KR102100739B1 - Method for predicting fine dust occurrence of target area - Google Patents

Abstract

According to the present invention, disclosed is a fine dust predicting method which generates a multilinear model regression equation using fine dust data, air pollution data, and weather observation data and predicts fine dust generation using the multilinear model regression equation.

Description

METHOD FOR PREDICTING FINE DUST OCCURRENCE OF TARGET AREA}

The present invention relates to a method for predicting the occurrence of fine dust, and more particularly, to a method for predicting the occurrence of fine dust in a target region using a multi-linear model regression equation.

Referring to the 'fine dust forecast accuracy status' described in Non-Patent Document 1, Korea's fine dust forecast accuracy averages 80%, including 84% in 2014, 87% in 2015, 86% in 2016, and 89% in 2017. Remained in the second half. However, the accuracy of forecasts calculated by computer numerical models is only about 50% on average because the forecaster's judgment is about 30% on average.

As the Ministry of Environment started fine dust forecasting in 2014, the fine dust prediction model developed by the U.S. Environmental Agency was evaluated and optimized using U.S. observational data, using the fine dust prediction model developed by the U.S. Environmental Agency. It is a model of the old program, and it is known that it has been mainly used for air quality analysis rather than air quality forecasting.

Therefore, conventionally, the fine dust prediction model has not been evaluated using Korea's weather observation data and fine dust data. In addition, in the related art, not only the fine dust is affected by the weather, but also may be affected by air pollutants or environmental conditions in a specific region. have.

1.http: //www.dailyt.co.kr/news/articleView.html? Idxno = 19356

In order to solve the above problems, the present invention generates a multilinear model regression equation using fine dust data, air pollution data, and weather observation data, and predicts the occurrence of fine dust in a target region using the multilinear model regression equation.

The present invention analyzes the correlation between the final independent variable affecting fine dust and the target area information, or analyzes the correlation between the final independent variable affecting fine dust and the final independent variable affecting plant growth state.

The present invention simulates the occurrence of fine dust according to the abnormal climate or the concentration of abnormal air pollutants using a multilinear model regression equation.

The method for predicting fine dust generation using a fine dust generation prediction apparatus including an input / output unit, a calculation unit, an analysis unit, and a control unit of the present invention for the above-mentioned problems to be solved is: Inputting observation values of fine dust set as a dependent variable using data and inputting observation values of a plurality of air pollutants and meteorological elements set as independent variables; Calculating non-standardized coefficient values and constant values for each independent variable by using a regression function between the dependent variable and the independent variable in the operation unit; Calculating a standardization coefficient for each independent variable in the operation unit, selecting a standardization coefficient exceeding a first coefficient threshold value in the control unit, and setting the independent variable having the selected standardization coefficient as a higher independent variable; Calculating a determination coefficient for each upper independent variable in the calculation unit, selecting a determination coefficient exceeding a second coefficient threshold, and setting an independent variable having a determination coefficient selected by the control unit as a final independent variable; Generating a multilinear model regression equation including the non-standardized coefficients and constant values of the final independent variable in the calculation unit, and generating fine dust by using the multilinear model regression equation as air pollutants and meteorological factors in the analysis unit And a step of predicting.

The analysis unit may be characterized by deriving the cause of the fine dust by analyzing the correlation between the final independent variable and the target area information set using the fine dust data, air pollution data, and weather observation data of the target area.

The analysis unit inputs a plant state observation value set as a dependent variable and a plurality of growth environment observation values set as an independent variable using plant growth state data and plant growth environment data, and uses a regression function between the dependent variable and the independent variable. The final independent variable affecting the state of plant growth is selected, and the correlation between the final independent variable affecting the fine dust and the final independent variable affecting the plant growth state may be analyzed.

The analysis unit receives the simulated values of the final independent variables and substitutes them into the multiple linear model regression equation to calculate the simulated results of the dependent variables, and uses the simulated results to generate fine dust according to the abnormal climate or the concentration of abnormal air pollutants. It may be characterized by simulating a.

The present invention can improve the accuracy of forecasting fine dust by generating a multilinear model regression equation using fine dust data, air pollution data, and weather observation data, and predicting occurrence of fine dust using the multilinear model regression equation. .

The present invention can analyze the correlation between the final independent variable and the target area information to find out the cause of the fine dust and derive a method for reducing the fine dust.

The present invention can utilize analysis information analyzing the correlation between the final independent variable affecting fine dust and the final independent variable affecting plant growth state for plant growth, and can contribute to high-quality plant production.

The present invention can derive a coping method according to the occurrence of fine dust by analyzing various result values that simulate the occurrence of fine dust according to the abnormal climate or the concentration of abnormal air pollutants.

1 is a block diagram illustrating a fine dust generation prediction apparatus 100 according to an embodiment of the present invention.
Figure 2a is an example showing a scatter plot between PM ₁₀ and the average underground temperature of 1.0m.
Figure 2b is an example showing a scatter plot between PM ₁₀ and the average relative humidity.
Figure 2c is an example showing the scatter plot between PM ₁₀ and carbon monoxide.
Figure 2d is an example showing the scatter plot between PM ₁₀ and large total evaporation.
Figure 2e is an example showing the scatter plot between PM ₁₀ and the highest temperature.
3 is an example showing a regression line graph between the independent variable and the dependent variable.
4 is a flowchart illustrating a method for predicting the occurrence of fine dust in a target area according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and the contents described in the accompanying drawings, but the present invention is not limited or limited by the embodiments.

In the atmospheric environment conservation method, air pollutants were defined as 61 types including gaseous and particulate matter, and among them, fine dust is classified into PM ₁₀ , PM _2.5 , and PM _1.0 by particle size. PM ₁₀ is a fine dust having a diameter of 10 micrometers or less, PM _2.5 is a very fine dust, and PM _1.0 is a very fine dust.

Fine dust sources are divided into artificial and natural sources, and artificial sources include fumes in industrial complexes, exhaust gases from automobiles, flying dust at construction sites, and smoke from incinerators, and natural sources are soil dust, salt from seawater And pollen.

1 is a block diagram showing a fine dust generation prediction apparatus 100 according to an embodiment of the present invention, and the fine dust generation prediction apparatus 100 predicts fine dust generation in a target area. The target area can be divided into city, province, county, town, or ri units, but is not limited thereto.

The fine dust generation prediction apparatus 100 includes an input / output unit 110, a calculation unit 120, an analysis unit 130, and a control unit 140. The input / output unit 110 inputs observation values of fine dust set as a dependent variable using fine dust data, air pollution data, and weather observation data, and a plurality of air pollutants set as independent varibles And Observation values of meteorological factors are input. In more detail, the input / output unit 110 inputs observation values of fine dust set as a dependent variable using fine dust data, air pollution data and weather observation data of the target area, and a plurality of air pollutants set as independent variables. Observation values of weather elements can be input. The control unit 140 may receive air pollution data and weather observation data and set independent variables for each air pollutant and weather elements. The independent variable is a variable that can be changed independently of other variable changes, and the dependent variable is a variable that changes as the independent variable changes.

Air pollutants may be ozone (O ₃ ), carbon monoxide (CO), nitrogen oxides (NO _X ) or sulfur oxides (SO _X ), and are various substances that adversely affect organisms or substances. The nitrogen oxides may be nitrogen monoxide (NO), nitrogen dioxide (NO ₂ ), nitrous oxide (N ₂ O) or dinitrogen tetraoxide (N ₂ O ₄ ), and are various compounds composed of compounds of nitrogen and oxygen. The sulfur oxide may be sulfur dioxide (SO ₂ ) or sulfur trioxide (SO ₃ ), and is a variety of compounds composed of a compound of sulfur and oxygen.

Meteorological factors can be ground, ground temperature, temperature, precipitation, wind, humidity, air pressure, sunlight, solar radiation, evaporation, snow or clouds, and are various factors that can measure the phenomenon occurring in the atmosphere. The ground and ground temperature are average ground temperature, average minimum ground temperature, minimum ground temperature, 0.05m average ground temperature, 0.1m average ground temperature, 0.2m average ground temperature, 0.3m average ground temperature, 0.5m average ground temperature, 1.0m It can be divided into average ground temperature, 1.5m average ground temperature, 3.0m average ground temperature, or 5.0m average ground temperature.

The temperature can be divided into an average temperature, an average maximum temperature, an average minimum temperature, a maximum temperature, or a minimum temperature. Precipitation can be divided into monthly precipitation, maximum precipitation, maximum precipitation per hour, or 10 minutes maximum precipitation. The wind may be divided into an average wind speed, a maximum wind speed, a maximum instantaneous wind speed, a maximum wind speed wind direction, or a maximum instantaneous wind speed wind direction. Humidity can be divided into average relative humidity or minimum relative humidity.

The air pressure can be classified into an average local barometric pressure, an average water vapor pressure, an average dew point temperature, an average sea level pressure, a maximum sea level pressure, a minimum sea level pressure, a maximum water vapor pressure, or a minimum water vapor pressure. The amount of sunshine and insolation can be divided into the sum of sunshine time, sunshine rate, or total sun sunshine. The amount of evaporation can be divided into small total evaporation, small maximum evaporation, large total evaporation, or large daily evaporation. Snow can be represented by average neutral cloud cover, and clouds can be represented by average cloud cover. The amount of evaporation is the amount of water evaporated in a unit area over a period of time.

The input / output unit 110 may provide a communication function, and may receive fine dust data, air pollution data, and weather observation data of a target area from a statistical data server (not shown). The statistical data server may be a server operated by a local government or a country in the target area. For example, fine dust data and air pollution data may be data provided by the Environmental Management Corporation, the City Hall or the county office, and weather observation data may be data provided by the Meteorological Agency's weather data opening portal.

Figure 2a is an example showing the scatter plot between PM ₁₀ and the average ground temperature of 1.0 m, Figure 2b is an example showing the scatter plot between PM ₁₀ and the average relative humidity, Figure 2c is an example showing the scatter plot between PM ₁₀ and carbon monoxide, Figure 2d Is an example showing the scatter plot between PM ₁₀ and large total evaporation, and FIG. 2E is an example showing the scatter plot between PM ₁₀ and the maximum temperature. The scatter plot is a statistical graph created by taking observations on a plane of rectangular coordinates to see the relationship between two continuous variables such as dependent and independent variables.

The calculating unit 120 calculates a non-standardized coefficient and a constant for each independent variable using a regression function between the dependent variable and the independent variable. More specifically, the calculation unit 120 uses a regression function for a regression line that passes closest to a plurality of points distributed in the scatter plots shown in FIGS. 2A to 2E to determine the non-standardized coefficients and constants for each independent variable. To calculate. [Equation 1] is an equation for calculating the regression line.

[Equation 1]

는 종속변수이고,

는 독립변수이며,

는 독립변수의 인덱스이고,

은

의 추정값에 대한 평균이며,

는 상수이고,

는 비표준화계수이다.here,

Is a dependent variable,

Is an independent variable,

Is the index of the independent variable,

silver

Is the mean for the estimate of

Is a constant,

Is the non-standardization factor.

If the unit of the independent variable is different, the non-standardized coefficient will also be different. Since the present invention is difficult to compare the magnitude of the relative influence as a non-standardized coefficient for each independent variable, the operator 120 calculates the standardization coefficient for each independent variable to compare the magnitude of the relative influence of each independent variable. To calculate. Because the standardization coefficient is independent of units, it is used to compare the relative influence of independent variables. For example, if the standardization coefficient is large, it can be said that the influence of the independent variable on the dependent variable is also large. The calculation unit 120 uses [Equation 2] when calculating the standardization coefficient.

[Equation 2]

는 표준화계수이고,

는

의 관측값에 대한 평균이며,

는

의 관측값에 대한 평균이다.here,

Is the standardization coefficient,

The

Is the mean of the observed values of

The

Is the average of the observed values of.

The control unit 140 selects a standardization coefficient exceeding the first coefficient threshold, and sets the independent variable having the selected standardization coefficient as the upper independent variable. The first coefficient threshold is the value that becomes the boundary between the rejected and adopted areas in the hypothesis test of the standardized coefficients.

3 is an example showing a regression linear graph between the independent variable and the dependent variable, since the calculating unit 120 may have a variable having low explanatory power among the upper independent variables, calculates a determination coefficient for each upper independent variable.

로 표기되고, 관측값의 표본자료를 얼마나 잘 설명하는지에 대한 평가기준이 되는 계수이며, [수식 3]을 참조하며 1에 가까우면 표본자료를 잘 설명하는 것이고, 0에 가까우면 자료를 거의 설명하지 못한다는 것을 의미한다. 표본자료를 잘 설명한다는 것은 정밀도가 높다는 것을 의미하고, 표본자료를 거의 설명하지 못한다는 것은 정밀도가 낮다는 것을 의미한다.The coefficient of determination is an explanatory power

It is denoted as, and is a coefficient that serves as an evaluation criterion for how well the sample data of the observations are explained. Refer to [Equation 3] and close to 1, the sample data is well explained. If it is close to 0, the data is almost described. It means you can't. A good explanation of sample data means high precision, and a very little description of sample data means low precision.

[Equation 3]

의 최댓값이고, [수식 8]은

의 최솟값이다.

의 최댓값과 최솟값의 차이가 적을수록 결정계수의 신뢰도가 향상된다.[Equation 7] and [Equation 8] for calculating the coefficient of determination are derived from [Equation 4] to [Equation 6]. [Equation 7]

Is the maximum value of [Equation 8]

Is the minimum value of

The smaller the difference between the maximum and minimum values of, the more reliable the coefficient of determination.

[Equation 4]

[Equation 5]

[Equation 6]

[Formula 7]

[Equation 8]

는 관측값 또는 추정값의 인덱스이다. SSE(explained sum of squares)는 설명된 변동으로서

의 추정값에 대한 편차제곱의 합이고, SSR(residual sum of squares)은 설명 안된 변동으로서

의 잔차에 대한 제곱의 합이며, SST(total sum of squares)는 SSE와 SSR을 합한 총 변동으로서

의 개별적인 편차제곱의 합이다.here,

Is the index of the observed or estimated value. Explained sum of squares (SSE) is the described variation

The sum of squares of deviations from the estimate of, and the residual sum of squares (SSR) is an unexplained variance.

Is the sum of squares of the residuals of, and the total sum of squares (SST) is the total change of the sum of SSE and SSR.

Is the sum of the individual squared deviations.

The control unit 140 selects a determination coefficient that exceeds the second coefficient threshold, and sets an independent variable having the selected determination coefficient as a final independent variable. The second coefficient threshold is the value that becomes the boundary between the rejected region and the adopted region in the hypothesis test of the coefficient of determination.

The present invention improves the accuracy of forecasting fine dust for a target area by selecting the independent variables that affect the occurrence of fine dust in the target area by firstly selecting the upper independent variable and secondly selecting the final independent variable. I can do it.

The calculation unit 120 generates a multilinear model regression equation including the non-normalized coefficient values and constant values of the final independent variable. The multiple linear model regression equation is as shown in [Equation 9].

[Equation 9]

[Equation 10]

는 최종 독립변수의 개수이다.here,

Is the number of final independent variables.

Due to the geographical characteristics and the structure of the urban industry, the frequency and concentration of fine dust are different for each target region, so a predictive alarm model is required for each target region. The analysis unit 130 predicts the occurrence of fine dust in the target region using the multilinear model regression equation. In addition, the analysis unit 130 analyzes the correlation between the final independent variable and the target region information. The target area information includes at least one of geographic information, industrial status information, development status information, vehicle registration number status, and oil consumption status regarding geographical characteristics and urban industrial structure.

The analysis unit 130 analyzes the correlation between the final independent variable and geographic information because a fine dust concentration difference may occur depending on whether it is an inland or coastal area, an area influenced by the western wind, or a basin area.

The analysis unit 130 correlates the industry status information, development status information, vehicle registration number status or oil consumption status and the final independent variable associated with the fine dust artificial origin since the cause of the fine dust is mostly related to the artificial origin of fine dust. Analyzes. The industry status information may include information related to the establishment of industries related to artificial sources of fine dust in the target area, such as soot from factories, and the development status information may include artificial sources of fine dust in the target area, such as dust from construction sites. Whether there is a related construction status or not, it may include information related to the construction status, and the number of vehicle registrations or oil consumption status may include information on the number of vehicles or fuel consumption related to artificial sources of fine dust in the target area, such as soot of vehicles have.

The present invention can analyze the correlation between the final independent variable and the target area information to find out the cause of fine dust, and refer to the cause of fine dust generation to derive a method to reduce fine dust for each target area. Regional forecasting models can be provided.

Fine dust is suspended in the air and deposited on the leaf surface of plants, blocking pores of the leaves and blocking sunlight, thereby inhibiting photosynthesis, assimilation, respiration and evaporation, and adversely affecting plant growth.

Ozone (O ₃ ) is a secondary pollutant produced by the photochemical reaction of nitrogen oxides (NOx) and volatile organic compounds (VOCs) in the atmosphere by sunlight, and has strong oxidizing power. It provides action, but having a concentration above a certain level affects the crop yield reduction. Carbon monoxide (CO) contributes to tropospheric ozone production.

Sulfur dioxide (SO ₂ ) is artificially released when fossil fuels are burned, and the main sources of emissions are from power plants, heating systems, metal smelting plants, refineries and other industrial processes. Sulfur dioxide is a substance that lowers acid rain with nitrogen oxides, affects soil acidification, and affects plant vein damage or growth inhibition.

As described above, air pollutants, which are possible sources of fine dust or fine dust, affect plant growth, so it is necessary to find out the independent variables affecting plant growth conditions in the target area, and the final independence affecting fine dust It is necessary to analyze the correlation between the variable and the final independent variable affecting plant growth status.

The input / output unit 110 inputs a plant state observation value set as a dependent variable and a plurality of growth environment observation values set as an independent variable using plant growth state data and plant growth environment data. The calculating unit 120 calculates the variable value and the constant value through the input plant state observation and the growth environment observation value. The controller 140 selects the final independent variable that affects the state of plant growth by using a regression function between the dependent variable and the independent variable. The analysis unit 130 analyzes the correlation between the final independent variable affecting fine dust and the final independent variable affecting plant growth state.

The present invention can utilize analysis information analyzing the correlation between the final independent variable affecting fine dust and the final independent variable affecting plant growth state for plant growth, and can contribute to high-quality plant production. For example, the analysis unit 130 may derive common independent variables that commonly affect fine dust and plant growth conditions, and generate analysis information related to a solution of plant growth in response to observations of the common independent variables. You can. The plant growth apparatus that controls the greenhouse house may be equipped with a sensor unit equipped with measurement items related to common independent variables, and control the environment of the greenhouse house with reference to the analysis information.

The fine dust occurrence prediction apparatus 100 receives the sensor values corresponding to the final independent variables through the sensor unit installed in the plant growth apparatus located in the target area, substitutes them into the multilinear model regression equation, calculates the predicted result values of the dependent variables, and , The operation of the plant growth apparatus can be controlled by referring to the predicted result value and the plant growth information. The biological growth information may include plant state information indicating a plant growth state and growth environment information indicating a plant growth environment.

The fine dust occurrence prediction apparatus 100 receives the simulated value of the final independent variable, substitutes it into the multilinear model regression equation, calculates the simulated result of the dependent variable, and uses the simulated result to determine the abnormal climate or abnormal air pollutants. It is possible to simulate the occurrence of fine dust according to the concentration.

The present invention can derive a coping method according to the occurrence of fine dust by analyzing various result values that simulate the occurrence of fine dust according to the abnormal climate or the concentration of abnormal air pollutants.

4 is a flowchart illustrating a method for predicting occurrence of fine dust in a target area according to an embodiment of the present invention, and the method for predicting fine dust generation predicts occurrence of fine dust in the target area through the apparatus 100 for predicting fine dust generation.

The apparatus 100 for predicting occurrence of fine dust inputs observation values of fine dust set as a dependent variable using fine dust data, air pollution data, and weather observation data of a target area, and a plurality of air pollutants and weather set as independent variables Enter the observations of the element. Next, the non-normalized coefficient values and constant values for each independent variable are calculated using the regression function between the dependent variable and the independent variable. Next, a standardization coefficient for each independent variable is calculated, a standardization coefficient exceeding a first coefficient threshold is selected, and an independent variable having a selected standardization coefficient is set as a higher independent variable. Next, the coefficient of determination for each upper independent variable is calculated, the coefficient of determination that exceeds the second coefficient threshold is selected, and the independent variable having the selected coefficient is set as the final independent variable. Next, a multilinear model regression equation is generated including the non-normalized coefficient values and constant values of the final independent variable, and the fine dust occurrence in the target region is predicted using the multilinear model regression equation.

[Table 1] is a table showing the results of multiple regression analysis for PM ₁₀ by model, [Table 2] is a table showing the decision coefficient for each model, and [Table 3] is a table showing the multilinear model regression equation for each model .

[Table 1] to [Table 3] are the results obtained by extracting the fine dust data, air pollution data, and meteorological observation data of the Daejeon area from 2007 to 2017 on a monthly basis, and the apparatus 100 for predicting the occurrence of fine dust according to the present invention This is the result obtained by utilizing. The data are provided by the Korea Meteorological Administration and the Environmental Management Corporation. The probability of significance (P) is the probability that an extreme result will actually be observed when the null hypothesis is assumed to be correct, is a random variable defined in the sample space of the experiment, and has a value between 0 and 1.

As shown in [Table 1] to [Table 3], referring to Model 5, which has the highest coefficient of determination in Daejeon, the largest total evaporation amount was 0.492 in the positive influence size, followed by 0.264. In the magnitude of wealth influence, the average underground temperature of 1.0 m was highest at -0.566, and the highest at -0.325. More specifically, as the total total evaporation amount and the carbon monoxide concentration increased, the fine dust concentration also increased. On the other hand, when the average ground temperature and the maximum temperature decreased, the fine dust concentration decreased.

The present invention is to find out the final independent variable that affects the occurrence of fine dust in the target area, and analyze the characteristics of the final independent variable and the target area together to utilize it in the field of managing fine dust in the target area.

100: fine dust generation prediction device
110: input and output unit 120: operation unit
130: analysis unit 140: control unit

Claims (4)

In the fine dust generation prediction method using a fine dust generation prediction apparatus including an input and output unit, a calculation unit, an analysis unit and a control unit,
Inputting observation values of fine dust set as a dependent variable using fine dust data, air pollution data, and weather observation data in the input / output unit, and inputting observation values of a plurality of air pollutants and meteorological elements set as independent variables. ;
Calculating non-standardized coefficient values and constant values for each independent variable by using a regression function between the dependent variable and the independent variable in the operation unit;
Calculating a standardization coefficient for each independent variable in the operation unit, selecting a standardization coefficient exceeding a first coefficient threshold value in the control unit, and setting the independent variable having the selected standardization coefficient as a higher independent variable;
Calculating a determination coefficient for each upper independent variable in the calculation unit, selecting a determination coefficient exceeding a second coefficient threshold, and setting an independent variable having a determination coefficient selected by the control unit as a final independent variable;
Generating a multilinear model regression equation including the non-normalized coefficient values and constant values of the final independent variable in the operation unit, and
And predicting the occurrence of fine dust using a multi-linear model regression equation as an air pollutant and a meteorological element in the analysis unit. According to claim 1,
The analysis unit predicts the occurrence of fine dust by analyzing the correlation between the final independent variable and the target area information set using the fine dust data, air pollution data, and weather observation data of the target area. Way. According to claim 1,
The analysis unit inputs a plant state observation value set as a dependent variable and a plurality of growth environment observation values set as an independent variable using plant growth state data and plant growth environment data, and uses a regression function between the dependent variable and the independent variable. A method for predicting the occurrence of fine dust characterized by selecting the final independent variable affecting the state of plant growth and analyzing the correlation between the final independent variable affecting the fine dust and the final independent variable affecting the plant growth state . According to claim 1,
The analysis unit receives the simulated values of the final independent variables and substitutes them into the multiple linear model regression equation to calculate the simulated results of the dependent variables, and uses the simulated results to generate fine dust according to the abnormal climate or the concentration of abnormal air pollutants. Method for predicting the occurrence of fine dust, characterized in that the simulation.

Images