An Optimization Method and System for Air Quality Prediction Based on Deep Learning

Abstract

The present invention discloses an optimization method and system for air quality prediction based on deep learning, specifically relating to the technical field of air quality prediction optimization. The method includes: Step 1, collecting and analyzing various monitoring data of the fire area; Step 2, constructing a pollutant concentration prediction model; Step 3, correcting the pollutant concentration prediction model; and Step 4, predicting and optimizing the air quality of the fire area. The present invention uses the various monitoring data of the fire area collected in each historical fire, and adopts deep learning technology to train the prediction models of various types of pollutant concentrations, obtaining the corrected prediction models of various types of pollutant concentrations in the fire area, improving the accuracy of the prediction models of various types of pollutant concentrations. The present invention adopts a zoning technical means to realize the analysis and optimization of the air quality in risk areas with different air quality levels, ensuring the effect of the optimization process while saving resources.

Description

Technical Field

[0001] The present invention relates to the technical field of air quality prediction optimization, and specifically relates to an air quality prediction optimization method and system based on deep learning. Background Art

[0002] In recent years, due to global climate change, extreme weather events such as high temperatures, droughts, and strong winds have become more frequent globally, leading to a gradual increase in the frequency and intensity of fires. Fires directly release a large amount of pollutants into the atmosphere. These pollutants not only affect the air quality in the surrounding areas of the fire site but also can spread to farther places under the influence of the wind. These pollutants can cause acute or chronic health problems to the human body. Therefore, it is necessary to study the air quality prediction optimization technology after a fire.

[0003] The prior art, such as a data preprocessing method and system for predicting air quality based on deep learning disclosed in the invention patent application with the publication number of CN115237896B, includes: obtaining a real-time meteorological simulation data set and optimizing it using a weather forecasting model to generate a meteorological forecast data set; then extracting and splicing the meteorological forecast data at the current and target forecast times in the meteorological forecast data set; further using the trained convolutional autoencoder model to perform dimensionality reduction and fusion on the spliced data in space and time to obtain regional meteorological factors; finally, inputting the regional meteorological factors and pollutant monitoring data into the trained air quality forecast model to obtain the air quality forecast result. This invention can significantly reduce the interference of noise and data outliers on the prediction process and results, improve the robustness of the model, and thus effectively improve the numerical accuracy and forecast accuracy of air quality forecasting for a relatively long period in the future.

[0004] Combined with the above solution, it is found that in the prior art, the prediction optimization of air quality mainly constructs a prediction model based on the relationship between meteorological data and pollutants. However, it ignores that the air quality in the fire area is not only affected by meteorological data but also includes various data of fire characteristics and geographical environment. This makes it difficult for the air quality prediction model in the prior art to be adapted and applied to the air quality prediction in the fire area, resulting in insufficient prediction optimization for the air quality in the fire area, and causing a greater potential safety hazard to the human body due to the air quality pollution in the fire area and a greater damage to the ecological system near the fire area. On the other hand, the optimization measures for air quality in the prior art ignore the hierarchical governance according to air quality, leading to waste of resources and reduced optimization efficiency when optimizing the air quality in the fire area, resulting in poor optimization effect of the air quality in the fire area and difficulty in meeting the actual requirements for air quality optimization. Summary of the Invention

[0005] The object of the present invention is to provide an optimization method and system for air quality prediction based on deep learning, which solves the problems existing in the background technology.

[0006] To solve the above technical problems, in the first aspect of the present invention, an optimization method for air quality prediction based on deep learning is provided. The method includes: Step 1, collecting and analyzing various monitoring data of the fire area: obtaining the concentration data of various types of pollutants, various meteorological data, and various fire characteristic data at each collection point in the fire area during the monitoring period, calculating the atmospheric stability coefficient of the fire area, obtaining various geographical environment data of the fire area, and calculating the building density coefficient of the fire area.

[0007] Step 2, constructing a pollutant concentration prediction model: calculating the emission rate of various types of pollutants, and constructing a prediction model for the concentration of various types of pollutants in the fire area based on the building density coefficient and the atmospheric stability coefficient of the fire area.

[0008] Step 3, correcting the pollutant concentration prediction model: obtaining various monitoring data of the fire area collected in each historical fire, and correcting the prediction model for the concentration of various types of pollutants in the fire area.

[0009] Step 4, predicting and optimizing the air quality in the fire area: based on the corrected prediction model for the concentration of various types of pollutants, obtaining the concentration data of various types of pollutants at each monitoring point in the judgment area, calculating the air quality risk coefficient at each monitoring point in the judgment area, constructing a judgment area and dividing the air quality risk level of the judgment area, and implementing the air quality optimization measures corresponding to the air quality risk level of the judgment area.

[0010] In the second aspect of the present invention, an optimization system for air quality prediction based on deep learning is provided. The system specifically includes: a data collection and analysis module, a prediction model construction module, a prediction model correction module, an air quality prediction and optimization module, and a terminal display module.

[0011] The data collection and analysis module is used to obtain the concentration data of various types of pollutants, meteorological data, and various fire characteristic data at each collection point in the fire area during the monitoring period, obtain various geographical environment data of the fire area, and calculate the atmospheric stability coefficient and the building density coefficient of the fire area.

[0012] The prediction model construction module is used to construct a prediction model for the concentration of various types of pollutants in the fire area according to the building density coefficient and the atmospheric stability coefficient of the fire area.

[0013] The prediction model correction module is used to obtain various monitoring data of the fire area collected in each historical fire, and correct the prediction model for the concentration of various types of pollutants in the fire area.

[0014] The air quality prediction optimization module is used to obtain the concentration data of various types of pollutants at each monitoring point within the judgment area, calculate the air quality risk coefficient of each monitoring point within the judgment area, divide the risk areas with different air quality levels, and implement the air quality optimization measures corresponding to the air quality risk level of the judgment area.

[0015] The terminal display module is used to display the air quality compliance level of each judgment area and display the air quality optimization measures of each judgment area.

[0016] Compared with the prior art, the advantages of the present invention are as follows: First, the present invention constructs a prediction model for the concentration of various types of pollutants in the fire area by analyzing the building density and atmospheric stability coefficient in the fire area, fully considering the influence of meteorological data and geographical environment data on the diffusion movement of pollutants in the fire area, and improving the reliability of the actual scenario application of the prediction model for the concentration of various types of pollutants.

[0017] Second, the present invention uses the monitoring data of each fire area collected in each previous fire and adopts the idea of deep learning to train the prediction model for the concentration of various types of pollutants, so as to obtain the corrected prediction model for the concentration of various types of pollutants in the fire area, realizing the accuracy of the prediction model for the concentration of various types of pollutants and providing data support for the subsequent air quality optimization treatment of risk areas with different air quality levels.

[0018] Third, the present invention adopts a zoning technical means to analyze the air quality risk coefficient of different judgment areas, thereby determining the risk levels of different judgment areas, realizing the analysis and optimization of the air quality in risk areas with different air quality levels, ensuring the effect of the optimization treatment, saving resources at the same time, and reflecting the practicability and scientificity of the air quality prediction optimization system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic flow chart of the method of the present invention.

[0021] Figure 2 It is a system structure connection diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] For the convenience of understanding, some professional terms related to the present invention are explained below:

[0024] Satellite remote sensing technology refers to the technology of remotely detecting the ground, ocean, atmosphere, etc. through sensors carried on satellites, such as optical, infrared, radar and other sensors, and collecting various data. These data are usually used in multiple fields such as environmental monitoring, natural disaster assessment, climate research, resource exploration, and urban planning.

[0025] Deep learning refers to the ability to automatically learn features in data through large-scale data and powerful computing capabilities without manually designing features.

[0026] The vertical temperature transformation index refers to an important index in meteorology used to describe the change of temperature with height in the atmosphere, indicating the rate of change of temperature with height in the atmosphere, and is used to evaluate the atmospheric stability.

[0027] The wind speed shear index refers to an index used to measure the degree of change of wind speed with height, that is, the speed of change of wind speed with height, and is used to evaluate the atmospheric stability.

[0028] The lateral diffusion coefficient refers to a statistic of the degree of lateral diffusion of a substance in a fluid, which is affected by building density and atmospheric stability.

[0029] The vertical diffusion coefficient refers to a statistic of the degree of vertical diffusion of a substance in a fluid, which is affected by building density and atmospheric stability.

[0030] The combustion intensity refers to an important parameter used to quantify the intensity or heat release rate in the combustion process, and helps to describe the heat release rate and release efficiency of the combustion process.

[0031] Embodiment 1

[0032] Refer to Figure 1 As shown, in the first aspect of the present invention, an optimized method for air quality prediction based on deep learning is provided. The method includes: Step 1, collecting and analyzing various monitoring data in the fire area: obtaining the concentration data of various types of pollutants, various meteorological data, and various fire characteristic data at each collection point in the fire area during the monitoring period, calculating the atmospheric stability coefficient of the fire area, obtaining various geographical environment data of the fire area, and calculating the building density coefficient of the fire area.

[0033] Step 2: Construct a pollutant concentration prediction model: Calculate the emission rate of each type of pollutant, and based on the building density coefficient of the fire area and the atmospheric stability coefficient of the fire area, construct a prediction model for the concentration of each type of pollutant in the fire area.

[0034] Step 3: Modify the pollutant concentration prediction model: Obtain the monitoring data of each fire area collected in previous fires, and modify the prediction model for the concentration of each type of pollutant in the fire area.

[0035] Step 4: Predict and optimize the air quality in the fire area: Based on the modified prediction model for the concentration of each type of pollutant, obtain the concentration data of each type of pollutant at each monitoring point in the judgment area, calculate the air quality risk coefficient of each monitoring point in the judgment area, construct a judgment area and divide the air quality risk level of the judgment area, and implement the air quality optimization measures corresponding to the air quality risk level of the judgment area.

[0036] In a specific embodiment of the present invention, the method for calculating the building density coefficient of the fire area is as follows: Taking the center point of the fire area as the center, Construct an analysis area with a radius of, and extract the planar geographical feature picture of the analysis area from the database, where represents the preset radius of the analysis area extracted from the database.

[0037] Based on the planar geographical feature picture of the analysis area, extract the planar outlines of each building, and through computer image processing technology, obtain the actual floor area of each building in the analysis area and .

[0038] It should be noted that in a specific embodiment, the specific analysis method for obtaining the actual floor area of each building in the analysis area is as follows: Based on the planar feature picture of the analysis area, through computer deep learning technology, automatically identify the feature area blocks of each building, and count the total number of pixel points of the feature area blocks of each building , based on the area of each pixel point , through the formula: , calculate the area of the feature area block of the building in the planar feature picture of the analysis area , obtain the magnification ratio of the planar feature picture of the analysis area obtained by satellite remote sensing technology , through the formula: , calculate the actual floor area and of each building in each analysis area.

[0039] Through the formula: , calculate the building density coefficient , where represents pi.

[0040] In a specific embodiment of the present invention, for calculating the atmospheric stability coefficient of the fire area, the specific analysis method is as follows: Based on the meteorological data of each collection point in the fire area during the monitoring period, the meteorological data includes: the temperature at each height , the wind speed at each height , where represents the temperature of each collection point in the fire area, represents the wind speed of each collection point in the fire area, represents the height of each collection point in the fire area, , represents the number of each collection point in the fire area, represents the total number of collection points in the fire area.

[0041] It should be noted that in a specific embodiment, the collection points are different coordinates in a three-dimensional space coordinate system.

[0042] It should be noted that in a specific embodiment, for the temperature at different heights and the wind speed at different heights, when the x-axis and y-axis coordinates are determined, the temperature and wind speed at different coordinates on the z-axis are considered. The collection points of height are set in a discrete form by setting the height growth scale of the collection points. For example, with a height growth scale of 2 meters, the temperature is collected through a temperature sensor at the position points on the z-axis with heights of 0, 2, 4, 6, etc. in sequence, and the wind speed is collected through a wind speed sensor.

[0043] Through the formula: , the vertical temperature transformation index is calculated. Through the formula: , the wind speed shear index is calculated.

[0044] Through the formula: , the atmospheric stability coefficient is calculated, where represents the standard vertical temperature change index extracted from the database, represents the standard wind speed shear index extracted from the database.

[0045] In a specific embodiment of the present invention, for constructing the prediction model of the concentration of various types of pollutants in the fire area, the specific analysis method is as follows: The wind direction of the fire area is obtained through a wind direction sensor. Taking the center point of the fire area as the coordinate origin, the vertical direction as the z-axis, the wind direction of the fire area as the x-axis, and the direction horizontally perpendicular to the x-axis as the y-axis, a three-dimensional space coordinate system is established.

[0046] Through the Gaussian gas diffusion model: , a prediction model for the concentrations of various types of pollutants in the fire area is obtained, where represents the emission rate of each type of pollutant, represents at the collection point the wind speed at the location, represents the -th type of pollutant concentration at the collection point at the location, represents the diffusion coefficient of the pollutant in the lateral direction, represents the diffusion coefficient of the pollutant in the vertical direction, represents the natural constant, , where represents the number of each type of pollutant, represents the total number of pollutant types.

[0047] Through satellite remote sensing monitoring technology, the combustion intensity of the fire area is obtained , and through the formula: , the emission rate of each type of pollutant is calculated , where represents the emission factor of each type of pollutant extracted from the database.

[0048] It should be noted that in a specific embodiment, the emission factor of each type of pollutant refers to the emission amount of a certain type of pollutant generated per unit time.

[0049] Through the formula: , the diffusion coefficient of the pollutant in the lateral direction is calculated, where represents the correction factor in the lateral direction.

[0050] Through the formula: , the diffusion coefficient of the pollutant in the vertical direction is calculated, where represents the correction factor in the vertical direction.

[0051] It should be noted that in a specific embodiment, the diffusion coefficient of the pollutant in the lateral direction and the diffusion coefficient in the vertical direction are affected by the building density coefficient and the atmospheric stability coefficient.

[0052] The present invention constructs a prediction model for the concentrations of various types of pollutants in the fire area by analyzing the building density and atmospheric stability coefficient in the fire area, fully considering the influence of meteorological data and geographical environment data on the diffusion movement of pollutants in the fire area, and improving the reliability of the prediction model for the concentrations of various types of pollutants in practical scenario applications.

[0053] In a specific embodiment of the present invention, for the correction of the prediction model of the concentration of various types of pollutants in the fire area, the specific analysis method is as follows: Extract the monitoring data of the fire area collected in each historical fire from the database. First, substitute the emission rates of various types of pollutants, the concentrations of various types of pollutants at each collection point, and the wind speeds at each collection point collected in the first historical fire into the prediction model of the concentration of various types of pollutants in the fire area to obtain the diffusion coefficients in the horizontal direction and the diffusion coefficients in the vertical direction at each collection point in the first historical fire.

[0054] Substitute the atmospheric stability coefficient and the building density coefficient of the fire area collected in the first historical fire into the diffusion coefficient in the horizontal direction at each collection point in the first historical fire to obtain the correction factor of the diffusion coefficient in the horizontal direction at each collection point in the first historical fire. Perform mean processing on the correction factors of the diffusion coefficient in the horizontal direction at each collection point in the first historical fire, and the mean result is used as the correction factor of the diffusion coefficient in the horizontal direction in the first historical fire.

[0055] According to the calculation method of the correction factor of the diffusion coefficient in the horizontal direction in the first historical fire, obtain the correction factors of the diffusion coefficient in the horizontal direction in each historical fire. Perform mean processing on the correction factors of the diffusion coefficient in the horizontal direction in each historical fire to obtain the correction factor of the diffusion coefficient in the horizontal direction.

[0056] Process the correction factor of the diffusion coefficient in the vertical direction according to the calculation method of the correction factor of the diffusion coefficient in the horizontal direction to obtain the correction factor of the diffusion coefficient in the vertical direction.

[0057] Substitute the correction factor of the diffusion coefficient in the horizontal direction and the correction factor of the diffusion coefficient in the vertical direction into the prediction model of the concentration of various types of pollutants in the fire area to obtain the corrected prediction model of the concentration of various types of pollutants in the fire area.

[0058] The present invention uses the monitoring data of the fire area collected in each historical fire and adopts the idea of deep learning to train the prediction model of the concentration of various types of pollutants, so as to obtain the corrected prediction model of the concentration of various types of pollutants in the fire area, realizing the accuracy of the prediction model of the concentration of various types of pollutants and providing data support for the subsequent air quality optimization treatment of risk areas with different air quality levels.

[0059] In a specific embodiment of the present invention, for obtaining the concentration data of various types of pollutants at each monitoring point in the judgment area, the specific analysis method is as follows: Extract the judgment radius from the database , with the origin of coordinates as the center and the x-y axis as the plane, construct a judgment area, and use the random number generation technology with uniform distribution in the judgment area to obtain the coordinates of each monitoring point , , Indicates the numbers of each monitoring point, indicating the total number of monitoring points.

[0060] It should be noted that in a specific embodiment, the specific analysis method for obtaining the coordinates of each monitoring point is as follows: Through the random number generation technology with uniform distribution, coordinates on the x-axis and y-axis are randomly generated within the judgment area, and it is necessary to ensure that the coordinate ranges on the generated x-axis and y-axis do not exceed , and by using the random number generation technology, the coordinate of the z-axis is obtained, and it is necessary to ensure that the range of the obtained z-axis coordinate does not exceed the judgment radius .

[0061] The combustion intensity of the fire area is obtained through satellite remote sensing, the emission rates of various types of pollutants are calculated, and the wind speeds of each monitoring point are obtained through wind speed detection sensors.

[0062] Substitute the coordinates of each monitoring point, the wind speeds of each monitoring point, and the emission rates of various types of pollutants into the prediction model of the concentration of various types of pollutants after correction to obtain the predicted concentrations of various types of pollutants at each monitoring point within the judgment area .

[0063] In a specific embodiment of the present invention, the specific analysis method for calculating the air quality risk coefficient of each monitoring point within the judgment area is as follows: Based on the predicted concentrations of various types of pollutants at each monitoring point within the judgment area , through the formula: , calculate the abnormal judgment value of the predicted concentration of various types of pollutants at each monitoring point within the judgment area , where represents the standard interval of the concentration of various types of pollutants extracted from the database, and through the formula: , calculate the air quality risk coefficient of each monitoring point within the judgment area .

[0064] In a specific embodiment of the present invention, the specific analysis method for dividing the air quality risk level of the judgment area is as follows: Based on the air quality risk coefficients of each monitoring point within the judgment area, mark the monitoring points with as abnormal monitoring points, and count the number of each abnormal monitoring point , where represents the judgment threshold of the air quality risk coefficient extracted from the database, and through the formula: , calculate the non-compliance rate of the air quality within the judgment area , if

[0065] If , the judgment area will be divided into the second-level air quality risk area.

[0066] If , the judgment area will be divided into the third-level air quality risk area, where represents the judgment threshold of the first-level air quality risk area extracted from the database, represents the judgment threshold of the third-level air quality risk area extracted from the database, and the air quality compliance level of the judgment area with as the judgment radius is obtained.

[0067] It should be noted that in a specific embodiment, the division of each level of risk area of the air quality is artificially set by technicians through comprehensive analysis based on the impact of the actual air quality coefficient of each area on the human body and the ecosystem, and is stored in the database.

[0068] Extract the judgment radius growth scale from the database , with as the radius and the coordinate origin as the center of the circle, the part of the judgment area with as the judgment radius that does not intersect with the circle with as the radius is formed into a new judgment area, and the operation of dividing the air quality risk level of the judgment area is performed.

[0069] Successively increase the judgment radius by the growth scale , obtain a new judgment area, perform the operation of dividing the air quality risk level of the judgment area, and obtain risk areas of different levels of air quality.

[0070] It should be noted that in a specific embodiment, the judgment radius increases successively but does not exceed the judgment radius threshold of the preset value extracted from the database , for example, the successively increased judgment radius shall not exceed meters.

[0071] In a specific embodiment of the present invention, for the air quality optimization measures corresponding to the air quality risk level of the judgment area, the specific analysis method is: extract the coverage area of the mobile air purifiers of each level from the database , where , represents the number of the mobile air purifiers of each level.

[0072] It should be noted that in a specific embodiment, the intensity levels of the mobile air purifiers decrease successively according to the numbers.

[0073] The optimization measure for the first-level air quality risk area is: select the mobile air purifier of the first level and extract the area of the first-level air quality risk area , through the formula: , calculate the distribution quantity of the mobile air purifiers of the first level in the first-class air quality risk area , where represents the ceiling function, then in the first-class air quality risk area, mobile air purifiers of the first level are arranged and placed for air quality optimization.

[0074] It should be noted that in a specific embodiment, the specific analysis method for extracting the area of the first-class air quality risk area is: if the judgment radius of the judgment area is the initial , then the area of the first-class air quality risk area is , if the judgment radius of the judgment area is the increased one, then through the formula: , calculate the area of the first-class air quality risk area.

[0075] According to the analysis method of the optimization measures for the first-class air quality risk area, process the second-class air quality risk area and the third-class air quality risk area to obtain the optimization measures for the second-class air quality risk area and the third-class air quality risk area.

[0076] The optimization measures for the first-class air quality risk area, the second-class air quality risk area and the third-class air quality risk area constitute the air quality optimization measures corresponding to the air quality risk level of the judgment area.

[0077] Embodiment 2

[0078] Referring to Figure 2 shown, in the second aspect of the present invention, a deep learning-based air quality prediction and optimization system is provided. The system includes: a data collection and analysis module, a prediction model construction module, a prediction model correction module, an air quality prediction and optimization module, a terminal display module and a database: The data collection and analysis module is used to obtain the concentration data, meteorological data and various fire characteristic data of various types of pollutants at each collection point in the fire area during the monitoring period, obtain various geographical environment data of the fire area, and calculate the atmospheric stability coefficient and building density coefficient of the fire area.

[0079] The prediction model construction module is used to construct a prediction model for the concentration of various types of pollutants in the fire area according to the building density coefficient and atmospheric stability coefficient of the fire area.

[0080] The prediction model correction module is used to obtain the monitoring data of the fire area collected in each previous fire, and correct the prediction model for the concentration of various types of pollutants in the fire area.

[0081] The air quality prediction optimization module is used to obtain the concentration data of various types of pollutants at each monitoring point within the judgment area, calculate the air quality risk coefficient of each monitoring point within the judgment area, divide the risk areas with different air quality levels, and implement the air quality optimization measures corresponding to the air quality risk level of the judgment area.

[0082] The terminal display module is used to display the air quality compliance level of each judgment area and display the air quality optimization measures of each judgment area.

[0083] The database is used to store the planar geographical feature pictures of the analysis area, store the standard vertical temperature change index and the standard wind speed shear index, store the emission factors of various types of pollutants and the fire duration, store the standard intervals of the concentrations of various types of pollutants, store the judgment thresholds of the air quality risk coefficient, store the judgment thresholds of the first-level air quality risk area and the third-level air quality risk area, and store the judgment radius and the judgment radius growth scale.

[0084] It should be noted that in a specific embodiment, in an air quality prediction optimization system based on deep learning, the data acquisition and analysis module is connected to the prediction model construction module, the prediction model construction module is connected to the prediction model correction module, the prediction model correction module is connected to the air quality prediction optimization module, the air quality prediction optimization module is connected to the terminal display module, and the database is connected to the data acquisition and analysis module, the prediction model construction module, the prediction model correction module, and the air quality prediction optimization module.

[0085] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods to replace them. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. A deep learning-based air quality prediction and optimization method, characterized in that: The method comprises: Step 1: Collect and analyze various monitoring data of the fire area: obtain the concentration data of various types of pollutants, meteorological data, and fire characteristics data of each collection point in the fire area during the monitoring period, calculate the atmospheric stability coefficient of the fire area, obtain the geographical environment data of the fire area, and calculate the building density coefficient of the fire area; The specific analysis method for calculating the building density coefficient of the fire area is as follows: The analysis area is constructed with the center point of the fire area as the center and R as the radius, and the plane geographic feature image of the analysis area is extracted from the database, where R represents the radius of the preset analysis area extracted from the database; Based on the planar geographic feature images of the analysis area, the planar outlines of each building are extracted, and the actual area and S of each building in the analysis area are obtained through computer image processing technology; By formula: The building density coefficient η is calculated, where π represents pi; The specific analysis method for calculating the atmospheric stability coefficient of the fire area is as follows: Based on the meteorological data of each collection point in the fire area during the monitoring period, the meteorological data include: the temperature at each height (T i ,h i ), wind speed at each height (V i ,h i ), where T i Indicates the temperature of each collection point in the fire area, V i Represents the wind speed at each collection point in the fire area, h i Indicates the height of each collection point in the fire area, i = 1, 2, ..., m, i represents the number of each collection point in the fire area, and m represents the total number of collection points in the fire area; By formula: The vertical temperature transformation index Γ1 is calculated by the formula: The wind speed shear index Γ2 is calculated; By formula: The atmospheric stability coefficient Ri is calculated, where Γ1′ represents the standard vertical temperature variation index extracted from the database, and Γ2′ represents the standard wind speed shear index extracted from the database; Step 2: Construct a pollutant concentration prediction model: Calculate the emission rate of various types of pollutants, and construct a prediction model for the concentration of various types of pollutants in the fire area based on the building density coefficient and the atmospheric stability coefficient of the fire area; The prediction model for the concentration of various types of pollutants in the fire area is constructed, and the specific analysis method is as follows: The wind direction of the fire area is obtained through the wind direction sensor, and a three-dimensional spatial coordinate system is established with the center point of the fire area as the coordinate origin, the vertical direction as the Z axis, the wind direction of the fire area as the X axis, and the direction perpendicular to the X axis as the Y axis; Using the Gaussian gas diffusion model: The prediction model of the concentration of various types of pollutants in the fire area is obtained, where Q j represents the emission rate of each type of pollutant, V(x,y,z) represents the wind speed at the collection point (x,y,z), C(x,y,z) j represents the concentration of the j-th type of pollutant at the collection point (x, y, z), σ y represents the lateral diffusion coefficient of pollutants, σ z represents the diffusion coefficient of pollutants in the vertical direction, e represents a natural constant, j = 1, 2, ..., n, where j represents the number of each type of pollutant, and n represents the total number of pollutant types; Through satellite remote sensing monitoring technology, the burning intensity F of the fire area is obtained, and the formula is: Q j =F*E j , calculate the emission rate Q of each type of pollutant j , where E j It represents the emission factors of each type of pollutant extracted from the database; By formula: The lateral diffusion coefficient of the pollutant is calculated, where α1 represents the lateral correction factor; By formula: The diffusion coefficient of pollutants in the vertical direction is calculated, where α2 represents the correction factor in the vertical direction; Step 3: Correct the pollutant concentration prediction model: obtain the monitoring data of the fire area collected from each historical fire, and correct the prediction model of the concentration of each type of pollutant in the fire area; Step 4. Predict and optimize the air quality in the fire area: Based on the revised prediction model for the concentration of each type of pollutant, obtain the concentration data of each type of pollutant at each monitoring point in the judgment area, calculate the air quality risk coefficient of each monitoring point in the judgment area, construct the judgment area and divide the air quality risk level of the judgment area, and implement air quality optimization measures corresponding to the air quality risk level of the judgment area.

2. The air quality prediction and optimization method based on deep learning according to claim 1, characterized in that: The specific analysis method for correcting the prediction model of the concentration of various types of pollutants in the fire area is as follows: The monitoring data of the fire area collected from each historical fire are extracted from the database. First, the emission rate of each type of pollutant collected from the first historical fire, the concentration of each type of pollutant at each collection point and the wind speed at each collection point are substituted into the prediction model of the concentration of each type of pollutant in the fire area to obtain the horizontal diffusion coefficient and the vertical diffusion coefficient of each collection point of the first historical fire. Substitute the atmospheric stability coefficient and building density coefficient of the fire area collected during the first fire in history into the lateral diffusion coefficient of each collection point of the first fire in history, obtain the correction factor of the lateral diffusion coefficient of each collection point of the first fire in history, average the correction factors of the lateral diffusion coefficient of each collection point of the first fire in history, and use the average result as the correction factor of the lateral diffusion coefficient of the first fire in history; According to the calculation method of the correction factor of the horizontal diffusion coefficient of the first fire in history, the correction factors of the horizontal diffusion coefficient of each fire in history are obtained, and the correction factors of the horizontal diffusion coefficient of each fire in history are averaged to obtain the correction factors of the horizontal diffusion coefficient; The correction factor of the diffusion coefficient in the vertical direction is processed according to the calculation method of the correction factor of the diffusion coefficient in the horizontal direction to obtain the correction factor of the diffusion coefficient in the vertical direction; Substituting the correction factor of the diffusion coefficient in the lateral direction and the correction factor of the diffusion coefficient in the vertical direction into the prediction model of the concentration of each type of pollutant in the fire area, a corrected prediction model of the concentration of each type of pollutant in the fire area is obtained.

3. The air quality prediction optimization method based on deep learning according to claim 2 is characterized in that: The concentration data of each type of pollutant at each monitoring point in the determination area are obtained, and the specific analysis method is as follows: Extract the judgment radius r from the database, take the coordinate origin as the center of the circle and the xy axis as the plane to construct the judgment area, and use the uniformly distributed random number generation technology within the judgment area to obtain the coordinates (x, y, z) of each monitoring point. k , k=1,2,...,p, k represents the number of each monitoring point, p represents the total number of monitoring points; The burning intensity of the fire area is obtained through satellite remote sensing, the emission rate of various types of pollutants is calculated, and the wind speed at each monitoring point is obtained through wind speed detection sensors; Substitute the coordinates of each monitoring point, the wind speed of each monitoring point and the emission rate of each type of pollutant into the corrected prediction model of each type of pollutant concentration to obtain the predicted concentration C of each type of pollutant at each monitoring point in the judgment area. kj .

4. The air quality prediction and optimization method based on deep learning according to claim 3 is characterized in that: The specific analysis method of calculating and judging the air quality risk coefficient of each monitoring point in the area is as follows: Based on the predicted concentration C of each type of pollutant at each monitoring point in the judgment area kj , through the formula: Calculate the abnormal judgment value (QC) of the predicted concentration of each type of pollutant at each monitoring point in the judgment area kj ,in The standard interval of concentration of each type of pollutant extracted from the database is expressed by the formula: Calculate the air quality risk coefficient ρ of each monitoring point in the judgment area k .

5. The air quality prediction and optimization method based on deep learning according to claim 4 is characterized in that: The specific analysis method for the air quality risk level of the division and judgment area is as follows: Based on the air quality risk coefficient of each monitoring point in the judgment area, ρ k >ρ′ are recorded as abnormal monitoring points, and the number λ of each abnormal monitoring point is obtained by statistics, where ρ′ represents the judgment threshold of the air quality risk coefficient extracted from the database, through the formula: The non-compliance rate γ of the air quality in the judgment area is calculated. If γ>θ2, the judgment area is divided into the first-level air quality risk area; If θ1≤γ≤θ2, the judgment area is divided into the secondary air quality risk area; If γ<θ1, the judgment area is divided into three air quality risk areas, where θ2 represents the judgment threshold of the first air quality risk area extracted from the database, and θ1 represents the judgment threshold of the third air quality risk area extracted from the database, and the air quality compliance level of the judgment area with r as the judgment radius is obtained; Extract the judgment radius growth scale f from the database, take (1+f)*r as the radius, and the coordinate origin as the center of the circle, obtain the judgment area with r as the judgment radius and the non-intersecting part of the circle with (1+f)*r as the radius, form a new judgment area, and perform the air quality risk level division operation of the judgment area; The judgment radius is increased in sequence with the growth scale f to obtain a new judgment area, and the operation of dividing the air quality risk level of the judgment area is performed to obtain risk areas with different air quality levels.

6. The air quality prediction and optimization method based on deep learning according to claim 5, characterized in that: The specific analysis method of the air quality optimization measures corresponding to the air quality risk level of the implementation judgment area is as follows: Extract the coverage area D of each level of mobile air purifier from the database g , where g = 1, 2, 3, g represents the number of each level of mobile air purifier; The optimization measures for the first-level air quality risk area are: select the first-level mobile air purifier, extract the area G1 of the first-level air quality risk area, and use the formula: The distribution number τ1 of the first-level mobile air purifiers in the first-level air quality risk area is calculated, where represents a round-up function, then τ1 first-level mobile air purifiers are placed in the first-level air quality risk area to optimize the air quality; According to the analysis method of optimization measures for air quality level 1 risk areas, the air quality level 2 risk areas and level 3 risk areas are processed to obtain optimization measures for air quality level 2 risk areas and level 3 risk areas; The optimization measures for the air quality level 1 risk areas, level 2 risk areas and level 3 risk areas constitute the air quality optimization measures corresponding to the air quality risk level of the judgment area.

7. A system for executing the air quality prediction optimization method based on deep learning according to any one of claims 1 to 6, characterized in that: The system comprises: The data collection and analysis module is used to obtain the concentration data of various types of pollutants, meteorological data and fire characteristics data of each collection point in the fire area during the monitoring period, obtain the geographical environment data of the fire area, and calculate the atmospheric stability coefficient and building density coefficient of the fire area; A prediction model building module is used to build a prediction model for the concentration of various types of pollutants in the fire area based on the building density coefficient and the atmospheric stability coefficient; The prediction model correction module is used to obtain the monitoring data of the fire area collected from each historical fire and correct the prediction model of the concentration of various types of pollutants in the fire area; The air quality prediction and optimization module is used to obtain the concentration data of various types of pollutants at each monitoring point in the judgment area, calculate the air quality risk coefficient of each monitoring point in the judgment area, divide the risk areas of different air quality levels, and implement air quality optimization measures corresponding to the air quality risk level of the judgment area; The terminal display module is used to display the air quality compliance level of each judgment area and the air quality optimization measures of each judgment area.

Images