CN112163375A - Long-time sequence near-surface ozone inversion method based on neural network - Google Patents
Long-time sequence near-surface ozone inversion method based on neural network Download PDFInfo
- Publication number
- CN112163375A CN112163375A CN202011037528.8A CN202011037528A CN112163375A CN 112163375 A CN112163375 A CN 112163375A CN 202011037528 A CN202011037528 A CN 202011037528A CN 112163375 A CN112163375 A CN 112163375A
- Authority
- CN
- China
- Prior art keywords
- ozone
- neural network
- data
- training
- ground
- 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.)
- Granted
Links
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 title claims abstract description 129
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000003062 neural network model Methods 0.000 claims abstract description 40
- 238000012544 monitoring process Methods 0.000 claims abstract description 27
- 238000012360 testing method Methods 0.000 claims abstract description 24
- 238000004088 simulation Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 230000001419 dependent effect Effects 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 7
- 230000002354 daily effect Effects 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 7
- 230000003203 everyday effect Effects 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000012886 linear function Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 abstract description 2
- 230000001932 seasonal effect Effects 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000007789 gas Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000010355 oscillation Effects 0.000 description 2
- 210000002345 respiratory system Anatomy 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 208000018569 Respiratory Tract disease Diseases 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 230000036737 immune function Effects 0.000 description 1
- 230000007794 irritation Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 230000009257 reactivity Effects 0.000 description 1
- 239000005437 stratosphere Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000005436 troposphere Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/02—Instruments for indicating weather conditions by measuring two or more variables, e.g. humidity, pressure, temperature, cloud cover or wind speed
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
Abstract
The invention discloses a long-time sequence near-ground ozone inversion method based on a neural network, which comprises the following steps of: s1, acquiring the concentration data of the satellite remote sensing monthly scale ozone column according to the satellite remote sensing data; s2, carrying out multi-temporal interpolation calculation on the meteorological data of the ground monitoring meteorological station to obtain meteorological data with spatial distribution characteristics; the meteorological data of the ground monitoring meteorological station comprise temperature, wind speed, air pressure, relative humidity and sunshine duration; s3, establishing a near-ground ozone inversion neural network model according to the concentration data of the satellite remote sensing monthly-scale ozone column and the meteorological data with the space distribution characteristics and training the near-ground ozone inversion neural network model; and S4, carrying out simulation test on the near-surface ozone inversion neural network model. The inversion method is high in precision and simple to operate, can realize accurate near-ground ozone concentration measurement, and can obtain a near-ground ozone concentration data set which can accurately reflect seasonal changes, annual changes and spatial distribution.
Description
Technical Field
The invention relates to the field of remote sensing inversion, in particular to a long-time sequence near-ground ozone inversion method based on a neural network.
Background
Ozone is an important trace gas, about 90% of the atmospheric ozone is present in the stratosphere, and only 10% is present in the troposphere, which can block ultraviolet rays and protect the earth's biosphere, while near-surface ozone is a pollutant. Near-ground ozone is an important greenhouse gas, and the increase of the concentration of ozone directly causes the temperature of the ground surface to increase; meanwhile, as a measuring gas with extremely high reactivity and extremely strong oxidability, ozone can generate certain damage to respiratory systems and plants of people and animals, and high-concentration ozone has very strong irritation to eyes and respiratory tracts, can cause respiratory tract diseases and destroy the immune function of human bodies; in addition, ozone can enter the crop body through the air holes to damage the crop tissues.
In recent years, with the rapid development of Chinese economy and the rapid advance of urbanization process, the method uses O3The typical photochemical pollution tends to be increased year by year, and the ozone concentration of 337 and more cities in 2019 in the country is increased by 6.5 percent at the same ratio, which is the most important factor influencing the air quality of the cities. "environmental air space" implemented in 2013 in ChinaGas quality standard (GB 3095-2012), ozone is newly added as a daily routine monitoring project, which creates favorable conditions for developing ozone pollution research. The existing ozone acquisition mainly comprises two means of site monitoring and remote sensing monitoring, wherein the site monitoring mainly comprises the steps of laying a large number of ground monitoring sites, actually measuring the air pollution data close to the ground, obtaining the ozone concentration continuously monitored by a fixed point hour by hour, and having the characteristics of high time resolution and continuous and accurate data. However, due to the limitation of factors such as site construction cost and regions, the number of ground environment monitoring sites is limited, the spatial representativeness is poor, the sites in the center of a common city are distributed more densely, and the sites in the peripheral regions of the cities such as suburbs, rural areas and mountainous areas are distributed sparsely, so that the difficulty is created for researching the spatial distribution condition of the ozone concentration on the ground of the city. With the development of remote sensing technology, satellite remote sensing provides a technical means for acquiring the global distribution of ozone and the change of the ozone along with time, and provides a richer data source for ozone research. However, the total amount data of the ozone column provided by satellite remote sensing, the ozone profile data and the near-ground ozone concentration data have a certain difference, and the tropospheric ozone concentration cannot completely and accurately reflect the near-ground ozone change, for example, research shows that the tropospheric ozone concentration is relatively stable in recent years, and the near-ground ozone concentration shows a gradually increasing trend. Therefore, in order to research the time-space change characteristics of near-ground ozone, it is necessary to fuse the ozone data of a ground monitoring station and the satellite remote sensing ozone data to obtain a long-time sequence near-ground ozone concentration data set with good time and space continuity. Liu xian (2016) et al used the OMI satellite data to multiply near-ground ozone using an optimized estimation method based on climate prior information and verified that the accuracy R was 0.73 using ground site data.
Disclosure of Invention
The invention provides a long-time sequence near-surface ozone inversion method based on a neural network, aiming at the problems that the station observation data is lack of space representativeness and the difference exists between the remote sensing observation data and the near-surface ozone data.
The invention provides a long-time sequence near-ground ozone inversion method based on a neural network, which comprises the following steps of:
s1, acquiring the concentration data of the satellite remote sensing monthly scale ozone column according to the satellite remote sensing data;
s2, carrying out multi-temporal interpolation calculation on the meteorological data of the ground monitoring meteorological station to obtain meteorological data with spatial distribution characteristics; the meteorological data of the ground monitoring meteorological station comprise temperature, wind speed, air pressure, relative humidity and sunshine duration;
s3, establishing a near-ground ozone inversion neural network model according to the concentration data of the satellite remote sensing monthly-scale ozone column and the meteorological data with the space distribution characteristics and training the near-ground ozone inversion neural network model;
and S4, carrying out simulation test on the near-surface ozone inversion neural network model.
Optionally, the step S1 further includes the following steps:
s11, eliminating cloud cluster coverage and valueless areas under the rainy weather condition in the daily satellite remote sensing ozone column concentration data;
s12, obtaining the monthly mean value of the concentration of the ozone column according to the concentration data of the ozone column of the satellite remote sensing every day;
s13, extracting the ozone column concentration monthly mean value in the 3 x 3 pixels at the ground monitoring weather station, namely selecting the pixels P (a, b) where the longitude and latitude of the ground monitoring weather station are located, wherein the a and the b are pixel row numbers, traversing the 3 x 3 pixels around the pixels, and obtaining the ozone column concentration monthly mean value in the pixel range.
Optionally, the step S2 further includes the following steps:
s21, eliminating invalid values in the meteorological data of the daily ground monitoring meteorological site, and calculating the monthly change of the meteorological data based on the invalid values;
and S22, obtaining the meteorological data with spatial distribution characteristics from the meteorological data of the ground monitoring meteorological station according to a Krigin interpolation method.
Optionally, the step S3 further includes the following steps:
s31, setting a training data set and a verification data set, wherein 80% of samples are used for training the model, and 20% of samples are used for checking the model precision;
s32, building and training a near-surface ozone inversion neural network model, namely building and training the near-surface ozone inversion neural network model by utilizing a plurality of dependent variables and independent variables in the training data set prepared in the step S31; the dependent variable comprises the concentration, the temperature, the wind speed, the air pressure, the relative humidity and the sunshine hours of a satellite remote sensing ozone column every day; the dependent variable includes near-surface ozone column concentration.
Optionally, the step S32 includes the following steps:
s321, setting a hidden layer;
s322, configuring neural network parameters including iteration times, minimum training rate, allowable error and dynamic parameters;
s323: and establishing and training a neural network model, wherein each input layer uses a hyperbolic tangent function tansig as a transmission function, each output layer uses a linear function purelin as a transmission function, and an elastic gradient descent function trainrp is used for training.
Optionally, in step S321, the number of hidden layer nodes is determined according to an empirical formula and the corresponding modeling precision:
in the formula, h is the number of nodes of the hidden layer, m and n are the numbers of nodes of the input layer and the output layer respectively, and a is an adjusting constant between 1 and 10.
Optionally, in step S4, performing a simulation test on the near-surface ozone inversion neural network model by using the verification data set.
Optionally, the method further comprises a step S5 of evaluating the near-surface ozone inversion neural network model based on simulation test results.
Optionally, in the step S5, the confidence level of the model is evaluated by using the decision coefficient R2, the mean square error MSE, the root mean square error RMSE, and the mean absolute error MAE.
The technical scheme of the invention has the beneficial technical effects that: the inversion method is high in precision and simple to operate, can realize accurate near-ground ozone concentration measurement, and can obtain a near-ground ozone concentration data set which can accurately reflect seasonal changes, annual changes and spatial distribution.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a neural network-based long time series near-surface ozone inversion method of the present invention.
FIG. 2 is a schematic structural diagram of a near-surface ozone inversion neural network model of the present invention.
FIG. 3 is a schematic diagram of a near-surface ozone inversion neural network model training process according to the present invention.
FIG. 4 is a graph of near-surface ozone inversion neural network model modeling accuracy and testing accuracy.
FIG. 5 is a diagram of the near-surface ozone inversion neural network model simulation test error of the present invention.
FIG. 6 is a near-surface ozone inversion neural network model inversion 2015-2019 Kyojin Tang near-surface ozone distribution graph adopting the near-surface ozone inversion method.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of systems consistent with certain aspects of the invention, as detailed in the appended claims.
As shown in FIG. 1, the invention provides a long time series near-surface ozone inversion method based on a neural network, which comprises the following steps: s1: acquiring satellite remote sensing monthly scale ozone column concentration data according to the satellite remote sensing data; s2: performing multi-temporal interpolation calculation on meteorological data of the ground monitoring meteorological station to obtain meteorological data with spatial distribution characteristics; the meteorological data of the ground monitoring meteorological station comprise temperature, wind speed, air pressure, relative humidity and sunshine duration; s3: establishing a near-ground ozone inversion neural network model according to the satellite remote sensing monthly ozone column concentration data and the meteorological data with the spatial distribution characteristics and training the near-ground ozone inversion neural network model; s4: and carrying out simulation test on the near-surface ozone inversion neural network model.
According to an embodiment of the present invention, step S1 further includes the steps of: s11: eliminating cloud cluster coverage in daily satellite remote sensing ozone column concentration data and a value-free area under the rainy weather condition; s12: obtaining the monthly mean value of the concentration of the ozone column according to the concentration data of the ozone column of the satellite remote sensing every day; s13: extracting the monthly mean value of the concentration of the ozone column in the 3 x 3 pixels at the ground monitoring meteorological station, namely selecting the pixels P (a, b) where the longitude and latitude of the ground monitoring meteorological station are located, wherein the a and the b are pixel row numbers, traversing the 3 x 3 pixels around the pixels, and acquiring the monthly mean value of the concentration of the ozone column in the pixel range.
According to an embodiment of the present invention, step S2 further includes the steps of: s21: eliminating invalid values in the meteorological data of the daily ground monitoring meteorological station, and calculating the monthly change of the meteorological data based on the invalid values; s22: and acquiring meteorological data with spatial distribution characteristics from the meteorological data of the ground monitoring meteorological station according to a Krigin interpolation method.
In step S22, the following formula is used for calculation:
E[Y(s)]=μ (1)
var[Y(s)-Y(s+h)]=2[C(0)-C(h)]=2γ(|h|) (2)
wherein, { Y(s)1),......,Y(sn) Is n samples s1,......,snCorresponding value of E [ Y(s) ]]For the mathematical expectation of Y(s), | h | is the distance between points, γ is the variation function, and C is the covariance function.
In the formula (I), the compound is shown in the specification,is Y at an unknown point s0Estimate of (C)(s)i,sj) Is a point siAnd sjWith an unbiased estimation condition of the sum of the covariance functionsThis gives:
the above equation set is a common kriging system, and includes n +1 equations to solve n weight coefficients. The kriging weight represented by a matrix after solving is as follows:
wherein C is a covariance matrix, C0The column vector composed of covariance between unknown points and samples is 1, and the column vector composed of n 1. The unbiased estimation of ordinary kriging can be obtained by substituting all weights into the previous formula
From the above equation, the mathematical expectation of the random field
μ=(1TC-11)-1(1TC-lY) (9)。
According to an embodiment of the present invention, step S3 further includes the steps of: s31, setting a training data set and a verification data set, wherein 80% of samples are used for training the model, and 20% of samples are used for checking the model precision; s32, building and training a near-surface ozone inversion neural network model, namely building and training the near-surface ozone inversion neural network model by utilizing a plurality of dependent variables and independent variables in the training data set prepared in the step S31; the dependent variable comprises the concentration, the temperature, the wind speed, the air pressure, the relative humidity and the sunshine hours of a satellite remote sensing ozone column every day; the dependent variable includes near-surface ozone column concentration. .
According to an embodiment of the present invention, in step S32, 240 groups of data of 5 sites per month in kyjin tang area 2015-2019 are randomly selected as a training data set for training, and 60 groups of data are selected as a verification data set for verification.
FIG. 2 is a schematic structural diagram of a near-surface ozone inversion neural network model of the present invention. The near-surface ozone inversion neural network model comprises an input layer, a hidden layer and an output layer. According to an embodiment of the present invention, step S32 includes step S321, setting the hidden layer. The number of hidden layer nodes has a great influence on the performance of the neural network model, but a scientific method for determining the number of the hidden layer nodes does not exist at present. Generally, it is thought that the network error can be reduced by adding the hidden layer, but the complexity of the network is also increased, so that the fitting time of the network is increased, the generalization capability of the network is reduced, and the overfitting phenomenon is easy to occur. If the number of hidden layer nodes is too small, the network cannot establish a complex judgment boundary and cannot train a proper network model. Therefore, the number of hidden layer nodes should be selected as few as possible on the premise of meeting the precision. The number of hidden layer nodes can be determined according to an empirical formula and corresponding modeling precision:
in the formula, h is the number of nodes of the hidden layer, m and n are the numbers of nodes of the input layer and the output layer respectively, and a is an adjusting constant between 1 and 10. According to a preferred implementation of the embodiment of the invention, the number of hidden layer nodes is set to 13 based on the above experience.
According to an embodiment of the present invention, step S32 further includes step S322 of configuring neural network parameters, including iteration number, minimum training rate, allowable error, and dynamic parameters. The iteration times are the maximum iteration times allowed by the neural network when the iteration result is not converged under the configured parameter condition, and according to a preferred implementation manner of the embodiment of the invention, the iteration times are set to be 1000; in the neural network model, the larger the training rate is, the larger the weight change is, and the faster the convergence is, but too fast the convergence rate may cause the oscillation of the system, resulting in the failure of convergence. Therefore, on the premise of no oscillation, the training rate as large as possible is selected, and according to a preferred implementation manner of the embodiment of the invention, the minimum training rate is set to be 0.9; when the error of the results of the two iterations is smaller than the allowable error, the system ends the iteration and outputs the results. According to a preferred implementation of the embodiment of the invention, the allowed error is set to 0.0001; according to a preferred implementation of the embodiment of the invention, the dynamic parameter is empirically chosen and set to 0.8.
According to an embodiment of the present invention, step S32 further includes step S323 of building and training a neural network model. Each input layer uses a hyperbolic tangent function tansig as a transfer function, each output layer uses a linear function purelin as a transfer function, and each output layer is trained by an elastic gradient descent function trainrp.
FIG. 3 is a schematic diagram of a near-surface ozone inversion neural network model training process according to the present invention. As shown in fig. 3, in the training process of the near-surface ozone inversion neural network model in step S3, the best accuracy result is obtained at cycle 36 of 0.0253, and the training is finished after cycle 42, so that the model building accuracy is high.
According to an embodiment of the present invention, in step S4, a simulation test is performed on the near-surface ozone inversion neural network model using the validation data set. And bringing the test data set into a network established by the training data set, outputting a result corresponding to the test data set, comparing the output result with a true value, and judging the model precision. The correlation coefficient R of the predicted near-surface ozone value and the original near-surface ozone value is 0.9476, and when N is 240, the correlation is significant when the confidence interval of 95% is exceeded statistically.
Further, wherein the correlation coefficient R is calculated according to the following formula:
in the formula, yiThe near-surface ozone values output by the training model for the training data set,to the original near-surface ozone values in the training dataset,to the original near-surface ozone value mean in the training dataset,is the original near-surface ozone value mean in the training data set, and n is the total training data.
FIG. 4 is a graph of near-surface ozone inversion neural network model modeling accuracy and testing accuracy. As shown in fig. 4, the left graph shows the correlation between the real value and the error value in the model training process in step S3, i.e. the training precision of the model is 0.9491. After the model is built, in the simulation test process of step S4, the error between the true value and the predicted value, i.e. the prediction accuracy of the model is 0.9423. The training precision and the prediction precision of the neural network model are proved to be higher, and the requirement of the ozone inversion precision is completely met. FIG. 5 is a comparison graph of the predicted value and the true value of the model and the corresponding error value in the simulation test process of the model in step S4. FIG. 6 shows the mean value of ozone concentration and the spatial distribution of Jingjin Tang near the ground in 2015-2019 based on the final inversion of the neural network. According to an embodiment of the invention, the method further comprises the step S5 of evaluating the near-surface ozone inversion neural network model based on simulation test results.
According to a preferred implementation of an embodiment of the invention, the confidence of the model is evaluated using the decision coefficient R2, the mean square error MSE, the root mean square error RMSE and the mean absolute error MAE.
The evaluation reliability of the coefficient R2 is determined by the following equation (12)
In the formula, yiThe near-surface ozone values output by the training model for the test data set,to test the original near-surface ozone values in the data set,is the mean of the original near-surface ozone values in the test data set, and n is the total amount of the test data. The decision coefficient R2 is 0.8979, and the model has high precision.
The evaluation reliability of the mean square error MSE, the mean square error RMSE, and the mean absolute error MAE is expressed by the following equations (13), (14), and (15).
In the above formulae, yiThe near-surface ozone values output by the training model for the test data set,is the original near-surface ozone value in the test data set, and N is the total amount of test data. The average absolute error MAE can better reflect the actual situation of the error of the predicted value, the value is 0.1731, the mean square error is 115.3849, and the root mean square error is 10.7417, so that the accuracy requirement is met.
The features and benefits of the present invention are illustrated by reference to the examples. Accordingly, the invention is expressly not limited to these exemplary embodiments illustrating some possible non-limiting combination of features which may be present alone or in other combinations of features.
The above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (9)
1. A long-time sequence near-ground ozone inversion method based on a neural network is characterized by comprising the following steps:
s1, acquiring the concentration data of the satellite remote sensing monthly scale ozone column according to the satellite remote sensing data;
s2, carrying out multi-temporal interpolation calculation on the meteorological data of the ground monitoring meteorological station to obtain meteorological data with spatial distribution characteristics; the meteorological data of the ground monitoring meteorological station comprise temperature, wind speed, air pressure, relative humidity and sunshine duration;
s3, establishing a near-ground ozone inversion neural network model according to the concentration data of the satellite remote sensing monthly-scale ozone column and the meteorological data with the space distribution characteristics and training the near-ground ozone inversion neural network model;
and S4, carrying out simulation test on the near-surface ozone inversion neural network model.
2. The neural network-based long-time-series near-surface ozone inversion method of claim 1, wherein the step S1 further comprises the steps of:
s11, eliminating cloud cluster coverage and valueless areas under the rainy weather condition in the daily satellite remote sensing ozone column concentration data;
s12, obtaining the monthly mean value of the concentration of the ozone column according to the concentration data of the ozone column of the satellite remote sensing every day;
s13, extracting the ozone column concentration monthly mean value in the 3 x 3 pixels at the ground monitoring weather station, namely selecting the pixels P (a, b) where the longitude and latitude of the ground monitoring weather station are located, wherein the a and the b are pixel row numbers, traversing the 3 x 3 pixels around the pixels, and obtaining the ozone column concentration monthly mean value in the pixel range.
3. The neural network-based long-time-series near-surface ozone inversion method of claim 1, wherein the step S2 further comprises the steps of:
s21, eliminating invalid values in the meteorological data of the daily ground monitoring meteorological site, and calculating the monthly change of the meteorological data based on the invalid values;
and S22, obtaining the meteorological data with spatial distribution characteristics from the meteorological data of the ground monitoring meteorological station according to a Krigin interpolation method.
4. The neural network-based long-time-series near-surface ozone inversion method of claim 1, wherein the step S3 further comprises the steps of:
s31, setting a training data set and a verification data set, wherein 80% of samples are used for training the model, and 20% of samples are used for checking the model precision;
s32, building and training a near-surface ozone inversion neural network model, namely building and training the near-surface ozone inversion neural network model by utilizing a plurality of dependent variables and independent variables in the training data set prepared in the step S31; the dependent variable comprises the concentration, the temperature, the wind speed, the air pressure, the relative humidity and the sunshine hours of a satellite remote sensing ozone column every day; the dependent variable includes near-surface ozone column concentration.
5. The neural network-based long-time-series near-surface ozone inversion method of claim 4, wherein the step S32 comprises the steps of:
s321, setting a hidden layer
S322, configuring neural network parameters including iteration times, minimum training rate, allowable error and dynamic parameters;
s323: and establishing and training a neural network model, wherein each input layer uses a hyperbolic tangent function tansig as a transmission function, each output layer uses a linear function purelin as a transmission function, and an elastic gradient descent function trainrp is used for training.
6. The neural network-based long-time series near-surface ozone inversion method of claim 4, wherein in step S321, the number of hidden layer nodes is determined according to an empirical formula and corresponding modeling accuracy:
in the formula, h is the number of nodes of the hidden layer, m and n are the numbers of nodes of the input layer and the output layer respectively, and a is an adjusting constant between 1 and 10.
7. The neural network-based long-time-series near-surface ozone inversion method of claim 4, wherein in the step S4, the near-surface ozone inversion neural network model is subjected to simulation testing by using the verification data set.
8. The long-time series near-surface ozone inversion method based on the neural network as claimed in claim 1, further comprising the step S5 of evaluating the near-surface ozone inversion neural network model based on simulation test results.
9. The neural network-based long-time-series near-surface ozone inversion method of claim 7, characterized in that: in step S5, the confidence level of the model is evaluated using the decision coefficient R2, the mean square error MSE, the root mean square error RMSE, and the mean absolute error MAE.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011037528.8A CN112163375B (en) | 2020-09-28 | Long-time sequence near-ground ozone inversion method based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011037528.8A CN112163375B (en) | 2020-09-28 | Long-time sequence near-ground ozone inversion method based on neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112163375A true CN112163375A (en) | 2021-01-01 |
CN112163375B CN112163375B (en) | 2024-05-10 |
Family
ID=
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112052627A (en) * | 2020-08-21 | 2020-12-08 | 海南星瞰信息咨询中心(有限合伙) | Method, device, medium and equipment for estimating near-surface ozone space distribution |
CN113189014A (en) * | 2021-04-14 | 2021-07-30 | 西安交通大学 | Ozone concentration estimation method fusing satellite remote sensing and ground monitoring data |
CN113486295A (en) * | 2021-07-30 | 2021-10-08 | 生态环境部华南环境科学研究所 | Fourier series-based total ozone change prediction method |
CN113514606A (en) * | 2021-04-25 | 2021-10-19 | 中科三清科技有限公司 | Method and device for forecasting ozone concentration by using ozone potential index |
CN113657023A (en) * | 2021-07-19 | 2021-11-16 | 生态环境部卫星环境应用中心 | Near-surface ozone concentration inversion method based on combination of machine learning and deep learning |
CN115631218B (en) * | 2022-12-21 | 2023-04-11 | 广东中浦科技有限公司 | Ozone transmission change analysis method based on common kriging interpolation |
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 |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310386A (en) * | 2020-02-13 | 2020-06-19 | 北京中科锐景科技有限公司 | Near-surface ozone concentration estimation method |
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310386A (en) * | 2020-02-13 | 2020-06-19 | 北京中科锐景科技有限公司 | Near-surface ozone concentration estimation method |
Non-Patent Citations (1)
Title |
---|
丁愫;陈报章;王瑾;陈龙;张晨雷;孙少波;黄丛吾;: "基于决策树的统计预报模型在臭氧浓度时空分布预测中的应用研究", 环境科学学报, no. 08, pages 277 - 290 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112052627A (en) * | 2020-08-21 | 2020-12-08 | 海南星瞰信息咨询中心(有限合伙) | Method, device, medium and equipment for estimating near-surface ozone space distribution |
CN113189014A (en) * | 2021-04-14 | 2021-07-30 | 西安交通大学 | Ozone concentration estimation method fusing satellite remote sensing and ground monitoring data |
CN113514606A (en) * | 2021-04-25 | 2021-10-19 | 中科三清科技有限公司 | Method and device for forecasting ozone concentration by using ozone potential index |
CN113657023A (en) * | 2021-07-19 | 2021-11-16 | 生态环境部卫星环境应用中心 | Near-surface ozone concentration inversion method based on combination of machine learning and deep learning |
CN113486295A (en) * | 2021-07-30 | 2021-10-08 | 生态环境部华南环境科学研究所 | Fourier series-based total ozone change prediction method |
CN113486295B (en) * | 2021-07-30 | 2023-05-30 | 生态环境部华南环境科学研究所 | Fourier series-based ozone total amount change prediction method |
CN115631218B (en) * | 2022-12-21 | 2023-04-11 | 广东中浦科技有限公司 | Ozone transmission change analysis method based on common kriging interpolation |
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 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109580003B (en) | Method for estimating near-ground atmospheric temperature by thermal infrared data of stationary meteorological satellite | |
Elminir et al. | Estimation of solar radiation components incident on Helwan site using neural networks | |
CN111078678A (en) | Satellite precipitation data correction method based on multi-source information fusion and scale reduction | |
CN110232471B (en) | Rainfall sensor network node layout optimization method and device | |
CN112785024A (en) | Runoff calculation and prediction method based on watershed hydrological model | |
CN112287294B (en) | Space-time bidirectional soil water content interpolation method based on deep learning | |
Ahn et al. | The reproducibility of surface air temperature over South Korea using dynamical downscaling and statistical correction | |
CN112016696B (en) | PM integrating satellite observation and ground observation 1 Concentration inversion method and system | |
CN110595968B (en) | PM2.5 concentration estimation method based on geostationary orbit satellite | |
CN113344149B (en) | PM2.5 hourly prediction method based on neural network | |
CN108983324A (en) | A kind of temperature forecast method and system based on Kalman filtering | |
CN110968926B (en) | Method for predicting atmospheric parameters based on improved background error covariance matrix | |
Khesali et al. | A method in near-surface estimation of air temperature (NEAT) in times following the satellite passing time using MODIS images | |
CN115544706A (en) | Wavelet and XGboost model integrated atmospheric fine particle concentration estimation method | |
CN111597692B (en) | Method, system, electronic equipment and storage medium for estimating surface net radiation | |
TWI717796B (en) | System, method and storage medium for estimating the amount of sunshine in geographic location by artificial intelligence | |
CN112560633A (en) | Plant key phenological period time point prediction method and system based on deep learning | |
Zhang et al. | The CA model based on data assimilation | |
CN117251989A (en) | Meteorological site carbon water flux information determination method, system and electronic equipment | |
CN109948175B (en) | Satellite remote sensing albedo missing value inversion method based on meteorological data | |
Hema et al. | Reconstructing missing hourly real-time precipitation data using a novel intermittent sliding window period technique for automatic weather station data | |
CN115859789A (en) | Method for improving inversion accuracy of polar atmosphere temperature profile | |
CN112163375B (en) | Long-time sequence near-ground ozone inversion method based on neural network | |
CN112163375A (en) | Long-time sequence near-surface ozone inversion method based on neural network | |
CN115730524A (en) | Machine learning-based numerical simulation virtual anemometry error correction method |
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 | ||
GR01 | Patent grant |