KR102401823B1 - System and method for computing thermal environment application index by disaster type, and weather factor of intense heat and freeze - Google Patents

Abstract

Provided are a system and a method for computing a thermal environment application index. A system for computing a thermal environment application index according to an embodiment of the present invention comprises: a ground surface analysis unit which builds a ground surface analysis model having a first local scale resolution set based on geographic information system (GIS) information, and uses the ground surface analysis model to establish a plurality of terrain variables which affect local weather; a learning unit which collects a plurality of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution, applies a machine learning technique to the first meteorological office neighborhood forecast information and the terrain variable corresponding to the first meteorological office neighborhood forecast, so as to extract the association between the first meteorological office neighborhood forecast information and a topographical variable corresponding to the first meteorological office neighborhood forecast, and generates weather information on the scale of the first resolution by detailing the first meteorological office neighborhood forecast information on the scale of the first resolution based on the extracted association; and a weather information calculation unit which refines the second meteorological office neighborhood forecast information into weather information on the scale of the first resolution based on the association and the weather information stored in the learning unit when the second meteorological office neighborhood forecast information is collected and outputs the detailed weather information.

Description

SYSTEM AND METHOD FOR COMPUTING THERMAL ENVIRONMENT APPLICATION INDEX BY DISASTER TYPE, AND WEATHER FACTOR OF INTENSE HEAT AND FREEZE

Embodiments of the present invention are related to a technology for calculating a thermal environment application index by using artificial intelligence (Artificial Intelligence) technology to refine the local weather forecast of the Korea Meteorological Administration on a local scale that is a small range of several kilometers.

The Meteorological Administration's neighborhood forecast is a forecast service that has been officially operated since 2008. Spatially, it is provided in units of eup/myeon/dong, and in terms of time, it is provided at 1 hour/3 hour intervals. Recently, studies using local forecasts from the Korea Meteorological Agency have been actively conducted, and in particular, as the demand for local forecasts from the Korea Meteorological Agency increases in the field of living environment, the use of these forecasts is also expanding.

However, such local weather forecasts from the Korea Meteorological Administration are currently provided with a resolution of 5 km, and more detailed information is needed to provide more accurate information in living environment-related fields. In particular, in order to produce meteorological information that more closely affects real life, it is necessary to consider the effects of atmospheric processes and local conditions of the earth's surface for meteorological factors on a local scale, ranging from several kilometers.

Korean Patent Publication No. 10-2098356 (2020.04.01)

Embodiments of the present invention are for calculating statistically detailed weather information by applying the influence on the local conditions of the earth's surface to the local weather forecast information of the Korea Meteorological Administration.

According to an exemplary embodiment of the present invention, a ground surface analysis model having a first resolution of a local scale set based on Geographic Information System (GIS) information is constructed, and using the ground surface analysis model, a surface analysis unit that calculates a number of topographical variables that have an influence; A plurality of first meteorological office neighborhood forecast information for a scale of a second resolution larger than the first resolution is collected, and machine learning (machine learning) is performed on the first meteorological office neighborhood forecast information and the terrain variable corresponding to the first meteorological office neighborhood forecast information. learning) technique to extract the correlation between the first meteorological office neighborhood forecast information and the topographical variable corresponding to the first meteorological agency neighborhood forecast, and based on the extracted correlation, the first meteorological agency neighborhood forecast information to the first resolution a learning unit for generating meteorological information on the scale of the first resolution by detailing on the scale of ; And when the second meteorological office neighborhood forecast information is collected, based on the association and weather information stored in the learning unit, the second meteorological office neighborhood forecast information is detailed as meteorological information for the scale of the first resolution, and the detailed weather information A thermal environment application index calculation system is provided, including a meteorological information calculation unit that outputs

The ground surface analysis unit converts the ground surface analysis input data including the numerical topographic map regarding the height and shape of the terrain, the numerical building map regarding the shape and height of the building, and the land cover map regarding the land use into a set coordinate system, and the converted The ground surface analysis model may be constructed by constructing the input data as raster data having a first resolution of the local scale.

The topographical variables include the latitude of the topography, the longitude of the topography, the elevation of the topography, the slope of the topography, the distance from the shoreline of the topography, the land cover of the topography, the slope in the north-south or east-west direction of the topography, the recessed depth of the topography, and the slope of the topography. It may include at least one of the length, the height of the building, the ratio of the side area of the building, the ratio of the floor area of the building, the roughness length of the building, and the area ratio of the specific land cover.

The learning unit selects set meteorological elements from among the first meteorological office neighborhood forecast information, converts the selected meteorological elements into a set file format, and the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast machine learning techniques can be applied to

The learning unit may apply a Gaussian Process Regression Model (GPRM) to the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast.

The learning unit may apply the GPRM after reducing the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information by a set ratio.

The set ratio may be 1/4.

The learning unit may verify the predicted weather information using observation data collected from an external server.

According to another exemplary embodiment of the present invention, the method comprising: constructing, in a surface analysis unit, a surface analysis model having a first resolution of a local scale set based on GIS information; calculating, in the ground surface analysis unit, a plurality of topographic variables affecting local weather by using the ground surface analysis model; collecting, in the learning unit, a plurality of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution; In the learning unit, by applying a machine learning technique to the terrain variable corresponding to the first meteorological office neighborhood forecast information and the first meteorological office neighborhood forecast, the first weather agency neighborhood forecast information and the first weather agency neighborhood forecast corresponding to the terrain extracting associations between variables; generating, in the learning unit, meteorological information on the scale of the first resolution by refining the neighborhood forecast information of the first meteorological agency based on the extracted correlation to the scale of the first resolution; When the second meteorological office neighborhood forecast information is collected in the meteorological information calculator, detailing the second meteorological office neighborhood forecast information into meteorological information for the scale of the first resolution based on the correlation and meteorological information stored in the learning unit ; and outputting the detailed meteorological information from the meteorological information calculator.

The step of constructing the ground surface analysis model includes setting the ground surface analysis input data including the numerical topographic map regarding the height and shape of the terrain, the numerical building map regarding the shape and height of the building, and the land cover map regarding the land use into a set coordinate system. The ground surface analysis model may be constructed by converting and constructing the converted input data as raster data having a first resolution of the local scale.

The topographical variables include the latitude of the topography, the longitude of the topography, the elevation of the topography, the slope of the topography, the distance from the shoreline of the topography, the land cover of the topography, the slope in the north-south or east-west direction of the topography, the recessed depth of the topography, and the slope of the topography. It may include at least one of the length, the height of the building, the ratio of the side area of the building, the ratio of the floor area of the building, the roughness length of the building, and the area ratio of the specific land cover.

The step of extracting the correlation includes selecting set meteorological elements from among the first meteorological office neighborhood forecast information, converting the selected meteorological elements into a set file format, and corresponding to the converted meteorological elements and the first meteorological office neighborhood forecast A machine learning technique can be applied to the terrain variable to be

In the extracting of the correlation, GPRM may be applied to the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast.

In the extracting of the correlation, the GPRM may be applied after reducing the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information by a set ratio.

The set ratio may be 1/4.

The method for calculating the thermal environment application index may further include, in the learning unit, verifying the predicted weather information using observation data collected from an external server, after predicting and storing the weather information. .

According to embodiments of the present invention, by calculating detailed weather information from the statistical correlation between various local conditions, such as terrain, structures, vegetation influence, etc. according to the Meteorological Agency neighborhood forecast information and urban surface composition, after machine learning, life It can help to provide more accurate information in environment-related fields.

1 is a block diagram showing the detailed configuration of a thermal environment application index calculation system according to an embodiment of the present invention;
2 is a block diagram illustrating a detailed configuration of a ground surface analysis unit according to an embodiment of the present invention;
3 is an example of a terrain variable calculated from a ground surface analysis model according to an embodiment of the present invention;
4 is a block diagram illustrating a detailed configuration of a learning unit according to an embodiment of the present invention;
5 is an example of the first meteorological office neighborhood forecast information collected by the learning unit according to an embodiment of the present invention;
6 is a graph showing a cross-validation result for each prediction element for each machine learning technique that can be applied to the learning unit according to an embodiment of the present invention;
7 is a graph showing a comparison result with external observation data for each machine learning technique that can be applied to the learning unit according to an embodiment of the present invention;
8 is a view showing a process of reducing the resolution of a sample space to a set ratio in the process of applying GPRM in the learning unit according to an embodiment of the present invention and calculation time performance accordingly
9 is an example of weather information predicted by the learning unit according to an embodiment of the present invention;
10 is a diagram for explaining a process of predicting weather information in a learning unit according to an embodiment of the present invention;
11 is a graph illustrating a process of verifying the accuracy of predicted weather information in a learning unit according to an embodiment of the present invention;
12 is a flowchart illustrating a method for calculating a thermal environment application index according to an embodiment of the present invention;
13 is a block diagram illustrating and describing a computing environment including a computing device suitable for use in example embodiments;

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The following detailed description is provided to provide a comprehensive understanding of the methods, devices, and/or systems described herein. However, this is merely an example, and the present invention is not limited thereto.

In describing the embodiments of the present invention, if it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. And, the terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to intentions or customs of users and operators. Therefore, the definition should be made based on the content throughout this specification. The terminology used in the detailed description is for the purpose of describing embodiments of the present invention only, and should in no way be limiting. Unless explicitly used otherwise, expressions in the singular include the meaning of the plural. In this description, expressions such as “comprising” or “comprising” are intended to indicate certain features, numbers, steps, acts, elements, some or a combination thereof, one or more other than those described. It should not be construed to exclude the presence or possibility of other features, numbers, steps, acts, elements, or any part or combination thereof.

1 is a block diagram showing a detailed configuration of a thermal environment application index calculation system 100 according to an embodiment of the present invention. As shown in FIG. 1 , the thermal environment application index calculation system 100 according to an embodiment of the present invention includes a ground surface analysis unit 102 , a learning unit 104 , and a weather information calculation unit 106 .

The ground surface analysis unit 102 builds a ground surface analysis model having a first resolution of a local scale set based on Geographic Information System (GIS) information, and uses the ground surface analysis model to affect local weather It yields a number of terrain variables.

2 is a block diagram showing a detailed configuration of the ground surface analyzer 102 according to an embodiment of the present invention. Referring to FIG. 2 , the surface analysis unit 102 according to an embodiment of the present invention may include a surface analysis input data collection unit 202 , a surface analysis model construction unit 204 , and a terrain variable calculation unit 206 . can

The ground surface analysis input data collection unit 202 collects ground surface analysis input data based on GIS information. The ground surface analysis input data may include, for example, a numerical topographic map related to the height and shape of the topography, a numeric building map related to the shape and height of a building, and a land cover map related to land use. The above-mentioned ground surface analysis input data is basic data for calculating meteorological variables on a national level. Here, the numerical topographic map includes the elevation and shape data of the topography, the numeric building map includes the shape and height data of the building, and the land cover map includes a set number of categories (eg, land use set by the Ministry of Environment). 7 categories) based on land use data, etc. may be included. In addition to the numerical topographical map, the numerical building map, and the land cover map, remote sensing data such as satellite data of topography and aerial lidar observation data may be used as input data for the analysis of the ground surface.

The ground surface analysis model building unit 204 builds a ground surface analysis model having a first resolution of a set local scale by using the ground surface analysis input data. Here, the first resolution of the set local scale may be, for example, a resolution of 1 km scale. In general, the resolution of the local weather forecast of the Korea Meteorological Administration is at the level of 5 km, and more detailed information is needed to provide more accurate information in living environment-related fields. For example, even within a radius of 5 km, the local temperature may vary depending on the local area or topographical characteristics. Accordingly, the ground surface analysis model building unit 204 may construct a ground surface analysis model in a range of 1 km, which is smaller than a range of 5 km. That is, the ground surface analysis model is a model in which the ground surface analysis input data is digitized with a first resolution of a local scale. The ground surface analysis model building unit 204 converts the ground surface analysis input data into a set coordinate system (eg, UTM-k coordinate system), and converts the converted input data into raster data having a first resolution of the local scale. It is possible to construct the above-mentioned ground surface analysis model by building as

The terrain variable calculation unit 206 calculates a plurality of terrain variables affecting local weather by using the ground surface analysis model. Since local weather varies depending on the size of the region, topographical characteristics, city type, and the like, the topographic variable calculating unit 206 calculates a plurality of topographical variables affecting the local weather by using the ground surface analysis model. Here, the terrain variable is, for example, the latitude of the terrain, the longitude of the terrain, the altitude of the terrain, the inclination of the terrain, the distance from the shoreline of the terrain, the land cover of the terrain, the slope in the north-south or east-west direction of the terrain, the depression of the terrain It may include depth, the length of the slope of the terrain, the height of the building, the ratio of the side area of the building, the ratio of the floor area of the building, the roughness length of the building, the area ratio of a specific land cover, and the like.

3 is an example of a terrain variable calculated from a ground surface analysis model according to an embodiment of the present invention.

Referring to FIG. 3 , the terrain variable calculation unit 206 uses the ground surface analysis model to influence the local weather, including land cover, slope, aspect, and building height. height) can be calculated.

Returning to FIG. 1 , the learning unit 104 generates and learns weather information having a first resolution of a local scale based on the terrain variable calculated by the ground surface analysis unit 102 . The learning unit 104 collects a plurality of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution, and the topography corresponding to the first meteorological office neighborhood forecast information and the first meteorological office neighborhood forecast By applying a machine learning technique to a variable, a correlation between the first meteorological office neighborhood forecast information and a terrain variable corresponding to the first meteorological office neighborhood forecast is extracted, and based on the extracted correlation, the first meteorological office neighborhood forecast Information on the scale of the first resolution may be detailed to generate weather information on the scale of the first resolution.

4 is a block diagram showing a detailed configuration of the learning unit 104 according to an embodiment of the present invention. Referring to FIG. 4 , the learning unit 104 according to an embodiment of the present invention includes a meteorological agency neighborhood forecast information collection unit 402 , a weather element selection unit 404 , a file transformation unit 406 , and a machine learning technique application unit. 408 and a post-processing unit 410 may be included.

The meteorological agency neighborhood forecast information collection unit 402 collects a plurality of first meteorological agency neighborhood forecast information. Here, the first meteorological office neighborhood forecast information is the meteorological office neighborhood forecast information collected in a nationwide unit, and may have a resolution of a second resolution larger than the first resolution. In this case, the second resolution may be, for example, a resolution of 5 km scale. The meteorological agency neighborhood forecast information collection unit 402 may collect a plurality of first meteorological agency neighborhood forecast information in real time, for example, from a server (not shown) of the Korea Meteorological Institute of Technology.

5 is an example of the first local weather forecast information collected by the learning unit 104 according to an embodiment of the present invention.

Referring to FIG. 5 , the first local weather forecast information is, for example, temperature (3 hours), daily maximum temperature, daily minimum temperature, east-west wind component, north-south wind component, wind direction, wind speed, sky condition, precipitation form, precipitation It may include information such as probability, precipitation (6 hours), new snow (6 hours), humidity, and wave height. In this case, the first meteorological office neighborhood forecast information may have a second resolution scale, for example, a 5 km scale resolution. The first meteorological office neighborhood forecast information may be in the form of, for example, a GRIB (GRIdded Binary) file.

Returning to FIG. 4 again, the meteorological element selection unit 404 selects the set meteorological elements from among the first meteorological office neighborhood forecast information. As an example, the meteorological element selection unit 404 may select 8 set meteorological elements among 14 meteorological elements shown in FIG. 5 . Note that the number and types of meteorological elements shown in FIG. 5 and the number and types of meteorological elements selected by the meteorological element selection unit 404 are merely examples and are not limited thereto.

The file conversion unit 406 converts the selected weather elements into a set file format. The file converter 406 may convert, for example, meteorological elements in a GRIB (GRIadded Binary) file format into a text file format (.TXT).

The machine learning technique application unit 408 applies the machine learning technique to the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast. Here, the machine learning technique may be, for example, a statistical downscaling method using a Gaussian Process Regression Model (GPRM), a Support Vector Machine (SVM) regression technique, or a Random Forest (RF) regression technique. The machine learning technique application unit 408 corresponds to the first meteorological agency neighborhood forecast information and the first meteorological agency neighborhood forecast by applying a machine learning technique to the converted meteorological elements and the terrain variables corresponding to the first meteorological office neighborhood forecast. It is possible to extract the correlation between the topographical variables that become

In the present embodiments, cross-validation was performed to select a machine learning technique with the highest accuracy among the various machine learning techniques (or statistical techniques) as described above. As an example, in the present embodiments, the entire data is classified into a verification set and a learning set at a ratio of 1:9 for each technique, and after repeatedly calculating statistical values for this, the calculation time for each technique, the sensitivity to applying the terrain variable, Spatial prediction error was calculated.

6 is a graph showing cross-verification results for prediction elements for each machine learning technique that can be applied in the learning unit 104 according to an embodiment of the present invention.

As shown in Figure 6 and Table 1 below, RF and SVM were the best in computational time performance, and terrain variable application sensitivity excelled in the order of SVM, RF, and GPRM, and spatial prediction error excelled in the order of GPRM, RF, and SVM. it happened Accordingly, in the embodiments of the present invention, GPRM was selected as an optimal machine learning technique, and the accuracy of each technique was verified using observation data collected from an external server as shown in FIG. 7 to be described later.

GPRM RF SVM calculation time 15 to 20 minutes within 1 minute within 1 minute Variable sensitivity lowest usually highest Spatial prediction error lowest usually highest

7 is a graph showing a comparison result with external observation data for each machine learning technique that can be applied in the learning unit 104 according to an embodiment of the present invention.

As an example, in the embodiments of the present invention, the accuracy of each machine learning technique was verified based on observation data (ie, weather data) collected from a server (not shown) of a KT mobile communication company.

As shown in FIG. 7 , it was found that the error between predicted weather information and observation data was the smallest when GPRM was used compared to other techniques.

Accordingly, the machine learning technique application unit 408 applies GPRM to the converted meteorological elements and the terrain variables corresponding to the first meteorological office neighborhood forecast to the first meteorological office neighborhood forecast information and the first meteorological office neighborhood forecast. It is possible to extract the correlation between the corresponding topographical variables. At this time, the machine learning technique application unit 408 determines the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information in order to shorten the time it takes to calculate the inverse matrix of the variance and covariance matrices, which occupies the largest portion of the calculation time. After reducing to a set ratio, the GPRM can be applied.

8 is a diagram illustrating a process of reducing the resolution of a sample space to a set ratio in the process of applying the GPRM in the learning unit 104 according to an embodiment of the present invention, and the calculation time performance accordingly.

As described above, the machine learning technique application unit 408 may apply the GPRM after reducing the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information by a set ratio, in this case, the process of applying the GPRM It is possible to reduce the time it takes to perform and the memory occupancy according to the application of GPRM.

Referring to (a) of FIG. 8 , the machine learning technique application unit 408 may apply the GPRM after reducing the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information to 1/4, for example. can After reducing the resolution of the sample space to 1/4, 1/2, etc., sensitivity experiments were conducted several times when GPRM was applied. As a result, the reduction in calculation time and the efficiency of maintaining accuracy were the highest in the 1/4 ratio. .

In addition, Met_R shown in FIG. 8(b) represents the detailed result using the RF (Random Forest) regression technique, and Met-G shown in FIG. 8(b) represents the detailed result using the GPRM technique, in FIG. The Met-GQ shown in (c) of 8 shows the detailed result using the HEG technique to which the above-described sample space resolution reduction is applied. Referring to (b) of FIG. 8 , it can be seen that the detailed result of Met-GQ, which has excellent calculation time performance, appears similar to that of Met-G.

Returning to FIG. 4 again, the machine learning technique application unit 408 applies the GPRM to the converted meteorological elements and the terrain variables corresponding to the first meteorological office neighborhood forecast, thereby providing the first meteorological office neighborhood forecast information and the first Extracting the correlation between the topographical variables corresponding to the regional forecast of the Korea Meteorological Agency, and refining the first regional forecast information of the Korea Meteorological Agency based on the extracted correlation to the scale of the first resolution to generate meteorological information for the scale of the first resolution can do.

As an example, the machine learning technique application unit 408 extracts a correlation between weather elements for 5 km resolution and a terrain variable corresponding to the corresponding weather elements, and then based on the extracted correlation, a local scale, for example, 1 km resolution. It is possible to predict/estimate weather information for That is, the machine learning technique application unit 408 uses the machine learning technique to refine the first meteorological office neighborhood forecast information on the scale of the second resolution to the scale of the first resolution smaller than the second resolution, It is possible to generate meteorological information about the scale.

9 is an example of weather information (ie, thermal environment application index) predicted by the learning unit 104 according to an embodiment of the present invention.

The 'grid' folder of FIG. 9 is an example of 1km grid-type weather information predicted by the learning unit 104, and the 'sigun' folder of FIG. 9 is an example of data obtained by averaging the 1km grid-type weather information in units of cities and counties.

As shown in FIG. 9 , it can be confirmed that the first meteorological office neighborhood forecast information having a resolution of 5 km is detailed as meteorological information having a resolution of 1 km by applying the above-described machine learning technique. In this case, the meteorological information includes, for example, the highest daily temperature on a 1km scale, the lowest daily temperature, the 3-hour unit humidity, the 3-hour unit temperature, the 3-hour unit wind speed, the daily average speed, the daily average temperature, the daily cumulative precipitation, etc. can do.

Returning to FIG. 4 again, the post-processing unit 410 uses the weather information to determine a Temperature Humidity Index (THI), a Wet Bulb Globe Temperature (WBGT), a sensible temperature, a daily cumulative precipitation, and an average by city and county. data, etc., can be calculated.

10 is a diagram for explaining a process of predicting weather information in the learning unit 104 according to an embodiment of the present invention.

Referring to FIG. 10 , as described above, the learning unit 104 may select set meteorological elements from among the first meteorological office neighborhood forecast information, and convert the selected meteorological elements into a set file format ( S102 ).

In addition, the learning unit 104 corresponds to the first meteorological office neighborhood forecast information and the first meteorological office neighborhood forecast by applying a machine learning technique to the converted weather elements and the terrain variable corresponding to the first meteorological office neighborhood forecast It is possible to extract the correlation between the terrain variables that become , S106). At this time, the city and county classification code data, urbanized agricultural land masking, livestock production land masking, orchard masking, field masking, neighborhood forecast coordinate information, output coordinate information, etc. shown in step S104 of FIG. 10 are one of the topographical variables in the correlation extraction process. It may be input as a parameter or input as additional additional information to be utilized.

In addition, the learning unit 104 may calculate a temperature-humidity index (THI), a heat index (WBGT), a sensible temperature, daily cumulative precipitation, average data for each city, and the like by using the weather information (S108). The average data for each city and county is, for example, divided into 167 city and county categories, and the maximum temperature, the minimum temperature, the average temperature, the average humidity, and the average wind speed information for each category can be calculated and converted.

Also, the learning unit 104 may verify the predicted weather information using observation data collected from an external server.

11 is a graph illustrating a process of verifying the accuracy of predicted weather information in the learning unit 104 according to an embodiment of the present invention.

11, the learning unit 104 includes observation data collected from a server (not shown) of KT mobile communication company, RMSE (Root-Mean-Square Deviation) collected from an agricultural meteorological observation network server (not shown), The predicted weather information may be verified using observation data such as correlation values, and a numerical value of the verification result may be converted into a database.

Returning to FIG. 1 , the weather information calculating unit 106 generates the second meteorological office neighborhood forecast information based on the association and weather information stored in the learning unit 104 when the second meteorological office neighborhood forecast information is collected. It is possible to refine the meteorological information for a scale of 1 resolution, and output the detailed meteorological information. That is, the meteorological information calculator 106 may collect the second meteorological office neighborhood forecast information at a predetermined time interval or every fixed date, and based on the result learned by the learning unit 104, the second meteorological office neighborhood forecast information can be detailed as weather information for the scale of the first resolution, and the detailed weather information can be output.

The second meteorological office neighborhood forecast information and detailed weather information may be input as input data in the process of applying the preceding machine learning technique, and by repeating this process, the learning model including the machine learning technique is updated or can be optimized.

According to embodiments of the present invention, by calculating detailed weather information from the statistical correlation between various local conditions, such as terrain, structures, vegetation influence, etc. according to the Meteorological Agency neighborhood forecast information and urban surface composition, after machine learning, life It can help to provide more accurate information in environment-related fields.

12 is a flowchart illustrating a method for calculating a thermal environment application index according to an embodiment of the present invention. In the illustrated flowchart, the method is described by dividing the method into a plurality of steps, but at least some of the steps are performed in a different order, are performed together in combination with other steps, are omitted, are performed in sub-steps, or are not shown. One or more steps may be added and performed.

In step S1002 , the ground surface analysis unit 102 builds a ground surface analysis model having a first resolution of a local scale set based on Geographic Information System (GIS) information.

In step S1004, the ground surface analysis unit 102 calculates a plurality of topographical variables affecting local weather by using the ground surface analysis model.

In step S1006, the learning unit 104 collects a plurality of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution.

In step S1008, the learning unit 104 applies a machine learning technique to the first meteorological office neighborhood forecast information and the terrain variable corresponding to the first meteorological office neighborhood forecast information to obtain the first meteorological office neighborhood forecast information and the The correlation between the topographical variables corresponding to the neighborhood forecast of the first Meteorological Agency is extracted.

In step S1010 , the learning unit 104 refines the first meteorological office neighborhood forecast information at the scale of the first resolution based on the extracted correlation to predict the weather information on the scale of the first resolution.

In step S1012, the meteorological information calculating unit 106 collects the second meteorological agency neighborhood forecast information.

In step S1014, the meteorological information calculating unit 106 refines the second meteorological office neighborhood forecast information into meteorological information with respect to the scale of the first resolution based on the correlation and weather information stored in the learning unit.

In step S1016, the weather information calculating unit 106 outputs the detailed weather information.

13 is a block diagram illustrating and describing a computing environment including a computing device suitable for use in example embodiments. In the illustrated embodiment, each component may have different functions and capabilities other than those described below, and may include additional components other than those described below.

The illustrated computing environment 10 includes a computing device 12 . In an embodiment, the computing device 12 may be the thermal environment application index calculation system 100 or one or more components included in the thermal environment application index calculation system 100 .

Computing device 12 includes at least one processor 14 , computer readable storage medium 16 , and communication bus 18 . The processor 14 may cause the computing device 12 to operate in accordance with the exemplary embodiments discussed above. For example, the processor 14 may execute one or more programs stored in the computer-readable storage medium 16 . The one or more programs may include one or more computer-executable instructions that, when executed by the processor 14, configure the computing device 12 to perform operations in accordance with the exemplary embodiment. can be

Computer-readable storage medium 16 is configured to store computer-executable instructions or program code, program data, and/or other suitable form of information. The program 20 stored in the computer readable storage medium 16 includes a set of instructions executable by the processor 14 . In one embodiment, computer-readable storage medium 16 includes memory (volatile memory, such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash It may be memory devices, other forms of storage medium accessed by computing device 12 and capable of storing desired information, or a suitable combination thereof.

Communication bus 18 interconnects various other components of computing device 12 , including processor 14 and computer readable storage medium 16 .

Computing device 12 may also include one or more input/output interfaces 22 and one or more network communication interfaces 26 that provide interfaces for one or more input/output devices 24 . The input/output interface 22 and the network communication interface 26 are coupled to the communication bus 18 . Input/output device 24 may be coupled to other components of computing device 12 via input/output interface 22 . Exemplary input/output device 24 may include a pointing device (such as a mouse or trackpad), a keyboard, a touch input device (such as a touchpad or touchscreen), a voice or sound input device, various types of sensor devices, and/or imaging devices. input devices, and/or output devices such as display devices, printers, speakers and/or network cards. The exemplary input/output device 24 may be included in the computing device 12 as a component constituting the computing device 12 , and may be connected to the computing device 12 as a separate device distinct from the computing device 12 . may be

Although the present invention has been described in detail through representative embodiments above, those of ordinary skill in the art to which the present invention pertains can make various modifications to the above-described embodiments without departing from the scope of the present invention. will understand Therefore, the scope of the present invention should not be limited to the described embodiments, and should be defined by the claims described below as well as the claims and equivalents.

100: Thermal environment application index calculation system
102: surface analysis unit
104: study department
106: weather information calculation unit
202: surface analysis input data collection unit
204: Surface analysis model construction unit
206: terrain variable calculation unit
402: Meteorological Agency neighborhood forecast information collection department
404: weather element selection unit
406: file conversion unit
408: machine learning technique application unit
410: post-processing unit

Claims (16)

Building a surface analysis model with a first resolution of a local scale set based on Geographic Information System (GIS) information, and calculating a number of topographical variables affecting local weather using the surface analysis model surface analysis unit;
A plurality of pieces of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution are collected, and the resolution of a sample space corresponding to the first meteorological office neighborhood forecast information is reduced by a set ratio. By applying a machine learning technique to the regional forecast information of the Korea Meteorological Agency and the topographical variable corresponding to the first regional forecast of the Korea Meteorological Agency, the correlation between the first meteorological office neighborhood forecast information and the topographical variable corresponding to the first meteorological office neighborhood forecast a learning unit that extracts and refines the first meteorological office neighborhood forecast information at the scale of the first resolution based on the extracted correlation to generate weather information on the scale of the first resolution; and
When the second meteorological office neighborhood forecast information is collected, the second meteorological office neighborhood forecast information is detailed based on the association and weather information stored in the learning unit into meteorological information for the scale of the first resolution, and the detailed weather information It includes a weather information calculation unit to output,
In the process of extracting the correlation, city and county classification code data, urbanization farmland classification code data, livestock district classification code data, orchard classification code data, and field classification code data are input as one parameter of the terrain variable, so that the scale of the first resolution Weather information for city and county units. It is calculated by averaging each city-gun unit of urbanized agricultural land, city-gun unit of livestock production area, city-gun unit of orchard, and city-gun unit of field.
The second Meteorological Agency neighborhood forecast information and the detailed weather information are input as input data in the process of applying the machine learning technique,
The second Meteorological Agency neighborhood forecast information and the detailed weather information is input as the input data is repeatedly performed, whereby the learning model including the machine learning technique is updated or optimized, a thermal environment application index calculation system.
The method according to claim 1,
The ground surface analysis unit converts the ground surface analysis input data including the numerical topographic map regarding the height and shape of the terrain, the numerical building map regarding the shape and height of the building, and the land cover map regarding the land use into a set coordinate system, and the converted A thermal environment application index calculation system for building the ground surface analysis model by constructing the input data as raster data having a first resolution of the local scale.
The method according to claim 1,
The terrain variables include the latitude of the terrain, the longitude of the terrain, the altitude of the terrain, the slope of the terrain, the distance from the shoreline of the terrain, the land cover of the terrain, the north-south or east-west slope of the terrain, the depression depth of the terrain, the slope of the terrain A thermal environment application index calculation system, comprising at least one of a length, a height of a building, a side area ratio of a building, a floor area ratio of a building, a rough length of a building, and an area ratio of a specific land cover.
The method according to claim 1,
The learning unit selects set meteorological elements from among the first meteorological office neighborhood forecast information, converts the selected meteorological elements into a set file format, and the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast A thermal environment application index calculation system that applies machine learning techniques to
5. The method according to claim 4,
The learning unit, a thermal environment application index calculation system for applying a Gaussian Process Regression Model (GPRM) to the topographic variables corresponding to the converted weather elements and the first meteorological office neighborhood forecast.
delete The method according to claim 1,
The set ratio is 1/4, the thermal environment application index calculation system.
The method according to claim 1,
The learning unit, using the observation data collected from an external server to verify the predicted weather information, a thermal environment application index calculation system.
Building a surface analysis model having a first resolution of a local scale set based on the GIS information in the surface analysis unit;
calculating, in the ground surface analysis unit, a plurality of topographic variables affecting local weather by using the ground surface analysis model;
collecting, in the learning unit, a plurality of first meteorological office neighborhood forecast information for a scale of a second resolution greater than the first resolution;
In the learning unit, machine learning is performed on the first meteorological office neighborhood forecast information and the terrain variable corresponding to the first meteorological office neighborhood forecast in a state in which the resolution of the sample space corresponding to the first meteorological office neighborhood forecast information is reduced by a set ratio. extracting a correlation between the first meteorological office neighborhood forecast information and a terrain variable corresponding to the first meteorological office neighborhood forecast by applying a technique;
generating, in the learning unit, meteorological information for the scale of the first resolution by refining the neighborhood forecast information of the first meteorological agency based on the extracted correlation to the scale of the first resolution;
When the second meteorological office neighborhood forecast information is collected in the meteorological information calculator, detailing the second meteorological office neighborhood forecast information into meteorological information for the scale of the first resolution based on the correlation and meteorological information stored in the learning unit ; and
and outputting the detailed weather information from the weather information calculating unit,
In the process of extracting the correlation, city and county classification code data, urbanization farmland classification code data, livestock district classification code data, orchard classification code data, and field classification code data are input as one parameter of the terrain variable, so that the scale of the first resolution Weather information for city and county units. It is calculated by averaging each city-gun unit of urbanized agricultural land, city-gun unit of livestock production area, city-gun unit of orchard, and city-gun unit of field.
The second Meteorological Agency neighborhood forecast information and the detailed weather information are input as input data in the process of applying the machine learning technique,
The second meteorological office neighborhood forecast information and the detailed weather information is input as the input data is repeatedly performed so that the learning model including the machine learning technique is updated or optimized, a thermal environment application index calculation method.
10. The method of claim 9,
The step of constructing the ground surface analysis model includes setting the ground surface analysis input data including the numerical topographic map regarding the height and shape of the terrain, the numerical building map regarding the shape and height of the building, and the land cover map regarding the land use into a set coordinate system. Converting, and constructing the ground surface analysis model by constructing the converted input data as raster data having a first resolution of the local scale, a thermal environment application index calculation method.
10. The method of claim 9,
The terrain variables include the latitude of the terrain, the longitude of the terrain, the altitude of the terrain, the slope of the terrain, the distance from the shoreline of the terrain, the land cover of the terrain, the north-south or east-west slope of the terrain, the depression depth of the terrain, the slope of the terrain A method of calculating a thermal environment application index, including at least one of length, height of a building, ratio of side area of a building, ratio of floor area of building, length of roughness of building, and area ratio of specific land cover.
10. The method of claim 9,
The step of extracting the correlation includes selecting set meteorological elements from among the first meteorological office neighborhood forecast information, converting the selected meteorological elements into a set file format, and corresponding to the converted meteorological elements and the first meteorological office neighborhood forecast A method of calculating a thermal environment application index by applying a machine learning technique to the terrain variable being
13. The method of claim 12,
The extracting of the correlation comprises applying the GPRM to the converted meteorological elements and the terrain variable corresponding to the first meteorological office neighborhood forecast, a thermal environment application index calculation method.
delete 10. The method of claim 9,
The set ratio is 1/4, the thermal environment application index calculation method.
10. The method of claim 9,
After predicting and storing the weather information,
In the learning unit, the method further comprising the step of verifying the predicted weather information using the observation data collected from the external server, thermal environment application index calculation method.

Images