CN114898820A - Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model - Google Patents
Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model Download PDFInfo
- Publication number
- CN114898820A CN114898820A CN202210378084.7A CN202210378084A CN114898820A CN 114898820 A CN114898820 A CN 114898820A CN 202210378084 A CN202210378084 A CN 202210378084A CN 114898820 A CN114898820 A CN 114898820A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- air quality
- ozone
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 title claims abstract description 66
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000012544 monitoring process Methods 0.000 claims abstract description 34
- 238000013528 artificial neural network Methods 0.000 claims abstract description 30
- 239000000443 aerosol Substances 0.000 claims abstract description 19
- 239000013618 particulate matter Substances 0.000 claims abstract description 16
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 12
- 239000002245 particle Substances 0.000 claims abstract description 12
- 230000000694 effects Effects 0.000 claims abstract description 11
- 230000007774 longterm Effects 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims description 25
- 238000004140 cleaning Methods 0.000 claims description 20
- 230000005855 radiation Effects 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 230000001681 protective effect Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 15
- 238000002310 reflectometry Methods 0.000 claims description 15
- 230000005540 biological transmission Effects 0.000 claims description 14
- 238000012806 monitoring device Methods 0.000 claims description 14
- 238000004088 simulation Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 11
- 238000003062 neural network model Methods 0.000 claims description 9
- 238000009825 accumulation Methods 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 230000003203 everyday effect Effects 0.000 claims description 5
- 230000033001 locomotion Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 238000011160 research Methods 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000004907 flux Effects 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 239000003570 air Substances 0.000 description 67
- 230000008859 change Effects 0.000 description 11
- 239000003344 environmental pollutant Substances 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 8
- 231100000719 pollutant Toxicity 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000003014 reinforcing effect Effects 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000012821 model calculation Methods 0.000 description 3
- MWUXSHHQAYIFBG-UHFFFAOYSA-N nitrogen oxide Inorganic materials O=[N] MWUXSHHQAYIFBG-UHFFFAOYSA-N 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 239000002243 precursor Substances 0.000 description 3
- 239000000779 smoke Substances 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000001311 chemical methods and process Methods 0.000 description 2
- 230000002354 daily effect Effects 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000003908 quality control method Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 239000012855 volatile organic compound Substances 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 239000000809 air pollutant Substances 0.000 description 1
- 231100001243 air pollutant Toxicity 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 239000012080 ambient air Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008033 biological extinction Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000004062 sedimentation Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 230000007306 turnover Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Chemical & Material Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Primary Health Care (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting and early warning ozone and particulate matters based on a multi-mode air quality model based on a confidence background, which comprises the following steps: acquiring meteorological prediction data, satellite remote sensing data, a pollution source emission list and air quality monitoring data of a target area; inputting the obtained data into a CMAQ air quality model, a CAMx air quality model, a particulate matter ozone neural network prediction model and a satellite remote sensing aerosol inversion model to obtain a model prediction value; adopting a model framework of a neural network, and performing cyclic training on a model predicted value and an actual monitoring value on the basis of long-term historical related data, thereby establishing a multi-mode air quality model; inputting the latest relevant data of the target area into the multi-mode air quality model, and predicting and early warning the ozone and the particulate matters in the target area according to the output result of the multi-mode air quality model. The invention has good prediction and early warning effect on particles and ozone.
Description
Technical Field
The invention relates to the field of meteorological prediction, in particular to a method for predicting and early warning ozone and particulate matters based on a multi-mode air quality model.
Background
At present, the particulate matters and the ozone become main air pollutants which affect the air quality of cities and areas in China, and the cooperative control of the particulate matters and the ozone becomes the focus of the air quality improvement and the key of winning the environmental protection war in blue. There is a complex association between particulate matter and ozone, both of which not only share common precursors, but also interact with each other in the atmosphere through a variety of pathways.
In the aspect of the synergistic treatment of the particulate matter ozone, the existing scientific research means faces challenges. The mechanism of the mutual influence of the ozone and the particulate matters is not clear, the technology of predicting and early warning the ozone and the particulate matters is not mature, and the control system of the government administration department for the ozone and the particulate matters is not complete.
In the aspect of prediction and early warning of particles and ozone, two methods of numerical prediction and statistical prediction are mainly adopted. The numerical prediction mainly utilizes an air quality mode to systematize complex atmospheric physical and chemical modes, establishes a model related to pollutant emission, weather and chemical reactions and simulates the change of air quality; the statistical prediction is established on the basis of the on-line automatic monitoring historical data of the air quality, and the future air quality is predicted according to the change rule of each pollution factor. The two methods have advantages and disadvantages respectively, and data can not be mutually supported, so that the accuracy of a prediction result is unstable.
In the specific implementation process, a large amount of data used for analysis cannot be obtained in time due to instability of long-distance transoceanic transmission, the calculated amount required by high-resolution mode simulation is huge, the existing main calculation frame cannot meet the requirement of over calculation, and the timeliness of prediction and early warning cannot be guaranteed.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects in the prior art and respond to the call of 'insisting on information technology application innovation roads' in China, the invention provides the method for predicting and early warning the ozone and the particulate matters based on the multi-mode air quality model, and the effect of predicting and early warning the particulate matters and the ozone is good.
The technical scheme is as follows: in order to achieve the purpose, the method for predicting and early warning the ozone and the particulate matters based on the multi-mode air quality model comprises the following steps:
the method comprises the following steps: acquiring long-term historical relevant data of a target area, wherein the relevant data comprises meteorological prediction data, satellite remote sensing data, a pollution source emission list and air quality monitoring data;
step two: respectively substituting the acquired historical relevant data into a CMAQ and CAMx third-generation air quality model, a particulate matter ozone neural network prediction model and a satellite remote sensing aerosol inversion model to acquire a model prediction value;
step three: adopting a model framework of a neural network, and performing cyclic training on a model predicted value and an actual monitoring value on the basis of long-term historical related data, thereby establishing a multi-mode air quality model;
step four: inputting the latest relevant data of the target area into the multi-mode air quality model, and predicting and early warning the ozone and the particulate matters in the target area according to the output result of the multi-mode air quality model.
Further, in step two, the specific implementation steps of substituting the data into the two third-generation air quality models of CMAQ and CAMx are as follows:
(a1) establishing a geographical area simulated by a model, determining grid parameters, generating weather prediction data through a WRF weather research and prediction model, and obtaining hourly emission inventory data through an emission source inventory algorithm;
(a2) inputting weather forecast data and emission list data, and respectively simulating SO2, NO2, CO, ozone, PM2.5, temperature, humidity, wind speed, wind direction and rainfall of each grid in a key area in each time period through two air quality models of CMAQ and CAMx;
(a3) performing optimization fitting on the prediction result according to the fitting function by a nonlinear least square fitting algorithm; the nonlinear least squares formula is:
f(x)≈p n (x＝a 0 +a 1 (x-x 0 )+a 2 (x-x 0 ) 2 +...+a n (x-x 0 ) n ；
in the formula, f is a fitting value, and p is a predicted value;
and calculating the error by using a least square method, wherein the error function is as follows:
performing machine learning to obtain a proper final expression of f (x); and then using a fitting formula f (x) to respectively output the predicted fitting values of the CMAQ model and the CAMX model.
Further, in the second step, the specific implementation steps of substituting the data into the particle ozone neural network prediction model are as follows:
(b1) deriving hourly historical data of all automatic air quality monitoring stations in a specific area from a platform database;
(b2) generating a training data list and a verification data list according to time sequence;
(b3) training the training data by using a circulating neural network, taking a plurality of days as a period, and outputting the hourly concentration of ozone and particulate matters for a plurality of days in the period as an output result:
(b4) after training is finished, the cyclic neural network model is stored; after 0 point in the morning every day, the model predicts hourly concentrations of particulate matters and ozone for a plurality of days in a future period according to 24-hour monitoring data of the previous day;
(b5) after entering a period, the neural network is trained again to obtain better model prediction accuracy.
Further, in the step (b1), data cleaning and normalization processing are performed on historical data which are not suitable for model training, such as data missing, numerical value abnormality, and the like; the normalized equation is:
further, in the second step, the specific implementation steps of substituting the data into the satellite remote sensing aerosol inversion model for prediction are as follows:
(c1) downloading MODIS data from a NASA website;
(c2) inverting the apparent reflectivity of the target by a polarized radiation transmission formula, wherein the polarized radiation transmission formula is as follows:
wherein R is apparent reflectivity (apparent reflectivity), F0 is extraterrestrial solar radiation flux, I is atmospheric top radiation rate, mu is the cosine of the zenith angle of the angle of view of the satellite, mu 0 is the cosine of the zenith angle of the sun,
analyzing the relative direction angle of the scattered radiation and the incident solar direction;
(c3) according to a target area, inverting the 0.412um wavelength expression reflectivity, meanwhile, according to the temperature brightness values inverted by the 11nm and 12nm wavelengths and the spatial variation of the 1.38um MODIS reflectivity, judging the meteorological conditions of the target area, if cloud layers or surface ice and snow cover exists, stopping the inversion, and using the predicted value of a historical model as the operation result of the model;
(c4) calculating parameters such as a normalized vegetation index and a normalized ice and snow index to obtain an AOT pre-evaluation value preliminarily;
(c5) identifying the land type and the aerosol type in the target area grid, adjusting corresponding parameters, and performing parameter AOT simulation value;
(c6) fitting the inversion effect of the model, and finally outputting the final optimized AOT value of the target area;
(c7) by means of an algorithm, the future AOT value is predicted.
Further, in step three, the specific implementation steps for establishing the multi-mode model are as follows:
(d1) generating training data, namely acquiring various historical prediction data of the multi-mode model through a database, generating the training data and verification data, taking a plurality of days as a period, taking the data in the period as an input training set, and taking the data in the next period as a verification set;
(d2) data validity processing and data normalization processing;
(d3) establishing an LSMT model, wherein the first layer of the LSMT model is an LSTM network, and the second layer of the LSMT model is a full connection layer; setting Mean Square Error (MSE) as a loss function of the neural network, wherein the calculation method of the loss function is to solve the square sum of the distance between a predicted value and a real value, and the formula is as follows:
then, an adaptive gradient descent algorithm AdaGrad is adopted as backward transmission iteration, and the updating process is as follows:
where s is the accumulation of the square of the gradient, the learning rate divided by the square root of this accumulation plus a small value, which favors the movement of the parameter in the direction closer to the base of the slope, accelerating convergence when updating the parameter;
(d4) after training is finished, the cyclic neural network model is stored, after 0 hour every morning, the model can predict hourly concentration of particulate matters and ozone in a future period according to 24-hour monitoring data of the previous day, and a plurality of days of data are extracted and output;
(d5) and entering the next period, and training the LSTM neural network model again.
Further, in the first step, GFS meteorological prediction data, MODIS satellite remote sensing data, data of air quality monitoring sites, a national pollution source emission mesoscale list and a target area pollution source emission small scale list are downloaded through a download server.
Further, the download server processes the data by using a distributed parallel operation architecture; the download server comprises a main server A, a basic server B and a calculation server resource pool; the main server A is used for being responsible for management of the whole distributed parallel operation, and the multi-mode air quality model is arranged on the main server A; the main server A and the basic server B provide basic calculation power for the multi-mode air quality model, and the basic server B and servers in the calculation power server resource pool provide super calculation power for the multi-mode air quality model; all servers communicate through a private network under the coordination of the main server a.
Further, in the first step, air quality monitoring data of a target area is obtained by setting an air quality monitoring station, and wind speed and wind direction data of the target area are obtained by setting a wind speed and wind direction monitoring device;
a protective frame is arranged on the outer side of the body of the wind speed and direction monitoring device; a cleaning rod is arranged on the outer side of the protection frame; the cleaning rod slides along the outer surface of the protective frame under the driving of the driving part; the whole protective frame is in an arc shape, the driving part is a rotating part driven by wind power, the rotating part is in rotating fit with the supporting seat below the wind speed and direction monitoring device, two wing pieces are arranged on the rotating part at an angle, the cleaning rod is positioned between the two wing pieces, and the wing pieces drive the rotating part to rotate when being subjected to wind; the supporting seat is provided with two limiting blocks for limiting the rotation limiting positions of the rotating part.
Has the advantages that: the method for predicting and early warning ozone and particulate matters based on the multi-mode air quality model has the following beneficial effects:
1) the ozone and the particulate matters are predicted and early-warned through the multi-mode air quality model, the accuracy of the prediction result is high, and the prediction result is more stable;
2) a download server with a distributed parallel operation architecture is designed, the requirement for over calculation is easily met, and the timeliness of prediction and early warning can be guaranteed.
Drawings
FIG. 1 is a block diagram of the overall process of the ozone and particulate prediction and early warning method of the present invention;
FIG. 2 is a flow chart diagram of a model predictive value numerical optimization method for CMAQ and CAMx air quality models;
FIG. 3 is a block diagram of a work flow of a particle ozone neural network prediction model;
FIG. 4 is a schematic structural diagram of a download server of the distributed parallel computing architecture according to the present invention;
FIG. 5 is a schematic diagram illustrating an operation of a download server according to the present invention;
fig. 6 is a schematic structural diagram of an air quality monitoring device according to the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings.
The method for the ozone and particulate matter prediction early warning based on the multi-mode air quality model as shown in the attached figures 1 to 6 comprises the following steps:
the method comprises the following steps: acquiring long-term historical relevant data of a target area, wherein the relevant data comprises meteorological prediction data, satellite remote sensing data, a pollution source emission list and air quality monitoring data;
step two: respectively substituting the acquired historical relevant data into two third-generation air quality models of CMAQ and CAMx, a particulate ozone neural network prediction model and a satellite remote sensing aerosol inversion model to acquire model prediction values;
step three: adopting a model framework of a neural network, and performing cyclic training on a model predicted value and an actual monitoring value on the basis of long-term historical related data, thereby establishing a multi-mode air quality model;
step four: inputting the latest relevant data of the target area into the multi-mode air quality model, and predicting and early warning the ozone and the particulate matters in the target area according to the output result of the multi-mode air quality model.
In the first step, GFS meteorological prediction data and MODIS satellite remote sensing data are downloaded from a target data server by a download server A and a download server B. Specifically, the latest data is obtained from a specific data category according to the requirements of the service scene.
In this step, the two servers are optimized on the physical line to ensure high-speed and stable connection with the overseas network.
And thirdly, considering that some target data servers limit the access quantity, the download speed and the download duration, and the access with large instant download quantity is limited, a set of download mechanism is worked out aiming at the limit, so that the target data servers are prevented from limiting the current.
And thirdly, each time the two target servers finish downloading one data file, the data files are synchronized to the content distribution server in real time, and the content distribution server ensures that the project servers distributed in various places can quickly and stably acquire the data through load balancing, content distribution, scheduling and other modes.
After the download server A and the download server B finish data uploading, the content distribution server checks the data, and generates a download finish mark file after judging that all files are complete and available.
The project server builds machine rooms in different areas according to the geographic position of a client, downloads files required by respective projects after confirming that the files are marked by the content distribution server after downloading is completed, and ensures that the files are transmitted quickly and stably by automatically optimizing a transmission path through an intelligent virtual network according to network congestion and corresponding speed.
All servers are built by using a Linux operating system, respective tasks are realized through a CSHELL script according to the data downloading mechanism, data are obtained from a target data server through a wget command, and file transmission is carried out among the servers through a scp command.
In step one, meteorological prediction data needs to be prepared. After GFS meteorological prediction data are obtained, GFS global meteorological prediction data are input through a WRF meteorological research and prediction model, future meteorological prediction results of multi-level medium and small-scale resolution ratios including China and surrounding areas, provincial surrounding areas, target city surrounding areas and the like are calculated, the future meteorological prediction is obtained, in specific implementation, high-resolution meteorological simulation values of key areas are obtained step by adopting a 27KM,9KM, 3KM and 1KM four-layer grid measurement nesting mode, radiation process parameters, physical process parameters, cloud accumulation convection parameters and road surface process parameters are set according to specific implementation requirements, and prediction accuracy is improved through refining basic data such as an underground bedding surface, LAC (location area code) and the like. The prediction time is determined according to the actual project requirement, and the prediction time is different from 72 hours to 168 hours.
In step one, the emissions inventory data needs to be prepared. Acquiring a national pollution source emission mesoscale list and a target regional pollution source emission small scale list, and calculating emission list data in a prediction early warning time period, wherein the emission list data comprises emission types, emission time and emission source identification; in the specific implementation process:
(e1) the discharge lists with different scales are different in the aspects of statistical mode, time classification, discharge type and data format, and relatively consistent lists need to be output according to project requirements;
(e2) when the source list of the large-scale grid is interpolated to the source list of the small grid, multiple interpolation modes exist, the accuracy of a model simulation result is influenced by different interpolation modes, and a test needs to be carried out at the initial stage of a project;
(e3) the emission list data is used for CMAQ and CAMx to simulate the ambient air, and for obtaining higher accuracy, a specific industry emission calculation model such as BVOCs (biological source volatile organic compound emission model), SMOKE (atmospheric emission source list processing model) and the like is needed.
In step one, preparation of historical air quality monitoring data is required. The air quality monitoring data can reflect the real-time condition of air quality in real environment, generally comprises six air quality data of SO2, NOx, CO, O3, PM2.5 and PM10 and four meteorological data of wind speed, wind direction, temperature, humidity and the like, and the data are uploaded in different time periods according to different equipment hardware.
The server acquires data of national 336 cities of air quality monitoring national control sites in real time, the national control site equipment is maintained timely, the data quality control requirement is strict, and the data reflect the most authoritative air quality of the environment.
In the actual implementation process, reliable station data such as provincial control and city control can be selected according to the actual conditions of the project, and the reliable station data is also included in the multi-mode air quality model calculation. Through a multi-mode air quality model, two key pollution factors of PM2.5 and ozone in the future air quality are predicted 72 hours, 48 hours and 24 hours by hours. And quantitatively tracing the pollution generation mechanism.
The multi-mode air quality model comprises: the model comprises two third-generation air quality models of CMAQ and CAMx, a particle ozone neural network prediction model and an aerosol inversion model based on MODIS satellite remote sensing data.
The invention fully combines the advantages of the CMAQ and the CAMx, is provided with two chemical mechanisms for fitting the prediction result with the real observation result, and finally respectively outputs the ozone and PM2.5 prediction results in the three ranges of 72 hours, 48 hours and 24 hours in the future. Meanwhile, in order to trace the source of the pollution source, two modules, namely a CMAQ-ISAM module and a CAMx-PSAT module, are started.
The third generation air quality model uses each curved surface grid to simulate the physical and chemical processes of the atmospheric environment, and the continuous change rate of the concentration of the pollution factor of each grid can be expressed by a gas continuous equation formula:
wherein Ci is the concentration of the pollution factor, u is the transmission concentration change rate, K is the diffusion concentration change rate, E is the discharge concentration change rate, S is the sedimentation concentration change rate, Rgas is the gaseous reaction concentration change rate, Rplate is the granular reaction concentration change rate, and Rphase is the gaseous conversion concentration change rate.
The air quality model needs to distribute pollutant emission data, meteorological prediction data, geographic environment data and boundary condition data to each grid according to different resolutions under multilayer nesting, then numerical simulation is carried out on each grid for tens of to hundreds of hours in the future through algorithms such as atmospheric physics and chemical modes, and the calculation amount and the data generation amount are huge.
The latest versions of international CMAQ and CAMx continue to use the traditional Fortran implementation mode, and do not consider computer cluster distributed and the latest efficient parallel computation.
In order to improve the accuracy of the air quality model, the output results predicted by the CMAQ and CAMx models are fitted with actual monitoring data in a nonlinear least square fitting mode to obtain a fitting function for further fitting the predicted results.
In the second step, the specific implementation steps of substituting the data into the particle ozone neural network prediction model are as follows:
(a1) according to the requirements of each project, a geographical area of model simulation is made, grid parameters are determined, a mode of nesting four layers of grids of 27KM,9KM, 3KM and 1KM is generally adopted, and the purpose of obtaining a high-resolution prediction result of a target area is achieved.
After the model grid parameters are determined, weather forecast data are generated through WRF, and hourly emission inventory data are obtained through an emission source inventory algorithm.
Emission source data for this step; emission source data is optimized and labeled for overhead emission pollution sources and heavy point pollution sources generally through a SMOKE model under the precision of 1 × 1 KM.
In the step, in order to improve the prediction precision of the meteorological model, the precision of various data is improved by adopting the technologies of land underlying surface replacement, land utilization rate updating, target region SMOKE emission list manufacturing, urban canopy simulation and the like.
(a2) Inputting weather forecast data and emission list data, and respectively simulating SO2, NO2, CO, ozone, PM2.5, temperature, humidity, wind speed, wind direction and rainfall of each grid in a key area in each time period through two air quality models of CMAQ and CAMx;
at this step, the existing code is decades ago, and with the improvement of simulation area and resolution, the calculation amount is multiplied compared with the former, and more time is needed for completing the simulation of several days.
The test on the computing efficiency of CMAQ and CAMx shows that the CPU core number and the computing speed are not in a linear relation during parallel operation, but the more the CPU core number is, the faster the computing speed is, because the third generation air quality model adopts a relatively old technology during code implementation, the implementation codes of the bottom layers of the CMAQ and CAMx models are optimized: replacing a more efficient compiling library, and installing a distributed parallel operation dependent library; optimizing a parallel operation scheduling mechanism; and optimizing a file output mechanism to enable the model to be used for distributed computation.
Through optimization and comprehensive consideration of implementation cost, the model calculation efficiency is improved by 4 times on the basis of basic calculation capacity, and meanwhile, by accessing the distributed server cluster, the calculation capacity can be additionally increased and the model calculation is accelerated according to the abundant degree of calculation capacity resources.
(a3) And performing optimization fitting on the prediction result according to the fitting function by a nonlinear least square fitting algorithm.
Nonlinear least squares formulation:
fitting Using higher order polynomials
f(x)≈p n (x)＝a 0 +a 1 x-x 0 )+a 2 (x-x 0 ) 2 +...+a n (x-x 0 ) n ；
f is the fitting value, p is the predicted value, and in actual operation, n is often set to 3.
The error is calculated by the least square method, and the error function is:
and finally obtaining a proper final expression of f (x) through machine learning.
And (5) respectively outputting predicted fitting values of the CMAQ model and the CAMX model by using a fitting formula f (x).
In consideration of the calculation amount and the actual service requirement, the hourly data of the future 72 hours are generally selected for output in the specific implementation process.
At present, the air quality prediction modes are mainly classified into two types: a mechanistic predictive approach and a non-mechanistic predictive approach. The CMAQ and CAMx air quality models use a mechanical prediction mode, and the future air quality is predicted by establishing a relatively complete pollution source emission source list, relatively accurate meteorological field conditions, atmospheric pollutant diffusion processes and other physical and chemical processes.
The non-mechanistic prediction does not need a complex pollutant boundary field and a meteorological boundary field, and only needs to capture data characteristics through historical pollutant data to obtain a pollutant concentration change rule.
PM2.5 and ozone have common precursors, namely nitrogen oxides and volatile organic compounds, and the effects of the precursors in the process of forming particulate matters and ozone are different. The interaction relationship between them is complicated. The concentration of ozone and particulate matter can be predicted for a period of time in the future by training and learning historical air quality data and meteorological data using a machine-learned neural network algorithm.
The cyclic neural network takes sequence data as input, all nodes are connected in a chain mode, the forward recursion is realized, the nonlinear characteristic learning of the sequence is advantageous, and the cyclic neural network is particularly suitable for predicting the concentration of particulate matters and ozone.
In the second step, the specific implementation steps of substituting the data into the particle ozone neural network prediction model are as follows:
(b1) hourly historical data (e.g., 2 years) for all automatic air quality monitoring stations for a particular region is derived from the platform database. In the process, data loss or abnormality caused by equipment failure, data quality control adjustment and the like can occur, and the data is not processed for a while; the method comprises the following steps of preprocessing historical data, and cleaning and normalizing data which are not suitable for model training and are absent in data, abnormal in numerical value and the like, wherein a normalization equation is as follows:
because the data used by the model are from environmental air quality monitoring national control sites and provincial control sites, operation and maintenance teams are equipped 24 hours all the year round, the condition that data are abnormal or lost for a plurality of consecutive days is avoided, and for the data lost for one day exceeding 8 hours, all the data of the day are removed and are not listed in a training sequence. In consideration of the most extreme case, even if the data preprocessed by the data preprocessing may be missing for several days, the model accuracy can be guaranteed.
(b2) And generating a training data list and verifying a data structure of the training data of the data list according to time sequence.
(b3) Training data by using a recurrent neural network, wherein the input step length of each training is 7, so that the accumulation and the dissipation of atmospheric pollutants can be simulated conveniently. The particle ozone neural network prediction model adopts a four-layer hidden layer structure, and the output result is the hourly concentration of ozone and particles for 7 days in the future.
(b4) After training is finished, the cyclic neural network model is stored, and after 0 point in the morning every day, the model can predict hourly concentrations of particulate matters and ozone for 7 days in the future according to 24-hour monitoring data of the previous day.
(b5) And training the neural network again every 7 days to obtain better model prediction accuracy.
Obtaining regional earth surface and atmospheric information through a satellite remote sensing image inversion technology is one of research hotspots of environmental remote sensing. Aot (aerosol Optical thickness) Optical aerosol thickness is the path integral of the aerosol extinction function along the direction of propagation from the ground to the top atmosphere, a physical quantity that characterizes the degree of attenuation of the aerosol to solar radiation. Experts and scholars at home and abroad always search for a method for estimating the concentration of the surface atmospheric pollutants by using AOT information, but the prediction effect of the method which only depends on satellite remote sensing data is not ideal. Considering that aerosol thickness data has obvious gains for particle prediction and ozone prediction, an aerosol inversion model based on MODIS data is introduced to predict future AOT value every day.
The aerosol inversion model mainly uses a simplified deep blue algorithm and a 6S radiation transmission model, and aims to provide a parameter with strong correlation for O3 and particulate matters for the multi-mode model. The accuracy of the multi-mode model neural network fusion is improved.
In the second step, the specific implementation steps of substituting the data into the satellite remote sensing aerosol inversion model for prediction are as follows:
(c1) and (3) downloading MODIS data from an NASA (network application server) website, and directly downloading Level 1 data subjected to algorithm calibration and coordinate processing by considering factors such as aging of optical devices and sensors. And downloading related data according to the time points when the satellite sweeps China.
(c2) Inverting the apparent reflectivity of the target by a polarized radiation transmission formula, wherein the polarized radiation transmission formula is as follows:
wherein R is apparent reflectivity (apparent reflectivity), F0 is extraterrestrial solar radiation flux, I is atmospheric top radiation rate, mu is the cosine of the zenith angle of the angle of view of the satellite, mu 0 is the cosine of the zenith angle of the sun,
the direction angle relative to the incident solar direction is analyzed for scattered radiation propagation.
(c3) And (3) according to the target area, inverting the 0.412um wavelength expression reflectivity, simultaneously judging the meteorological conditions of the target area according to the temperature brightness values inverted by the 11nm and 12nm wavelengths and the spatial variation of the 1.38um MODIS reflectivity, stopping the inversion if cloud cover or surface ice and snow cover exists, and using the predicted value of the historical model as the operation result of the model.
(c4) And calculating parameters such as normalized vegetation index (NDVI) and normalized snow and ice index (NDSI) to obtain an AOT pre-estimated value preliminarily.
(c5) And adjusting corresponding parameters through identifying the land type (water body, city, vegetation and desert) and the aerosol type (sand dust type, city type, ocean type and biological combustion type) in the target area grid through the parameter AOT simulation value.
(c6) And fitting the model inversion effect according to 60 site data of AERONET in China, and finally outputting the final optimized AOT value of the target area.
(c7) By algorithm, the future 3-day daily value is predicted.
The multi-mode air quality model comprises four independent models, each model has a special advantage in particulate matter and ozone simulation, the CMAQ and CAMx models can predict all parameters of future air quality and weather, the prediction accuracy of the particulate matter ozone neural network prediction model on ozone and PM2.5 is high, and the aerosol inversion model can provide an AOT important index. The multi-mode air quality model of this application adopts the model framework of LSTM neural network, on the basis of long-term historical data, carries out the circulation training to model predicted value and actual monitoring value, except can remembering the long-term law of target area air pollution, can also synthesize the air quality of target area in 7 days recently, analysis short-term characteristic.
In step three, the specific implementation steps for establishing the multi-mode model are as follows:
(d1) and generating training data, namely acquiring various historical prediction data of the multi-mode model through a database, paying attention to historical monitoring data of the site, and generating training data and verification data. Seven days of 24 hours data were used as input training set and 7 days in the future were used as validation set.
(d2) Data validity processing and data normalization processing.
(d3) Establishing an LSMT model, wherein the first layer is an LSTM network, the second layer is a full connection layer, the mean square error MSE is set as a loss function of the neural network, and the formula of the loss function is as follows:
the calculation method is to calculate the square sum of the distance between the predicted value and the real value;
an adaptive gradient descent algorithm AdaGrad is adopted as backward transmission iteration: the updating process is as follows:
where s is the accumulation of the square of the gradient, the learning rate divided by the square root of this accumulation plus a small value facilitates the movement of the parameter in a direction closer to the base of the slope, thus speeding up convergence.
(d4) After training is finished, the cyclic neural network model is stored, after 0 point in the morning every day, the model can predict hourly concentration of particles and ozone in 7 days in the future according to 24-hour monitoring data of the previous day, and data of 3 days are extracted and output.
(d5) Every 7 days for one cycle, the LSTM neural network model was retrained again.
The multi-mode air quality model integrates four prediction models, and requires a large amount of data processing and numerical simulation calculation. In order to ensure the timeliness of model prediction and the guidance of daily environment management, the invention designs a distributed parallel operation framework specially used for scientific calculation.
The multi-mode air quality model is arranged on the main server A and is responsible for the management of the whole distributed parallel operation, and the multi-mode air quality model comprises the functions of task allocation, communication coordination of sub servers, memory management and the like. The debugging operation for the whole model is carried out on the main server A.
And the basic server B and servers in the calculation power server resource pool provide super calculation power for the multi-mode air quality model, and all the servers are communicated through a special network under the coordination of the main server.
The main server A and the basic server B provide basic computing power for the multi-mode air quality model, and meanwhile computing power resources are additionally loaded under the condition of abundant computing power by evaluating the load rate of the power resource pool.
In the first step, air quality monitoring data of a target area is obtained by setting an air quality monitoring station, and wind speed and wind direction data of the target area are obtained by setting a wind speed and wind direction monitoring device.
Traditional wind speed and direction monitoring devices are difficult to maintain when in use, and are easily influenced by external sundries in long-term use, for example, fallen leaves are accumulated in a large number and attached to cause the wind speed and direction monitoring devices to rotate, so that the reliability of monitoring results of the wind speed and direction monitoring devices is influenced. If only rely on the manual clearance debris of maintainer, the maintenance cycle is longer, hardly in time clears up the debris. Therefore, an automatic cleaning mechanism is added to the wind speed and direction monitoring device.
A protective frame 1 is arranged on the outer side of the body of the wind speed and direction monitoring device; the protective frame 1 can block foreign matters; a cleaning rod 2 is arranged on the outer side of the protective frame 1; the cleaning rod 2 is driven by the driving part to slide along the outer surface of the protective frame 1; thereby cleaning foreign matters attached to the surface of the protection frame 1.
The driving part can be electrically driven, a positioning module is arranged on the motor, and the motor drives the cleaning rod 2 to rotate back and forth for one time or multiple times at a certain interval.
The drive member may also be wind driven. Be provided with on the supporting seat 6 and rotate the piece, it is provided with the structure that receives the wind to rotate on rotating the piece, the structure that receives the wind drives under the wind-force effect rotate the piece and rotate, it drives to rotate when rotating the piece and rotate clearance pole 2 edge the surface of protective frame 1 slides, still be provided with on the supporting seat 6 and carry out spacing locating part to rotating a pivoted extreme position. When clearance pole 2 slided to one side of protection frame 1 back along the surface of protection frame 1, blocked by the locating part, prevented to rotate and continue to rotate, when opposite wind-force effect, clearance pole 2 slided to the opposite side of protection frame 1 along the surface of protection frame 1 to make clearance pole 2 can follow the surface reciprocating sliding of protection frame 1.
The wind-receiving structure comprises two fins 3 arranged at an angle, and the included angle of the two fins 3 is smaller than or equal to 90 degrees. Clearance pole 2 is located the mid-plane of two fin 3, does benefit to clearance pole 2 and slides along the surface of protection frame 1 is reciprocal.
Connecting rods 4 are respectively arranged below two ends of the cleaning rod 2, each connecting rod 4 is a vertical rod body, the upper end of each connecting rod 4 is connected with one end of the cleaning rod 2, a sleeve 5 is arranged at the lower end of each connecting rod 4, and an inserting shaft 7 which is matched with the sleeve 5 in a rotating mode is arranged on the supporting seat 6. Two fins 3 are all arranged on one sleeve 5, so that the structure is more simplified and reasonable.
The limiting part comprises two limiting blocks which are symmetrically arranged, and the two limiting blocks are respectively positioned at two sides of the connecting rod 4 and used for limiting the movement position of the wing piece 3. When the lower plate surface of one fin 3 is connected with the limiting block for limiting, the other fin 3 is in a vertical state approximately, so that the fin 3 drives the whole rotating part to rotate more easily when receiving wind.
The whole protection frame 1 is arc-shaped and consists of a plurality of arc-shaped rods which are arranged at intervals. The below of clearance pole 2 is provided with a plurality of tooth pieces, the tooth piece corresponds and is located the clearance of two adjacent arc poles, and the width of tooth piece is less than the interval of adjacent arc pole, and the setting of tooth piece can strengthen the clearance ability of clearance pole 2.
The two sides of the supporting seat 6 are provided with tooth-shaped grooves for accommodating the tooth blocks to pass through, the tooth blocks move to the tooth-shaped grooves along the gaps of the arc-shaped rods, foreign matters are easier to scrape from the outer surface of the protective frame 1, and the cleaning effect of the cleaning rod 2 is better.
The middle lower part of the arc-shaped rod is provided with a reinforcing frame, the reinforcing frame is composed of two vertical rods and a cross rod arranged at the upper ends of the two vertical rods, and the middle lower part of the arc-shaped rod is fixedly connected with the cross rod through a vertical reinforcing rod. The length of the reinforcing rod is greater than that of the tooth block, so that the tooth block is prevented from moving and interfering with the cross rod. The structural strength of the protective frame 1 can be improved by the arrangement of the reinforcing frame.
The upper surface of the supporting seat 6 is arc-shaped, and foreign matters falling on the upper surface of the supporting seat 6 are more easily slipped off.
When the driving part is driven by wind power, the rotating part always falls to one side under the balanced state due to the structure of the rotating part, one wing piece 3 is vertical, and the other wing piece 3 is horizontal; the vertical wing pieces 3 can drive the rotating piece to turn over after being acted by wind power, so that the cleaning rod 2 is driven to slide along the surface of the protective frame 1 in the positive direction to clean foreign matters until the original vertical wing pieces 3 become horizontal and the original horizontal wing pieces 3 become vertical, and balance is achieved again; when the wind direction changes and generates opposite wind force, the wing pieces 3 can drive the cleaning rod 2 to reversely slide along the surface of the protective frame 1; thereby make clearance pole 2 can follow protection frame 1's surface reciprocating motion under the wind-force effect, realize the clearance to debris, the clearance pole can in time clear up most debris, but when in actual use, also probably touch the debris that a small part can't be cleared up, can thoroughly clear away debris when the maintenance by wind speed wind direction monitoring devices's maintainer.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (9)
1. The method for predicting and early warning ozone and particulate matters based on the multi-mode air quality model is characterized by comprising the following steps of: the method comprises the following steps:
the method comprises the following steps: acquiring long-term historical relevant data of a target area, wherein the relevant data comprises meteorological prediction data, satellite remote sensing data, a pollution source emission list and air quality monitoring data;
step two: respectively substituting the acquired historical relevant data into two third-generation air quality models of CMAQ and CAMx, a particulate ozone neural network prediction model and a satellite remote sensing aerosol inversion model to acquire model prediction values;
step three: adopting a model framework of a neural network, and performing cyclic training on a model predicted value and an actual monitoring value on the basis of long-term historical related data, thereby establishing a multi-mode air quality model;
step four: inputting the latest relevant data of the target area into the multi-mode air quality model, and predicting and early warning the ozone and the particulate matters in the target area according to the output result of the multi-mode air quality model.
2. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 1, characterized in that: in the second step, the specific implementation steps of substituting the data into the CMAQ and CAMx third-generation air quality models are as follows:
(a1) establishing a geographical area simulated by a model, determining grid parameters, generating weather prediction data through a WRF weather research and prediction model, and obtaining hourly emission inventory data through an emission source inventory algorithm;
(a2) inputting weather forecast data and emission list data, and respectively simulating SO2, NO2, CO, ozone, PM2.5, temperature, humidity, wind speed, wind direction and rainfall of each grid in a key area in each time period through two air quality models of CMAQ and CAMx;
(a3) performing optimization fitting on the prediction result according to the fitting function by a nonlinear least square fitting algorithm; the nonlinear least squares formula is:
f(x)≈p n (x)＝a 0 +a 1 (x-x 0 )+a 2 (x-x 0 ) 2 +...+a n (x-x 0 ) n ；
in the formula, f is a fitting value, and p is a predicted value;
and calculating the error by using a least square method, wherein the error function is as follows:
performing machine learning to obtain a proper final expression of f (x); and then using a fitting formula f (x) to respectively output the predicted fitting values of the CMAQ model and the CAMX model.
3. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 2, characterized in that: in the second step, the specific implementation steps of substituting the data into the particle ozone neural network prediction model are as follows:
(b1) deriving hourly historical data of all automatic air quality monitoring stations in a specific area from a platform database;
(b2) generating a training data list and a verification data list according to time sequence;
(b3) training the training data by using a circulating neural network, taking a plurality of days as a period, and outputting the hourly concentration of ozone and particulate matters for a plurality of days in the period as an output result:
(b4) after training is finished, the cyclic neural network model is stored; after 0 point in the morning every day, the model predicts hourly concentrations of particulate matters and ozone for a plurality of days in a future period according to 24-hour monitoring data of the previous day;
(b5) after entering a period, the neural network is trained again to obtain better model prediction accuracy.
4. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 3, characterized in that: in the step (b1), data cleaning and normalization processing are performed on historical data which are not suitable for model training, such as data missing, numerical value abnormality and the like; the normalized equation is:
5. the method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 1, characterized in that: in the second step, the specific implementation steps of substituting the data into the satellite remote sensing aerosol inversion model for prediction are as follows:
(c1) downloading MODIS data from a NASA website;
(c2) inverting the apparent reflectivity of the target by a polarized radiation transmission formula, wherein the polarized radiation transmission formula is as follows:
wherein R is apparent reflectivity (apparent reflectivity), F0 is extraterrestrial solar radiation flux, I is atmospheric top radiation rate, mu is the cosine of the zenith angle of the angle of view of the satellite, mu 0 is the cosine of the zenith angle of the sun,
analyzing the relative direction angle of the scattered radiation and the incident solar direction;
(c3) according to a target area, inverting the 0.412um wavelength expression reflectivity, meanwhile, according to the temperature brightness values inverted by the 11nm and 12nm wavelengths and the spatial variation of the 1.38um MODIS reflectivity, judging the meteorological conditions of the target area, if cloud layers or surface ice and snow cover exists, stopping the inversion, and using the predicted value of a historical model as the operation result of the model;
(c4) calculating parameters such as a normalized vegetation index and a normalized ice and snow index to obtain an AOT pre-evaluation value preliminarily;
(c5) identifying the land type and the aerosol type in the target area grid, adjusting corresponding parameters, and performing parameter AOT simulation value;
(c6) fitting the inversion effect of the model, and finally outputting the final optimized AOT value of the target area;
(c7) by means of an algorithm, the future AOT value is predicted.
6. The method for ozone and particulate matter prediction and early warning based on the multi-mode air quality model according to claim 1, characterized in that: in step three, the specific implementation steps for establishing the multi-mode model are as follows:
(d1) generating training data, namely acquiring various historical prediction data of the multi-mode model through a database, generating the training data and verification data, taking a plurality of days as a period, taking the data in the period as an input training set, and taking the data in the next period as a verification set;
(d2) data validity processing and data normalization processing;
(d3) establishing an LSMT model, wherein the first layer of the LSMT model is an LSTM network, and the second layer of the LSMT model is a full connection layer; setting Mean Square Error (MSE) as a loss function of the neural network, wherein the calculation method of the loss function is to solve the square sum of the distance between a predicted value and a real value, and the formula is as follows:
then, an adaptive gradient descent algorithm AdaGrad is adopted as backward transmission iteration, and the updating process is as follows:
where s is the accumulation of the square of the gradient, the learning rate divided by the square root of this accumulation plus a small value, which favors the movement of the parameter in the direction closer to the base of the slope, accelerating convergence when updating the parameter;
(d4) after training is finished, the cyclic neural network model is stored, after 0 point every morning, the model can predict hourly concentration of particulate matters and ozone in a future period according to 24-hour monitoring data of the previous day, and a plurality of days of data are extracted and output;
(d5) and entering the next period, and training the LSTM neural network model again.
7. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 1, characterized in that: in the first step, GFS meteorological prediction data, MODIS satellite remote sensing data, air quality monitoring site data, a national pollution source emission mesoscale list and a target area pollution source emission small scale list are downloaded through a download server.
8. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 7, characterized in that: the download server processes data by using a distributed parallel operation architecture; the download server comprises a main server A, a basic server B and a calculation server resource pool; the main server A is used for being responsible for management of the whole distributed parallel operation, and the multi-mode air quality model is arranged on the main server A; the main server A and the basic server B provide basic calculation power for the multi-mode air quality model, and the basic server B and servers in the calculation power server resource pool provide super calculation power for the multi-mode air quality model; all servers communicate through a private network under the coordination of the main server a.
9. The method for ozone and particulate matter prediction early warning based on the multi-mode air quality model according to claim 1, characterized in that: in the first step, air quality monitoring data of a target area are obtained by setting an air quality monitoring station, and wind speed and wind direction data of the target area are obtained by setting a wind speed and wind direction monitoring device;
a protective frame (1) is arranged on the outer side of the body of the wind speed and direction monitoring device; a cleaning rod (2) is arranged on the outer side of the protective frame (1); the cleaning rod (2) slides along the outer surface of the protective frame (1) under the driving of the driving part; the whole protective frame (1) is arc-shaped, the driving part is a wind-driven rotating part, the rotating part is in rotating fit with a supporting seat (6) below the wind speed and direction monitoring device, two wing pieces (3) are arranged on the rotating part at an angle, the cleaning rod (2) is positioned between the two wing pieces (3), and the wing pieces (3) drive the rotating part to rotate when being windy; the supporting seat (6) is provided with two limiting blocks (8) for limiting the rotation limit positions of the rotating parts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210378084.7A CN114898820A (en) | 2022-04-12 | 2022-04-12 | Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210378084.7A CN114898820A (en) | 2022-04-12 | 2022-04-12 | Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114898820A true CN114898820A (en) | 2022-08-12 |
Family
ID=82718147
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210378084.7A Pending CN114898820A (en) | 2022-04-12 | 2022-04-12 | Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114898820A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116306251A (en) * | 2023-02-09 | 2023-06-23 | 中科三清科技有限公司 | Near-ground ozone concentration inversion method, near-ground ozone concentration inversion device, storage medium and chip |
| CN116596326A (en) * | 2023-04-11 | 2023-08-15 | 常州双炬智能科技有限公司 | Urban environment detection and comprehensive evaluation method based on remote sensing data |
| CN116737428A (en) * | 2023-08-14 | 2023-09-12 | 中科三清科技有限公司 | Air quality mode operation stability checking method and device and electronic equipment |
| CN116741311A (en) * | 2023-08-14 | 2023-09-12 | 中科三清科技有限公司 | Method and device for outputting natural source volatile organic compounds BVOCs emission list |
| CN117031582A (en) * | 2023-06-27 | 2023-11-10 | 华南理工大学 | Ozone hour concentration forecasting method based on recursive space-time learning and simulation monitoring fusion |
| CN117332906A (en) * | 2023-12-01 | 2024-01-02 | 山东大学 | Machine learning-based three-dimensional space-time grid air quality prediction method and system |
| CN118212548A (en) * | 2024-03-21 | 2024-06-18 | 宁波大学 | Cascade depth network-based particulate matter and ozone remote sensing cooperative estimation method |
-
2022
- 2022-04-12 CN CN202210378084.7A patent/CN114898820A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116306251A (en) * | 2023-02-09 | 2023-06-23 | 中科三清科技有限公司 | Near-ground ozone concentration inversion method, near-ground ozone concentration inversion device, storage medium and chip |
| CN116596326A (en) * | 2023-04-11 | 2023-08-15 | 常州双炬智能科技有限公司 | Urban environment detection and comprehensive evaluation method based on remote sensing data |
| CN116596326B (en) * | 2023-04-11 | 2024-04-26 | 泰州城发数字科技有限公司 | Urban environment detection and comprehensive evaluation method based on remote sensing data |
| CN117031582A (en) * | 2023-06-27 | 2023-11-10 | 华南理工大学 | Ozone hour concentration forecasting method based on recursive space-time learning and simulation monitoring fusion |
| CN116737428A (en) * | 2023-08-14 | 2023-09-12 | 中科三清科技有限公司 | Air quality mode operation stability checking method and device and electronic equipment |
| CN116741311A (en) * | 2023-08-14 | 2023-09-12 | 中科三清科技有限公司 | Method and device for outputting natural source volatile organic compounds BVOCs emission list |
| CN116741311B (en) * | 2023-08-14 | 2023-10-20 | 中科三清科技有限公司 | Method and device for outputting natural source volatile organic compounds BVOCs emission list |
| CN116737428B (en) * | 2023-08-14 | 2023-11-21 | 中科三清科技有限公司 | Air quality mode operation stability checking method and device and electronic equipment |
| CN117332906A (en) * | 2023-12-01 | 2024-01-02 | 山东大学 | Machine learning-based three-dimensional space-time grid air quality prediction method and system |
| CN117332906B (en) * | 2023-12-01 | 2024-03-15 | 山东大学 | Machine learning-based three-dimensional space-time grid air quality prediction method and system |
| CN118212548A (en) * | 2024-03-21 | 2024-06-18 | 宁波大学 | Cascade depth network-based particulate matter and ozone remote sensing cooperative estimation method |
| CN118212548B (en) * | 2024-03-21 | 2024-09-13 | 宁波大学 | Cascade depth network-based particulate matter and ozone remote sensing cooperative estimation method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114898820A (en) | Method for predicting and early warning ozone and particulate matters based on multi-mode air quality model | |
| Yang et al. | A review of solar forecasting, its dependence on atmospheric sciences and implications for grid integration: Towards carbon neutrality | |
| Zhang et al. | Real-time air quality forecasting, part II: State of the science, current research needs, and future prospects | |
| Carmichael et al. | The MICS-Asia study: model intercomparison of long-range transport and sulfur deposition in East Asia | |
| Manfren et al. | Paradigm shift in urban energy systems through distributed generation: Methods and models | |
| CN114818391B (en) | Pollution source concentration analysis method based on multi-tracing model | |
| CN109726867B (en) | High-resolution electric power weather forecasting method based on multi-mode set | |
| Cohan et al. | Dependence of ozone sensitivity analysis on grid resolution | |
| CN106651036A (en) | Air quality forecasting system | |
| KR102283487B1 (en) | A Forecasting System of Photovoltaic Generation Based on Machine-learning Using Realtime Satellite Data and Numerical Modeling Data | |
| Cheng et al. | A synoptic climatological approach to assess climatic impact on air quality in south-central Canada. Part I: Historical analysis | |
| KR101882874B1 (en) | Method and apparatus for scheduling power plant considering atmospheric environmental conditions | |
| Niska et al. | Evaluation of an integrated modelling system containing a multi-layer perceptron model and the numerical weather prediction model HIRLAM for the forecasting of urban airborne pollutant concentrations | |
| CN110322283A (en) | Electric energy alternative solution prediction technique and device based on Gauss regression combination prediction model | |
| Chen et al. | Simulating the effects of land urbanization on regional meteorology and air quality in Yangtze River Delta, China | |
| CN117114438A (en) | Building area energy system cold and hot load data driving prediction method with flexibility and interpretability | |
| Stoliarov | Efficient electricity generation forecasting from solar power plants using technology: Integration, benefits and prospects | |
| Petry et al. | Design and results of an AI-based forecasting of air pollutants for smart cities | |
| CN116297062B (en) | PM (particulate matter) oriented to multidimensional space-time factors 2.5 Concentration refinement space-time simulation and quantitative analysis method | |
| Karimian Shamsabadi et al. | Urban buildings configuration and pollutant dispersion of PM 2.5 particulate to enhance air quality | |
| Beddows et al. | Enhancements to the UK Photochemical Trajectory Model for simulation of secondary inorganic aerosol | |
| CN118364725B (en) | Model construction method and system of distributed photovoltaic system | |
| CN118115179B (en) | Method and device for identifying contribution concentration of artificial source and natural source | |
| Stoliarov | BULLETIN of Cherkasy State Technological University | |
| DE ANGELIS | Integrated Assessment Modelling to support decision makers in planning air quality and low carbon win-win policies. A socio-economic and health perspective. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |