KR20220078847A - Method for calculating the concentration of a plurality of gases and a device for calculating the concentration of a plurality of gases - Google Patents

Abstract

The method for calculating the concentration of a plurality of gases according to an embodiment comprises the steps of preparing reference data representing predetermined characteristics of a plurality of gases, (b) driving a radiation transfer model based on the reference data of the plurality of gases to generate a model radiation amount obtaining, (c) obtaining the observed radiation amount through a Geostationary Environmental Monitoring Spectrometer (GEMS), (d) calculating a difference value between the model radiation amount and the observed radiation amount, and (e) the difference value is It may include repeatedly performing (b) to (d) until it is less than a predetermined threshold.

Description

Method for calculating the concentration of a plurality of gases and a device for calculating the concentration of a plurality of gases

The present invention relates to a method for calculating the concentration of a plurality of gases and an apparatus for calculating the concentration of a plurality of gases, and more particularly, in a state where the model radiation through a radiation transfer model of a plurality of gases and the observed radiation calculated through GEMS are almost similar. The present invention relates to a method for calculating concentrations of a plurality of gases and a device for calculating concentrations of a plurality of gases for calculating concentrations of a plurality of gases.

Recently, the frequency of using spectrometers in satellite, terrestrial and aerial observations to calculate atmospheric host information on gaseous substances and aerosols in the atmosphere is increasing.

In particular, as environmental problems such as air pollution have emerged as a global issue, the activity of geostationary orbit satellites equipped with environmental payloads is becoming increasingly important.

In the case of the prior art, in order to know the concentration of a specific trace gas, it is necessary to assume the concentration of other trace gases in the same wavelength band. For example, referring to FIG. 4 , in order to calculate the concentration of O ₃ in the G1 section, it is necessary to assume the concentration of SO ₂ and NO ₂ , and the concentration of O ₃ in order to calculate the concentration of NO ₂ in the G2 section. An assumption is needed for

That is, when the concentration of the other gas is not relatively small within the interval for calculating the concentration of the gas having the largest absorption signal, it is necessary to know the concentration of the other gas in advance. At this time, if the assumption of the concentration of another gas is made wrong, an error occurs in the calculated concentration of the gas having the largest absorption signal.

In the case of the present invention, it was derived to solve the above-mentioned problems, and by simultaneously calculating the concentrations of a plurality of gases rather than calculating the concentration of a specific type of gas, interference by other gases is minimized to minimize the above false assumptions. The purpose of this is to reduce the error rate caused by

A method for calculating concentrations of a plurality of gases according to an embodiment includes the steps of: (a) preparing reference data indicating predetermined characteristics of the plurality of gases; (b) obtaining a model radiation amount by driving a radiation transfer model based on the reference data of the plurality of gases; (c) acquiring the observed radiation through a Geostationary Environmental Monitoring Spectrometer (GEMS); (d) calculating a difference value between the model radiation amount and the observed radiation amount; and (e) repeating steps (b) to (d) until the difference value is less than a predetermined threshold.

allocating an input value of a selected part of the reference data of the plurality of gases to a predetermined state variable, wherein (b) is performed based on the input value assigned to the predetermined state variable The radiation transfer model may be driven.

Step (e) may be performed by re-assigning another input value of the selected part of the reference data to the predetermined state variable.

The reference data may include at least one of a surface reflectivity of the plurality of gases, an altitude from the surface of the earth, a concentration of each gas, and a radiation amount absorption degree for each gas.

The copy transfer model may be a neural network model composed of an input node, a hidden node, and an output node, wherein (b) inputs the input value assigned to the predetermined state variable to the input node and receives the input from the output node. Inference on the radiation transfer model may be performed by outputting the model radiation amount.

The plurality of gases are O ₃ , CO _{2 ,} aerosol, SO ₂ , NO ₂ , It may be at least two selected from the group consisting of CH ₂ O.

Concentration calculating apparatus of a plurality of gases according to an embodiment includes a processor; and a memory including at least one instruction executable by the processor, wherein the processor prepares reference data representing predetermined characteristics of a plurality of gases, and based on the reference data of the plurality of gases The radiation transfer model is driven to obtain the model radiation, the observed radiation amount is obtained through a Geostationary Environmental Monitoring Spectrometer (GEMS), and a difference value between the model radiation amount and the observed radiation amount is calculated, and the difference value is less than a predetermined threshold. The steps (b) to (d) may be repeated until the

In the present invention, by simultaneously calculating the concentrations of a plurality of gases rather than calculating the concentrations of one specific type of gas, interference by other gases is minimized, thereby reducing the error occurrence rate.

By using the high resolution of GEMS, it is possible to calculate the concentration of a plurality of gases corresponding to a wide range of wavelengths, and thus the precision can be further improved.

1 is a block diagram of an apparatus for calculating concentrations of a plurality of gases according to an embodiment.
2 is a flowchart illustrating a method of calculating concentrations of a plurality of gases according to an embodiment.
FIG. 3 is a conceptual diagram for explaining a method of calculating concentrations of a plurality of gases according to an embodiment together with FIG. 2 .
4 is a diagram referenced to explain the problem of the method for calculating the concentration of a single gas according to the prior art.
5 is a diagram for explaining the internal configuration of a model when the radiation transfer model according to the embodiment is implemented as a neural network model.
6 is a diagram referenced to explain neural network learning for automatically selecting a plurality of gases according to an embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] Reference is made to the accompanying drawings, which show by way of illustration specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein with respect to one embodiment may be implemented in other embodiments without departing from the spirit and scope of the invention. In addition, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the detailed description set forth below is not intended to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all scope equivalents as those claimed. Like reference numerals in the drawings refer to the same or similar functions throughout the various aspects.

Before going into detail, the present invention may obtain a product from a Geostationary Environmental Monitoring Spectrometer (GEMS) to be mounted on a fixed multi-purpose satellite. The GEMS instrument may be an imaging spectrometer designed to measure backscattered radiation from the atmosphere and reflected radiation from the Earth's surface.

After a series of calibration assembly and scan mechanisms, the measured GEMS signal can be focused on the slit and dispersed into the spectrometer. The spectrometer may use a two-dimensional charge-coupled device (CCD). The signal from the spectrometer can be imaged at the focal plane, passed through the instrument control electronics, and then sent to the host spacecraft for transmission to the ground.

When the amount of trace gas in the atmosphere is too small or when several gases are mixed, a detailed and accurate analysis and calculation technique such as the present invention is required.

In particular, in the case of the present invention, the concentration of a plurality of gases corresponding to a wide range of wavelengths can be calculated together using the high resolution of the GEMS, so that the precision can be further improved.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings.

As shown in FIG. 1 , the apparatus 1 for calculating concentrations of a plurality of gases according to the embodiment may include a processor 10 and a memory 20 .

The processor 10 controls the overall system of the apparatus 1 for calculating the concentration of a plurality of gases.

The Radiative Transfer Model (11) is a model that physically calculates a series of processes in which light from the sun passes through the earth's atmosphere, is reflected from the earth's surface, and passes through the earth's atmosphere again. In the present invention, the concentration of gas A forward model that formulates the relationship between light and radiation can be used.

The radiation transfer model 11 may be manufactured in the form of a hardware chip including a software module and mounted on the apparatus 1 for calculating concentrations of a plurality of gases. For example, it may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or as part of an existing general-purpose processor (such as a CPU or application processor) or a graphics-only processor (such as a GPU) for multiple aircraft. It may be mounted on the concentration calculating device 1 of

As shown in FIG. 5 , the radiative transfer model 11 may be a neural network model composed of an input node, a hidden node, and an output node.

The copy propagation model 11 may include a number of associative parameters w between an input node and a hidden node and between a hidden node and an output node.

Each input node of the input layer may be fully connected to each output node of the output layer, or may be incompletely connected to some output nodes of the output layer.

The input node serves to transfer the input value assigned to a predetermined state variable to the hidden node. Then, the actual reasoning operation is performed in the hidden node. After generating the output data in the hidden node, the output node receives the output data and recalculates it to output the model copy amount. When calculation is performed at the hidden node and the output node, the input data input to the node is multiplied by a predetermined related parameter (or weight, w) and received and the calculation is performed. Then, after summing (weighted sum) all the calculation results performed at each node, it passes through a preset activation function to output the model copy amount.

Hidden nodes and output nodes have activation functions. Activation functions are a step function, a sign function, a linear function, a logistic sigmoid function, a hyper tangent function, a ReLU function, and a softmax ( softmax) function. An activation function may be appropriately determined by a person skilled in the art according to a learning method of a neural network. Neural networks learn machine learning by repeatedly updating (or modifying) the associated parameter (w) to an appropriate value.

The memory 20 stores at least one instruction executable by the processor 10 and a processing result by the processor 10 , and the processor 10 reads the instruction from the memory 20 and executes a program.

2 is a flowchart of a method for calculating concentrations of a plurality of gases according to an embodiment.

As shown in FIG. 2 , the processor 10 may prepare reference data indicating predetermined characteristics of a plurality of gases ( s210 ). According to the embodiment, the plurality of gases may be at least two selected from the group consisting of O ₃ , CO _{2 ,} aerosol, SO ₂ , NO _{2 ,} CH ₂ O, etc. can perform each process.

According to another embodiment, the processor 10 performs unsupervised learning through deep learning based on a neural network, such as O ₃ , CO _{2 ,} aerosol, SO ₂ , NO _{2 ,} CH ₂ O, etc. Gases composed of trace gases can be automatically classified into a predetermined number of clusters. In a state in which target data desired to be calculated by the neural network for input data is not determined, output data classified for similar input data can be obtained through unsupervised learning. In this case, the number of clusters for classification may be arbitrarily set by the user, but the number is not particularly limited.

Specifically, in the neural network of FIG. 6 , trace gas information may be input through an input node of an input layer, and a probability value belonging to each cluster may be output through an output node of an output layer. Then, it is classified into a specific cluster by referring to the probability value. For example, O ₃ and CO ₂ may be classified as a first cluster, SO ₂ , and NO ₂ may be classified as a second cluster, and aerosol and CH ₂ O may be classified as a third cluster according to specific criteria. For reference, the specific criterion may be the molecular weight of the gas, the degree of influence of the gas on contamination, and the like. The processor 10 may perform each step of FIG. 2 by defining each gas included in each group as a plurality of gases of the present invention.

The wavelength band may be set to be a wide wavelength band, preferably the entire wavelength band, enough to determine the absorption amounts of at least two preselected gases. For reference, in the present invention, a plurality of gases means different types of gases.

Reference data representing predetermined characteristics of a plurality of gases according to the embodiment include the surface reflectance of the plurality of gases, distribution information of the plurality of gases by altitude, from the surface of the earth, the concentration of each gas, the intensity of solar radiation, and the point observed by the satellite. It may include at least one of the angle of the satellite with respect to the sun, air pressure, temperature, and humidity information, and the amount of radiation absorbed per molecule of gas at a specific temperature and pressure condition.

In the case of the present invention, the reference data may be data obtained from an external device or using data previously calculated through a Geostationary Environmental Monitoring Spectrometer (GEMS).

In the present invention, the concentration of a plurality of gases can be simultaneously calculated using the observed radiation of a plurality of gases confirmed from all wavelength bands observed by GEMS rather than a single gas, thereby reducing errors caused by interference of other gases. It is possible to obtain the output that is minimized. It is the core of the present invention to minimize the error by considering the possible interference as much as possible in case of interference that may occur no matter how the wavelength band is selected.

The processor 10 may allocate input values of some selected from among reference data of a plurality of gases to a predetermined state vector (s220). For example, for iterative operation of the radiative transfer model, the input value (eg > 20 ppm) of some important data (eg > gas concentration) among the reference data is assigned to a state variable, and then the input value is changed to another value. Iterative operation can be performed through the process of reassigning to a state variable.

However, the present invention may be equally/similarly applied to a case in which not only input values corresponding to one of the reference data but also input values corresponding to at least two or more are allocated to state variables. For example, not only the concentration of the gas but also the amount of radiation absorbed per molecule of the gas under specific temperature and atmospheric pressure conditions may be assigned to the state variables to perform the above process.

The processor 10 drives the radiation transfer model 11 based on an input value assigned to a predetermined state variable to obtain a model radiation amount, and the observed radiation amount calculated through a Geostationary Environmental Monitoring Spectrometer (GEMS) of a plurality of gases obtained, it is possible to calculate a difference value between the model radiation amount and the observed radiation amount (s230 to s250).

As described above, the process of the processor 10 using a forward model implemented as a neural network model, inputting an input value assigned to a predetermined state variable to an input node of the forward model, and outputting a model copy amount through an output node Inference can be performed on the forward model through In the case of the present invention, by using a forward model implemented as a neural network model, a more accurate output value is obtained and the operation speed is increased.

The processor 10 determines whether the difference between the model radiation amount and the observed radiation amount is less than a predetermined threshold (s260), and if it is not less than the predetermined threshold, select some of the reference data until it reaches less than the predetermined threshold After allocating another input value (eg, >10 ppm) of (eg > concentration of gas) to a predetermined state variable ( s270 ), the above-described steps s230 to s250 may be repeated again. That is, the above process can be repeated so that the numerical values of the model radiation amount and the observed radiation amount become almost similar.

을 최소화하는 과정에 의해 수행될 수 있다.The process of re-assigning to a predetermined state variable until it reaches less than a predetermined threshold is expressed in Equation 1 below.

It can be carried out by a process that minimizes

Equation 1:

비용 함수

: 소정의 상태 변수,

: 상태 변수의 초기값, Sa: 모델 에러 공분산 행렬,

: 측정 스펙트럼,

: 복사 전달 모델 구동 결과의 모델 복사량,

: 측정 에러 공분산 행렬, T: Transpose 연산 b: 복사 전달 모델의 입력값 중 상태 변수 이외의 변수) (

cost function

: a given state variable,

: initial value of state variable, Sa: model error covariance matrix,

: measurement spectrum,

: Model radiation amount of the radiative transfer model driving result,

: Measurement error covariance matrix, T: Transpose operation b: Variables other than state variables among the input values of the radiative transfer model)

The model error covariance Sa can be thought of as the uncertainty of xa. It can be said that it corresponds to the variance of the equation of the multivariate normal distribution. For example, if the state variable x=(NO ₂ concentration, O ₃ concentration)^T is defined, then xa becomes a two-element vector and Sa becomes a 2x2 matrix. Here, the (1,1) element of Sa is the dispersion of the NO ₂ concentration, and the (2,2) element is the dispersion of the O ₃ concentration. Even if any value is set as the initial value of the NO ₂ and O ₃ concentrations, the values cannot be trusted indefinitely, so their uncertainty is expressed as a variance value. The terms (1,2) and (2,1) represent the covariance between the two variables.

Theoretically, the covariance matrix should be constructed with the variance and covariance values, but in practice, the matrix is constructed empirically because the true values of these variances and covariances cannot be known. For example, if observation values for m wavelengths are used, the measurement spectrum y becomes a 1 x m vector and Se becomes a m x m matrix. The element (i,i) of the Se matrix is the variance (uncertainty) of the observed radiation at the i-th wavelength and the element (i,j) is the covariance between the observation at the i-th wavelength and the observation at the j-th wavelength. The diagonal component of the matrix generally uses an observation noise value announced by a satellite observing entity, and values other than the diagonal component are assumed to be zero.

In addition, other input values of the selected part of the reference data may be calculated by Equation 2 below.

Equation 2:

i번째 반복 단계에서의 상태 변수,

: i+1번째 반복 단계에서의 상태 변수,

: Levenberg-Marquardt 방정식 매개 변수,

: i번째 반복 단계에서의 자코비안 행렬(

)) (

the state variable at the ith iteration step,

: the state variable in the i+1th iteration step,

: parameters of the Levenberg-Marquardt equation,

: Jacobian matrix at the ith iteration step (

)))

The gamma value is determined empirically. The larger the gamma value is, the smaller the value added to the xi term becomes, so the amount of change in x at each step becomes smaller. If the gamma value is too small and the change in x is too large, the x value may have an unrealistic value (eg, a negative concentration value). If the gamma value is too large, the change in x is too small and many steps must be taken to reach convergence.

A Jacobian matrix is a three-dimensional matrix. The (i,j,k) element of the Jacobian matrix is the change in the model radiation amount of the k-th wavelength according to the change of the j-th state variable in the i-th step. A variable indicating how sensitive the model radiation amount is to each element of the state variable.

In addition, when it is determined that less than the predetermined threshold is reached through the repetition, another input value of a selected part of the reference data may be output as a final calculated value and provided to the user (s280).

That is, the last input value assigned to the state variable in a state in which the model radiation amount and the observed radiation amount are almost similar may be provided to the user as a final output value.

3 is a block diagram illustrating the process described above in FIG. 2 .

s210 of FIG. 2 is referenced by s310 and s311 of FIG. 3, s220 is referenced by s320, s230 is referenced by s330 and s340, s240 is referenced by s350 and s340, and s250 to s270 are referenced by s360. , and s280 may be referenced by s370.

The embodiments described above may be implemented in the form of program instructions that can be executed through various computer components and recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination.

Examples of the computer-readable recording medium include a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, an optical recording medium such as a CD-ROM and DVD, and a magneto-optical medium such as a floppy disk. optical media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device may be configured to act as one or more software modules for carrying out the processing according to the present invention, and vice versa.

Aspects herein may take the form of a computer program product embodied on one or more computer readable media embodied in hardware entirely, software entirely (including firmware, resident software, microcode, etc.) or computer readable program code. .

Features, structures, effects, etc. described in the above embodiments are included in one embodiment of the present invention, and are not necessarily limited to one embodiment. Furthermore, features, structures, effects, etc. illustrated in each embodiment can be combined or modified for other embodiments by those of ordinary skill in the art to which the embodiments belong. Accordingly, the contents related to such combinations and modifications should be interpreted as being included in the scope of the present invention.

Claims (10)

(a) preparing reference data representing predetermined characteristics of a plurality of gases;
(b) obtaining a model radiation amount by driving a radiation transfer model based on the reference data of the plurality of gases;
(c) acquiring the observed radiation through a Geostationary Environmental Monitoring Spectrometer (GEMS);
(d) calculating a difference value between the model radiation amount and the observed radiation amount; and
(e) repeating (b) to (d) until the difference value is less than a predetermined threshold;
A method for calculating the concentration of a plurality of gases. The method of claim 1,
Allocating an input value of a selected part of the reference data of the plurality of gases to a predetermined state variable;
In (b), the radiative transfer model is driven based on the input value assigned to the predetermined state variable.
A method of calculating the concentration of a plurality of gases. 3. The method of claim 2,
wherein (e) is performed by re-assigning another input value of the selected part of the reference data to the predetermined state variable,
A method for calculating the concentration of a plurality of gases. 4. The method according to any one of claims 1 to 3,
The reference data includes at least one of a surface reflectivity of the plurality of gases, an altitude from the surface of the earth, a concentration for each gas, and a radiation amount absorption degree for each gas,
A method for calculating the concentration of a plurality of gases.
4. The method of claim 3,
The (e) is of the formula

is carried out by a process that minimizes

(

: a given state variable,

: initial value of state variable, Sa: model error covariance matrix,

: measurement spectrum,

: Model radiation amount of the radiative transfer model driving result,

: Measurement error covariance matrix, T: Transpose operation b: Variables other than state variables among the input values of the radiative transfer model)
A method of calculating the concentration of a plurality of gases. 6. The method of claim 5,
Other input values of the selected part of the reference data of (e) are calculated by the following formula,

(

the state variable at the ith iteration step,

: the state variable in the i+1th iteration step,

: parameters of the Levenberg-Marquardt equation,

: Jacobian matrix at the ith iteration step (

)))
A method for calculating the concentration of a plurality of gases. 3. The method of claim 2,
The copy transfer model is a neural network model consisting of an input node, a hidden node, and an output node,
The (b) is,
Inference on the radiation transfer model is performed by inputting the input value assigned to the predetermined state variable to the input node and outputting the model radiation amount from the output node,
A method for calculating the concentration of a plurality of gases. The method of claim 1,
The plurality of gases is at least two selected from the group consisting of O ₃ , CO _2, aerosol, SO ₂ , NO _2, CH ₂ O,
A method for calculating the concentration of a plurality of gases. 9. The method of claim 8,
The plurality of gases is the O ₃ , CO _2, aerosol, SO ₂ , NO _2, CH ₂ O is classified through a neural network to be included in a specific group,
A method for calculating the concentration of a plurality of gases. processor; and
Including; a memory including at least one instruction executable by the processor;
The processor is
Prepare reference data representing predetermined characteristics of a plurality of gases, drive a radiation transfer model based on the reference data of the plurality of gases to obtain a model radiation amount, and radiation amount observed through a Geostationary Environmental Monitoring Spectrometer (GEMS) , calculating the difference between the model radiation amount and the observed radiation amount, and repeatedly performing (b) to (d) until the difference value is less than a predetermined threshold,
A device for calculating the concentration of a plurality of gases.

Images