of the label.

[0025] The step of checking the label data of the second local volatility to generate label data of the third local volatility of multiple segmented areas, and constructing a segment boundary optimization model to obtain data after segment structure optimization specifically includes the following contents:

[0026] Starting from the first data, check the current second local volatility Whether the data type of the label is the same as the previous data label type, if the same, continue to check, if different, record the position of the current point and the serial number of the segment where the current point is located;

[0027] Restart checking the next segment until the end of the data, and finally generate a third local volatility with multiple segmented areas. Label data;

[0028] Will The corresponding labels, values, and labels of the adjacent points before and after are used as the input data set, and the left and right boundary point positions of the segment are used as the output data set. A segment boundary optimization model based on reinforcement learning is constructed, and the reward function is set to the sum of classification smoothness and classification continuity divided by the standard deviation of the classified data. The segment boundary optimization model is run to obtain the data after segment structure optimization. ;

[0029] according to The segment boundaries, in each segment area, are proportional Multiply by the length of the segment as the window of the segment smoothing filter. If the smoothing filter width is less than W 0 Point, no filtering operation is performed, and the filtered .

[0030] The S3 specifically includes the following contents:

[0031] Will Input it into the lidar water vapor profile solution model, select the two hypotheses that the vertical distribution of water vapor satisfies the exponential decay type and the Gaussian type, run the solution model, and obtain the inverted water vapor profile data under the two hypotheses and ;

[0032] Based on water vapor mixing ratio data collected from high-frequency radiosonde data According to the observation time, altitude, longitude and latitude information, the high-frequency radiosonde data and the dual-wavelength lidar observation data are matched in time and space under the conditions of minimum observation time difference and shortest spatial distance, and the dual-wavelength lidar detection data time and corresponding distance library information that matches the high-frequency radiosonde data in time and space are obtained;

[0033] According to the obtained time and the corresponding distance library information, combined with and Data, obtained after time and space matching and data;

[0034] Calculate the and Data and The data with the smaller difference is selected as the water vapor inversion value at the matching point, thus forming the independent variable data set. As the dependent variable data set;

[0035] The relationship between the independent variable and the dependent variable is fitted based on the least squares method to obtain the corrected lidar water vapor profile data. , respectively The first-order derivative, second-order derivative, and logarithmic first-order derivative are used as the screening condition, and the corresponding heights are recorded as H1, H2, and H3 respectively;

[0036] The RI index at different heights i of the collected data in S1 is calculated by the RI index calculation formula, and the corresponding data is calculated according to the judgment conditions, which specifically includes the following contents:

[0037] The collected high-frequency radiosonde data are input into the RI index calculation formula to calculate the RI index at different altitudes. Index, denoted as RI_D;

[0038] The maximum height corresponding to RI_D<0.25 is used as the judgment condition for calculating the height of the mixed layer, and the height of the mixed layer Hmix is ​​calculated;

[0039] Based on the collected near-surface urban atmospheric temperature Tu, suburban atmospheric temperature Tr, urban atmospheric humidity Qu, and suburban atmospheric humidity Qr, the maximum atmospheric temperature difference between urban and suburban areas ΔTmax=max(Tu-Tr) and the maximum atmospheric humidity ratio ΔQmax=max(Qu / Qr) are calculated respectively. The maximum atmospheric temperature difference in urban area ΔTumax=max(Tu)-min(Tu) and the maximum atmospheric temperature difference in suburban area ΔTrmax=max(Tr)-min(Tr) are also calculated. The maximum atmospheric humidity in urban area ΔQumax=max(Qu)-min(Qu) and the maximum atmospheric humidity in suburban area ΔQrmax=max(Qr)-min(Qr) are also calculated.

[0040] The construction of the original characteristic data set, obtaining the urban impact characteristic factor and the meteorological impact characteristic factor according to the mutual information analysis model, and obtaining the estimated data of the urban atmospheric mixing height according to the urban impact characteristic factor and the meteorological impact characteristic factor specifically include the following contents:

[0041] The collected urban geographic and environmental data, socio-economic data and calculated H1, H2, H3, as well as Tu, Tr, Qu, Qr, ΔTmax, ΔQmax, ΔTumax, ΔTrmax, ΔQumax, ΔQrmax together constitute the original feature data set Data_in_ori;

[0042] Take Hmix as the target data set Data_out_obj, input each type of data in Data_in_ori and the target data set into the mutual information analysis model to calculate the mutual information, sort the calculated mutual information values ​​in descending order from large to small, take the urban geography, environment and socio-economic data corresponding to the mutual information of the top Per ranking ratio to combine into the urban impact characteristic factor, take the meteorological data corresponding to the mutual information of the top Per ranking ratio to combine into the meteorological impact characteristic factor;

[0043] The urban impact characteristic factor and the meteorological impact characteristic factor are combined and recorded as [x 1 ,x 2 ,x 3 ,…,x n ], and then construct the interaction term X=[x 1 x 2 ,x 2 x 3 ,…,x n-1 x n ] and use it as the input data set, and Hmix as the output data set. The mutual information analysis model and the random forest regression model are combined to build an urban atmospheric mixing layer height estimation model. The urban atmospheric mixing layer height estimation model is run to obtain the estimated data of the urban atmospheric mixing layer height. .

[0044] The S5 specifically includes the following contents:

[0045] The collected dual-wavelength lidar radial velocity data and operating parameter data are input into the radar velocity azimuth integrated wind field inversion model to obtain the horizontal wind speed at different distances and heights.

[0046] Taking the horizontal wind speed at different heights of the collected high-frequency radiosonde data as the target, according to the principle of space-time matching with the minimum time difference between the lidar observation and the high-frequency radiosonde observation data and the shortest spatial distance of the observation target, the horizontal wind speed inverted by the lidar after space-time matching, the corresponding height and observation timestamp are selected as the input data set Data_in1;

[0047] The generated water vapor profile , the collected surface temperature and humidity are taken as the input data set Data_in2, and Data_in1 and Data_in2 are merged into the input data set Data_in;

[0048] The horizontal wind speed at different heights of high-frequency radiosonde after time-space matching is taken as the output data set Data_out, and Data_in is input into the autoregressive integrated moving average model and the spatial data analysis model respectively. The autoregressive integrated moving average model and the spatial data analysis model are run respectively to extract the time trend characteristics and spatial dependence characteristics of the input data set;

[0049] Data_in and the extracted time trend features and spatial dependence features are input into the extreme gradient ascent model. Data_out is used as the output data set. Evolutionary and genetic algorithm strategies are selected to optimize the parameter combination of the model. The extreme gradient ascent model is run to obtain the horizontal wind speed at different heights of the corrected lidar. .

[0050] The present invention has the following advantages: a method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar, combining the advantages of high spatiotemporal resolution of laser radar when detecting meteorological elements at different wavelengths, high time resolution and detection accuracy of radiosonde, and considering a variety of city-specific influencing factors, and comprehensively utilizing these data to correct the laser radar detection data, thereby realizing high spatiotemporal resolution urban atmospheric ventilation coefficient intelligent estimation based on dual-wavelength laser radar detection data, which can not only improve the accuracy of the estimation result, but also give full play to the advantages of various observation equipment and data, and provide more reliable support for urban environmental management and meteorological forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the process of the present invention;

[0052] Figure 2 A flowchart for realizing the generation of local volatility label data of the laser radar echo intensity profile of the present invention;

[0053] Figure 3 It is a flow chart for implementing the segmented filtering process of the local fluctuation rate of the laser radar echo intensity profile of the present invention;

[0054] Figure 4 This is a flowchart for implementing the preliminary estimation of the height of the mixing layer based on the water vapor profile inverted by the laser radar of the present invention;

[0055] Figure 5 A flowchart for realizing the urban atmospheric mixing layer height estimation of the present invention;

[0056] Figure 6 The present invention is a flowchart for realizing the correction of horizontal wind speed data at different heights of the laser radar. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided below in conjunction with the drawings is not intended to limit the scope of protection of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application. The present invention is further described below in conjunction with the drawings.

[0058] like Figure 1 As shown, the present invention specifically relates to a method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar, which specifically includes the following contents:

[0059] Step 1: Collect detection data and operating parameter data of dual-wavelength lidar deployed in urban areas; collect high-frequency radiosonde data close to the lidar deployment location; collect urban geographic and environmental data; collect near-surface atmospheric temperature and atmospheric humidity data in urban and suburban areas; and collect socioeconomic data in urban areas.

[0060] Furthermore, the dual-wavelength lidar detection data mainly refers to the water vapor Raman scattering echo intensity data and signal-to-noise ratio level data at different distance bins detected at the emission wavelength of 355 nm, and the radial velocity data detected at the emission wavelength of 1550 nm;

[0061] The operating parameter data of the dual-wavelength lidar mainly refers to the lidar wavelengths of 355 nm and 1550 nm, the beam scanning azimuth and elevation angles, the scanning time and cycle, the range library length of 30 m, the altitude and longitude and latitude of the instrument;

[0062] High-frequency radiosonde data mainly refers to the water vapor mixing ratio data, temperature, humidity, air pressure, horizontal wind speed, horizontal wind direction, altitude, longitude and latitude data at different times and altitudes measured at one-hour intervals;

[0063] Urban geography and environmental data mainly refer to urban terrain data such as slope, aspect and altitude, urban building data such as building density, material and height, road and traffic data such as traffic density and road coverage, and green space and water distribution data such as greening rate and water surface ratio.

[0064] Socio-economic data mainly refers to population density data and factory distribution data;

[0065] Step 2: If Figure 2 As shown, according to the dual-wavelength lidar detection data collected in step 1, the water vapor Raman scattering echo intensity data with a wavelength of 355nm is taken out and recorded as , i represents different heights. First, use the gradient formula to calculate Gradient ; Secondly, set the window width to 4, and use the local volatility calculation formula to calculate the data in the window starting from the first data The first local volatility , and then slide this window at intervals of 2 times the distance library length, and calculate The first local volatility of the data within each window in , until the tail of the entire data.

[0066] The gradient formula is: ;

[0067] The local volatility calculation formula is: , is the average value of the data in the window;

[0068] Step 3: The first local volatility generated from step 2 First, calculate the local volatility difference between each point and all other points, find the 5 points with the smallest local volatility difference, and arrange these 5 local volatility differences in ascending order. Use the index of the data point as the horizontal axis and the calculated local volatility difference as the vertical axis to draw a change characteristic curve, find the transition point between the flat area and the mutation area of ​​the curve, set the index position of the transition point to the neighborhood radius ε, and set the minimum number of neighbors to N_m=2ε. Secondly, according to the two parameters of the set neighborhood radius ε and the minimum number of neighbors N_m, input the local volatility data into the density-based clustering model, run the model to start clustering analysis, and divide the data into low volatility areas and high volatility areas. Afterwards, calculate the means of all data in the low volatility area and the high volatility area, respectively, and record them as and Finally, according to and , the first local volatility Each data in the chart is marked as high, medium and low local volatility points according to the volatility division rules, forming the second local volatility Label data.

[0069] Furthermore, the volatility division rule is: ;

[0070] Step 4: Figure 3 As shown, according to step 3 generated Label data, starting from the first data, check whether the label type of the current data is the same as the label type of the previous data. If yes, continue to check. If not, record the position of the current point and the serial number of the segment where the current point is located, and then restart the check of the next segment until the end of the data, and finally generate a third local volatility with multiple segmented areas Label data. Afterwards, The corresponding labels, values, and labels of the adjacent points before and after are used as the input data set, and the left and right boundary points of the segment are used as the output data set. A segment boundary optimization model based on reinforcement learning is constructed, and the reward function is set to the sum of classification smoothness and classification continuity divided by the standard deviation of the classified data. Running the model can obtain the data after segment structure optimization. Afterwards, according to The segment boundaries, in each segment area, are proportional Multiply by the length of the segment as the window of the segment smoothing filter. If the smoothing filter width is less than 5 points, no filtering operation is performed. .

[0071] Step 5: Figure 4 As shown, the output of step 4 The data is input into the lidar water vapor profile solution model. The water vapor vertical distribution satisfies the exponential decay type and Gaussian type. The model is run to obtain the inverted water vapor profile data under the two hypothetical schemes. and Based on the water vapor mixing ratio data from the high time frequency radiosonde data collected in step 1 The observation time, altitude, longitude and latitude information of the high-frequency radiosonde data are first matched in time and space with the dual-wavelength lidar observation data according to the conditions of minimum observation time difference and shortest spatial distance, and the dual-wavelength lidar detection time and corresponding distance library information that match the high-frequency radiosonde data in time and space are obtained; secondly, according to the obtained time and corresponding distance library information, combined with and Data can be obtained after time and space matching and Afterwards, we calculate the and Data and The data with the smaller difference is selected as the water vapor inversion value at the matching point to form the independent variable data set, and the data with the smaller difference is selected as the water vapor inversion value at the matching point to form the independent variable data set. Finally, the relationship between the independent variable and the dependent variable is fitted based on the least squares method to obtain the corrected lidar water vapor profile data. . The first-order derivative, second-order derivative, and logarithmic first-order derivative are used to find the minimum value of the derivative result as the screening condition, and the corresponding heights are recorded as H1, H2, and H3 respectively.

[0072] Among them, the shortest spatial distance means the sum of the square of the height difference, the square of the longitude difference and the square of the latitude difference between two points in space is the smallest;

[0073] Step 6: Input the high time frequency radiosonde data collected in Step 1 into In the index calculation formula, the index at different heights i can be calculated. The index is recorded as RI_D. The maximum height corresponding to RI_D < 0.25 is used as the judgment condition for calculating the height of the mixing layer, and the clear sky mixing layer height Hmix can be calculated; according to the near-surface urban atmospheric temperature Tu, suburban atmospheric temperature Tr, urban atmospheric humidity Qu, and suburban atmospheric humidity Qr collected in step 1, the maximum atmospheric temperature difference between the urban and suburban areas ΔTmax = max(Tu-Tr), the maximum atmospheric humidity ratio ΔQmax = max(Qu / Qr), the maximum urban atmospheric temperature difference ΔTumax = max(Tu)-min(Tu), the maximum suburban atmospheric temperature difference ΔTrmax = max(Tr)-min(Tr), the maximum urban atmospheric humidity ΔQumax = max(Qu)-min(Qu), and the maximum suburban atmospheric humidity ΔQrmax = max(Qr)-min(Qr) are calculated.

[0074] in, The index calculation formula is:

[0075] ,

[0076] in, is the acceleration due to gravity, is the height near the ground, is the potential temperature, To detect the height, and are the two components of horizontal wind speed respectively.

[0077] Step 7: Figure 5 As shown, the urban geographic and environmental data, socio-economic data collected in step 1, H1, H2, H3 calculated in step 5, and Tu, Tr, Qu, Qr, ΔTmax, ΔQmax, ΔTumax, ΔTrmax, ΔQumax, ΔQrmax calculated in step 6 together constitute the original feature data set Data_in_ori, and Hmix is ​​used as the target data set Data_out_obj. Each type of data in Data_in_ori and the target data set are respectively input into the mutual information analysis model to calculate the mutual information. The calculated mutual information values ​​are arranged in descending order from large to small. The urban geographic, environmental and socio-economic data corresponding to the mutual information with a top ranking of 2 / 3 are combined into the urban impact characteristic factors, and the meteorological data corresponding to the mutual information with a top ranking of 2 / 3 are combined into the meteorological impact characteristic factors. The urban impact characteristic factors and the meteorological impact characteristic factors are combined and recorded as [x 1 ,x 2 ,x 3 ,…,x n ], and then construct the interaction term X=[x 1 x2 ,x 2 x 3 ,…,x n-1 x n ] and use it as the input data set. Hmix is ​​used as the output data set, and the mutual information analysis model and the random forest regression model are combined to build an urban atmospheric mixing layer height estimation model. Running this model can obtain the estimated data of the urban atmospheric mixing layer height. .

[0078] Step 8: Figure 6 As shown, the dual-wavelength lidar radial velocity data and operating parameter data collected in step 1 are input into the radar velocity azimuth integrated wind field inversion model to obtain the horizontal wind speed at different distances and heights. Taking the horizontal wind speed at different heights of the high-frequency radiosonde data collected in step 1 as the target, according to the principle of space-time matching with the minimum time difference between lidar observations and high-frequency radiosonde observations and the shortest spatial distance of the observed target, the horizontal wind speed, corresponding height and observation timestamp inverted by the lidar after space-time matching are selected as the input data set Data_in1. The water vapor profile generated in step 5 , the surface temperature and humidity collected in step 1 are taken together as the input data set Data_in2. Merge Data_in1 and Data_in2 into the input data set Data_in, and take the horizontal wind speed at different heights of high-frequency radiosonde after time-space matching as the output data set Data_out. Input Data_in into the autoregressive integrated moving average model and the spatial data analysis model respectively. Running the models separately can extract the time trend characteristics and spatial dependence characteristics of the input data set; input Data_in together with the extracted time trend characteristics and spatial dependence characteristics into the extreme gradient ascent model, and use Data_out as the output data set. Select evolutionary and genetic algorithm strategies to optimize the parameter combination of the model. Running the model can obtain the horizontal wind speed at different heights of the corrected lidar. .

[0079] Step 9: Estimated data of urban atmospheric mixing layer height obtained in step 7 , combined with the horizontal wind speed at different heights of the corrected lidar obtained in step 8 According to the calculation formula of atmospheric ventilation coefficient, the estimated value of urban atmospheric ventilation coefficient can be calculated. .

[0080] The calculation formula of atmospheric ventilation coefficient is: .

[0081] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used for various other combinations, modifications and improvements, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar, characterized in that: The estimation method includes: S1. Collect dual-wavelength lidar detection data and dual-wavelength lidar operation parameter data deployed in urban areas, high-frequency radiosonde data close to the lidar deployment location, urban geography and environmental data, urban and suburban near-surface atmospheric temperature and atmospheric humidity data, and urban socioeconomic data, and calculate the first local volatility of dual-wavelength lidar detection data using a gradient formula; S2. Draw a change characteristic curve according to the generated first local volatility, find the transition point between the flat area and the mutation area of ​​the change characteristic curve, and obtain the label data of the second local volatility in combination with the clustering model, check the label data of the second local volatility to generate the label data of the third local volatility of multiple segmented areas, and construct a segment boundary optimization model to obtain the data after the segment structure is optimized; S3, input the data after the segmented structure optimization into the lidar water vapor profile solution model, obtain the corresponding water vapor profile data according to the two schemes set, and combine the data collected in S1 to obtain the water vapor profile data after time-space matching according to the time-space matching algorithm, and obtain the corrected lidar water vapor profile data based on the least squares method; S4, calculate the RI index at different heights of the data collected in S1 by using the RI index calculation formula, calculate the corresponding data according to the judgment conditions, construct the original feature data set, obtain the urban impact characteristic factor and the meteorological impact characteristic factor according to the mutual information analysis model, and obtain the estimated data of the urban atmospheric mixing height according to the urban impact characteristic factor and the meteorological impact characteristic factor; S5, input the data collected in S1 into the radar velocity azimuth integrated wind field inversion model to obtain the horizontal wind speeds at different distances and heights, construct input data sets and output data sets and input them into the analysis model, and input the obtained results into the extreme gradient ascent model to obtain the corrected horizontal wind speeds at different heights of the lidar; S6. According to the estimated data of the urban atmospheric mixing layer height and the horizontal wind speed at different heights of the calibrated laser radar, an estimated value of the urban atmospheric ventilation coefficient is calculated according to the atmospheric ventilation coefficient calculation formula; The S3 specifically includes the following contents: Will Input it into the lidar water vapor profile solution model, select the two hypotheses that the vertical distribution of water vapor satisfies the exponential decay type and the Gaussian type, run the solution model, and obtain the inverted water vapor profile data under the two hypotheses and ; Based on water vapor mixing ratio data collected from high-frequency radiosonde data According to the observation time, altitude, longitude and latitude information, the high-frequency radiosonde data and the dual-wavelength lidar observation data are matched in time and space under the conditions of minimum observation time difference and shortest spatial distance, and the dual-wavelength lidar detection data time and corresponding distance library information that matches the high-frequency radiosonde data in time and space are obtained; According to the obtained time and the corresponding distance library information, combined with and Data, obtained after time and space matching and data; Calculate the and Data and The data with the smaller difference is selected as the water vapor inversion value at the matching point, thus forming the independent variable data set. As the dependent variable data set; The relationship between the independent variable and the dependent variable is fitted based on the least squares method to obtain the corrected lidar water vapor profile data. , respectively The first-order derivative, second-order derivative, and logarithmic first-order derivative are used to find the minimum value of the derivative result as the screening condition, and the corresponding heights are recorded as H1, H2, and H3 respectively.

2. The method for estimating the urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 1 is characterized in that: The dual-wavelength laser radar detection data includes the emission wavelength The water vapor Raman scattering echo intensity data, signal-to-noise ratio level data, and emission wavelength data at different distance libraries detected at the same time Detect radial velocity data; The dual-wavelength laser radar operating parameter data includes the laser radar wavelength, beam scanning azimuth and elevation, scanning time and cycle, range library length, instrument height, longitude and latitude; The high-frequency radiosonde data include water vapor mixing ratio data, temperature, humidity, air pressure, horizontal wind speed, horizontal wind direction, altitude, longitude and latitude data at different times and altitudes measured at low time intervals; The urban geography and environment data include urban topography data of slope, aspect and altitude, urban building data of building density, material and height, road and traffic data of traffic density and road coverage, and green space and water body distribution data of greening rate and water surface ratio; The socio-economic data include population density data and factory distribution data.

3. The method for estimating the urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 2 is characterized in that: The method of calculating the first local fluctuation rate of the dual-wavelength laser radar detection data by using the gradient formula includes: According to the collected dual-wavelength lidar detection data, the wavelength is taken out The water vapor Raman scattering echo intensity data is recorded as , where i represents different heights; Calculate using the gradient formula Gradient , and set the window width to , using the local volatility calculation formula, starting from the first data window Data calculation within The first local volatility , and then slide this window at intervals of M times the distance library length, and calculate The first local volatility of the data within each window in , until the tail of the entire data.

4. The method for estimating the urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 3 is characterized in that: The step of drawing a change characteristic curve according to the generated first local volatility, finding the transition point between the flat region and the mutation region of the change characteristic curve, and obtaining the label data of the second local volatility in combination with the clustering model specifically includes the following contents: According to the generated first local volatility , calculate the local volatility difference between each point and all other points, find the K points with the smallest local volatility difference, and arrange these K local volatility differences in ascending order, and draw a change characteristic curve with the index of the data point as the horizontal axis and the calculated local volatility difference as the vertical axis; Find the transition point between the flat area and the mutation area of ​​the characteristic curve, set the index position of the transition point to the neighborhood radius ε, and set the minimum number of neighbors to N_m=2ε. According to the two parameters set for the neighborhood radius ε and the minimum number of neighbors N_m, the first local volatility The data is input into the density-based clustering model, and the clustering model is run to start clustering analysis. The data is divided into low volatility area and high volatility area, and the means of all data in the low volatility area and high volatility area are calculated respectively, which are recorded as and ; according to and , the first local volatility Each data in the chart is marked as high, medium and low local volatility points according to the volatility division rules, forming the second local volatility The data of the label.

5. The method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 4 is characterized in that: The step of checking the label data of the second local volatility to generate label data of the third local volatility of multiple segmented areas, and constructing a segment boundary optimization model to obtain data after segment structure optimization specifically includes the following contents: Starting from the first data, check the current second local volatility Whether the data type of the label is the same as the previous data label type, if the same, continue to check, if different, record the position of the current point and the serial number of the segment where the current point is located; Restart checking the next segment until the end of the data, and finally generate a third local volatility with multiple segmented areas. Label data; Will The corresponding labels, values, and labels of the adjacent points before and after are used as the input data set, and the left and right boundary point positions of the segment are used as the output data set. A segment boundary optimization model based on reinforcement learning is constructed, and the reward function is set to the sum of classification smoothness and classification continuity divided by the standard deviation of the classified data. The segment boundary optimization model is run to obtain the data after segment structure optimization. ; according to The segment boundaries, in each segment area, are proportional Multiply by the length of the segment as the window of the segment smoothing filter. If the smoothing filter width is less than the W0 point, no filtering operation is performed, and the filtered .

6. The method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 5 is characterized in that: The RI index at different heights of the collected data in S1 is calculated by the RI index calculation formula, and the corresponding data is calculated according to the judgment conditions, which specifically includes the following contents: The collected high-frequency radiosonde data are input into the RI index calculation formula to calculate the RI index at different heights i. Index, denoted as RI_D; The maximum height corresponding to RI_D<0.25 is used as the judgment condition for calculating the height of the mixed layer, and the height of the mixed layer Hmix is ​​calculated; Based on the collected near-surface urban atmospheric temperature Tu, suburban atmospheric temperature Tr, urban atmospheric humidity Qu, and suburban atmospheric humidity Qr, the maximum atmospheric temperature difference between urban and suburban areas ΔTmax=max(Tu-Tr) and the maximum atmospheric humidity ratio ΔQmax=max(Qu / Qr) are calculated respectively. The maximum atmospheric temperature difference in urban area ΔTumax=max(Tu)-min(Tu) and the maximum atmospheric temperature difference in suburban area ΔTrmax=max(Tr)-min(Tr) are also calculated. The maximum atmospheric humidity in urban area ΔQumax=max(Qu)-min(Qu) and the maximum atmospheric humidity in suburban area ΔQrmax=max(Qr)-min(Qr) are also calculated.

7. The method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 6 is characterized in that: The construction of the original characteristic data set, obtaining the urban impact characteristic factor and the meteorological impact characteristic factor according to the mutual information analysis model, and obtaining the estimated data of the urban atmospheric mixing height according to the urban impact characteristic factor and the meteorological impact characteristic factor specifically include the following contents: The collected urban geographic and environmental data, socio-economic data and calculated H1, H2, H3, as well as Tu, Tr, Qu, Qr, ΔTmax, ΔQmax, ΔTumax, ΔTrmax, ΔQumax, ΔQrmax together constitute the original feature data set Data_in_ori; Take Hmix as the target data set Data_out_obj, input each type of data in Data_in_ori and the target data set into the mutual information analysis model to calculate the mutual information, sort the calculated mutual information values ​​in descending order from large to small, take the urban geography, environment and socio-economic data corresponding to the mutual information of the top Per ranking ratio to combine into the urban impact characteristic factor, take the meteorological data corresponding to the mutual information of the top Per ranking ratio to combine into the meteorological impact characteristic factor; The urban impact characteristic factor is combined with the meteorological impact characteristic factor, recorded as [x1, x2, x3, …, x n ], and then construct the interaction term X=[x1x2,x2x3,…,x n-1 x n ] and use it as the input data set, and Hmix as the output data set. The mutual information analysis model and the random forest regression model are combined to build an urban atmospheric mixing layer height estimation model. The urban atmospheric mixing layer height estimation model is run to obtain the estimated data of the urban atmospheric mixing layer height. .

8. The method for estimating urban atmospheric ventilation coefficient based on dual-wavelength laser radar according to claim 6 is characterized in that: The S5 specifically includes the following contents: The collected dual-wavelength lidar radial velocity data and operating parameter data are input into the radar velocity azimuth integrated wind field inversion model to obtain the horizontal wind speed at different distances and heights. Taking the horizontal wind speed at different heights of the collected high-frequency radiosonde data as the target, according to the principle of space-time matching with the minimum time difference between the lidar observation and the high-frequency radiosonde observation data and the shortest spatial distance of the observation target, the horizontal wind speed inverted by the lidar after space-time matching, the corresponding height and observation timestamp are selected as the input data set Data_in1; The generated water vapor profile , the collected surface temperature and humidity are taken as the input data set Data_in2, and Data_in1 and Data_in2 are merged into the input data set Data_in; The horizontal wind speed at different heights of high-frequency radiosonde after time-space matching is taken as the output data set Data_out, and Data_in is input into the autoregressive integrated moving average model and the spatial data analysis model respectively. The autoregressive integrated moving average model and the spatial data analysis model are run respectively to extract the time trend characteristics and spatial dependence characteristics of the input data set; Data_in and the extracted time trend features and spatial dependence features are input into the extreme gradient ascent model. Data_out is used as the output data set. Evolutionary and genetic algorithm strategies are selected to optimize the parameter combination of the model. The extreme gradient ascent model is run to obtain the horizontal wind speed at different heights of the corrected lidar. .

Images