KR102169452B1 - METHOD FOR ENSURING STABILITY OF DATA COLLECTED IN IoT WEATHER ENVIRONMENT - Google Patents

Abstract

The present invention is a technology for securing the safety of data collected in the IoT meteorological environment.When the meteorological data collected through the IoT sensor is required to generate alternative data due to equipment and network failure, or communication failure, an error part is recorded in the collected data. It is a technology to solve when it adversely affects analysis and utilization. All collected data goes through quality control (QC) to improve the quality of the data, and after generating learning data through the improved data, it is reflected in machine learning algorithms to detect errors in raw weather data, and data determined as errors. The present invention relates to a method of correcting with an estimate using machine learning.

Description

How to secure the stability of data collected from the IoT weather environment{METHOD FOR ENSURING STABILITY OF DATA COLLECTED IN IoT WEATHER ENVIRONMENT}

The present invention relates to a method of securing the stability of data collected in an IoT meteorological environment, and more particularly, an error existing in the meteorological observation data by performing quality control (QC) on the weather observation data collected using an IoT sensor. It relates to a method of securing the stability of the collected data of the IoT meteorological environment to obtain refined weather observation data by searching and filtering.

In the sensed data received from a plurality of weather sensors, various error patterns exist according to the sensor and the observation environment, and there is a limit to error detection with the existing simple quality control technique.

In particular, there are cases in which the error pattern does not show a significant difference from normal data, and the detection data rapidly increases due to the spread of Internet of Things (IoT) technology, so there is a limit to quality control by humans.

In other words, it is not possible to determine whether there is an abnormality in the sensing data only with the numerical value for the mechanical error, there are various error situations for the sensing data within the normal range, and there may also be a normal measurement situation similar to the error. In some cases, it is impossible to accurately determine whether there is an abnormality.

In order to solve such a problem, weather through a complex judgment that complexly judges whether there is an error based on a plurality of error analysis methods classified into a plurality of error analysis groups according to the error judgment reliability of the detected data measured from the weather sensor. There is a way to detect sensor abnormalities.

This method adds daily change pattern analysis, probability distribution analysis, frequency analysis, correlation analysis, etc. to the general QC method, which is an error analysis method, and removes error data by complexly judging individual analysis results or determines the data grade for the detected data. It is classified and provided so that the detection data of the grade suitable for the purpose of use can be used.

However, such a method is said to judge individual analysis results in a complex manner, but it is not possible to selectively apply one more suitable for the meteorological data among multiple quality control (QC) methods in consideration of the characteristics of each of the various factors of meteorological data. When there are many errors in the meteorological data collected by the IoT meteorological environment collection sensor, it is not possible to perform quality control (QC) using the meteorological data collected from another IoT meteorological environment collecting sensor adjacent to it. And there is a problem that the accuracy of the correction value of the weather data determined as an error is poor.

Registered Patent No. 10-1519012 (Weather information error verification system and method)

The present invention is to solve the problem of the prior art, and after performing basic QC to check for errors by performing basic quality control (QC) for each meteorological factor, according to the user's selection or Not only the first quality control (QC) using only individual factors for each object item automatically according to the characteristics, and the second quality control (QC) using only the individual factors for each object item and meteorological data from one IoT meteorological environment collecting sensor By selecting one or more of spatial quality control (QC) using meteorological data from several neighboring IoT meteorological environment collecting sensors and performing quality control (QC), it is possible to create and create a learning model suitable for the characteristics of meteorological factors. To provide a method of securing the stability of IoT meteorological environment data that can be applied to machine learning algorithms to detect errors in raw weather data, and correct weather data determined as errors with an estimate using machine learning. will be.

In order to achieve the above object, the method of securing stability of the IoT meteorological environment collection data of the present invention includes: a collecting step of obtaining, by a collecting unit, meteorological data from an IoT meteorological environment collecting sensor; A preprocessing step of processing and converting the meteorological data collected by the preprocessor according to the type of the collection sensor; A basic QC step of checking whether or not there is an error by performing basic quality control (QC) on the meteorological data preprocessed by the verification unit for each meteorological factor; The verification unit uses meteorological data whose quality has been improved through the basic QC step as learning data, and the first quality control (QC) using only individual factors for each object item and a related factor linking different meteorological factors for each object item. 2 Quality management (QC) by selecting one or more of quality control (QC) and spatial quality control (QC) using meteorological data from several neighboring IoT meteorological environment collecting sensors, as well as meteorological data from one IoT meteorological environment collecting sensor. A machine learning QC step of generating learning data by performing ); The verification unit calculates the standard deviation (σ) using meteorological data measured in the past for a certain period of time, and uses the standard deviation (σ) to determine and correct the error in the learning data value generated in the machine learning QC step. Correcting the learning data; And the prediction unit applies the learning data generated through the learning data correction step to a machine learning algorithm to detect an error in raw weather data obtained from the IoT meteorological environment collection sensor, and convert the weather data determined as an error into an estimate using machine learning. It includes a prediction step of correcting.

In addition, the basic QC step performs basic quality control including a physical limit test, a time variability/step test, a persistence test, and an internal consistency test, and the physical limit test deviates from the upper and lower limit values set for each meteorological factor. If the difference between the value 1 minute ago and the current value for each meteorological factor exceeds the threshold value, it is determined as an error, and the time variability/step test is determined as an error. If the amount of fluctuation is less than a certain limit value, it is determined as an error, and the internal consistency test determines the error based on the coincidence of a pair of weather factors that are related to each other. And internal consistency tests are sequentially performed, and if all are satisfied, it is determined as normal.

In addition, in the machine learning QC step, the verification unit checks the error rate in the training data that has passed the basic QC step, and when it is determined that it exceeds a preset error rate, selects and performs spatial quality management (QC). .

In addition, in the machine learning QC step, spatial quality management (QC) is performed using meteorological data from another IoT meteorological environment collecting sensor belonging to a certain distance based on the location of the IoT meteorological environment collecting sensor. Quality control is performed by considering the terrain characteristics near the collection sensor and the terrain characteristics near other IoT weather environment collection sensors as variables, and among the terrain characteristics near other IoT weather environment collection sensors, the terrain characteristics near the IoT weather environment collection sensor Meteorological data from IoT meteorological environment collection sensors that show a difference of more than a certain value in the degree of similarity to and topographical are excluded from performing spatial quality control (QC).

Furthermore, the topography including mountains, rivers, fields, and buildings near each IoT meteorological environment collection sensor is scored using distance and height information away from the IoT meteorological environment collection sensor, and the topographical characteristics are compared. Calculate whether phosphorus similarity shows a difference of more than a certain value.

And, in the training data correction step, the verification unit calculates the standard deviation (σ) using meteorological data measured for a certain time in the past, and subtracts 3σ from the average value of the meteorological data measured for a certain time and for a certain time. If the learning data value generated in the machine learning QC step exists between the average value of the measured meteorological data plus 3σ, it is regarded as normal, and if it is out of this range, it is determined as an error, and the learning data value determined as an error Compensate by comparing with the nearest time value.

According to the method of securing the stability of the IoT meteorological environment data collected through the IoT sensor according to the present invention, when there is an error due to equipment and network failure or communication failure in the meteorological data collected through the IoT sensor, it is possible to correct it with an estimate using machine learning, It can increase the reliability of the analysis and use of weather observation data.

And, depending on the characteristics of the meteorological factors, one of the first quality control (QC) using only individual factors for each object item and the second quality control (QC) using a linkage factor linking different meteorological factors for each object item is selected, or Both can be selected and applied, and if it is determined that the error rate is high through the basic QC, the meteorological data of several IoT meteorological environment collecting sensors located nearby is used instead of the first quality control and the second quality control. Since spatial quality management (QC) can be selected, optimal learning data can be generated through quality management that comprehensively considers the characteristics of meteorological factors and the ratio of errors.

After all, by applying highly reliable learning data to a machine learning algorithm, errors in raw weather data can be accurately corrected.

1 is a flowchart of a method of securing stability of collected data of an IoT meteorological environment according to the present invention
2 is a flow chart of a basic QC method applied to the method of securing stability of collected data of an IoT meteorological environment according to the present invention
3 is a flowchart of a method of generating learning data using machine learning QC applied to the method of securing stability of IoT meteorological environment collection data according to the present invention, and determining and correcting the error of the generated learning data
4 is a system configuration diagram for implementing a method of securing stability of collected data of an IoT meteorological environment according to the present invention

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to embodiments to be described later in detail together with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various different forms.

In the present specification, the present embodiment is provided to complete the disclosure of the present invention, and to completely inform the scope of the invention to those of ordinary skill in the art to which the present invention belongs.

And the invention is only defined by the scope of the claims.

Thus, in some embodiments, well-known components, well-known operations, and well-known techniques have not been described in detail in order to avoid obscuring interpretation of the present invention.

In addition, throughout the specification, the same reference numerals refer to the same constituent elements, and terms used in the present specification (referred to) are intended to describe embodiments and are not intended to limit the present invention.

In this specification, the singular form also includes the plural form unless specifically stated in the phrase, and the components and actions referred to as'include (or, have)' do not exclude the presence or addition of one or more other components and actions. .

Unless otherwise defined, all terms (including technical and scientific terms) used in the present specification may be used as meanings that can be commonly understood by those of ordinary skill in the art to which the present invention belongs.

In addition, terms defined in a commonly used dictionary are not interpreted ideally or excessively unless defined.

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

It is important to provide accurate information for meteorological data collected using IoT sensors, and it is necessary to provide corrections for information determined as errors.

In other words, it is necessary to collect meteorological data that has been properly quality control (QC). Quality Control (QC) is the task of identifying defects and post-correction that exist in previously collected meteorological data. Performing QC on the meteorological data collected from the IoT meteorological environment collection sensor means performing the task of obtaining refined data by finding and filtering errors that exist in the collected weather observation data.

1 to 4, in the method of securing stability of the IoT meteorological environment collection data according to the present invention, first, the collecting unit 410 collects meteorological data from the IoT meteorological environment collecting sensor (S110). The collected meteorological data includes temperature, wind direction, wind speed, rainfall detection, humidity, ozone, ultraviolet rays, PM2.5, and solar radiation.

The pre-processing unit 420 processes and converts the collected meteorological data according to the type of the collection sensor (S120). For example, in the case of insolation data, the accumulated insolation data must be converted into 1 minute insolation data. And, wind direction and wind speed data are generated as independent variables using Radian transform and u and v transform.

The verification unit 430 performs basic QC to check whether there is an error by performing basic quality control (QC) on the meteorological data preprocessed by the preprocessor 420 for each meteorological factor (S130).

A detailed look at the basic QC steps performed by the verification unit 430 are as follows.

Basic quality control including physical limit inspection, temporal variability/step inspection, continuity inspection, and internal consistency inspection is performed step by step, and errors are checked.

First, the physical limit test (S210) determines as an error when it exceeds the upper and lower limit values set for each meteorological factor.

Physical limit test reference value Sensor type Lower limit Upper limit Temperature -40 60 Wind direction 0 360 Wind speed 0 75 Rainfall detection 0 One Humidity 0 100 Ilsa 0 1500 PM2.5 0.1 359.9

The time variability/step check (S220) determines as an error when the difference between the value 1 minute ago and the current value for each meteorological factor exceeds the threshold value. It targets temperature, air pressure, temperature, wind speed, and humidity. If the difference between the value 1 minute ago and the current value exceeds the threshold, it is judged as an error.

Maximum variation per factor Sensor type Maximum variation Temperature 0.1 Wind speed One Humidity One atmospheric pressure 0.2

The persistence check (S230) determines as an error when the variation of the target factor is less than a certain limit value for 60 minutes or more for each meteorological factor. It targets wind speed, temperature, humidity, and air pressure, and if the variation of the target factor for more than 60 minutes is less than a certain threshold, it is judged as an error.

Minimum variation per factor Sensor type Minimum variation Temperature 0.1 Wind speed 0.5 Humidity One atmospheric pressure 0.1

The internal consistency test (S240) determines an error based on whether a pair of weather factors that are related to each other are matched. In the example of Table 4, wind direction and wind speed, rainfall detection, and 1-hour cumulative rainfall were tested. If either of the wind direction and wind speed is 0 and the other factor is not 0, both are treated as errors. If the rainfall detection is 1 but the rainfall is 0, or the rainfall detection is 0 and the rainfall is greater than 0, it is treated as an error.

Criteria for determining internal consistency Sensor 1 Sensor 2 Criteria Wind direction Wind speed Error if either of them is 0 Rainfall detection fall Rainfall detection = 1, rainfall = 0 or
Rainfall detection = 0, if rainfall> 0, error

A physical limit test (S210), a time variability/step test (S220), a persistence test (S230), and an internal consistency test (S240) are sequentially performed, and if all are satisfied, it is determined as normal, and any one step in the process of performing the sequentially Even if it is an error, it is determined as an error, written as -999 (error value), and corrected in a later step.

The verification unit 430 uses the meteorological data with improved quality through the basic QC step (S130) as training data, and uses only individual factors for each object item, and a first quality control (QC) and different meteorological factors for each object item. Spatial quality control (QC) using meteorological data collected from several neighboring IoT meteorological environment collection sensors, as well as the second quality control (QC) using the linkage factor linked to and meteorological data collected from one IoT meteorological environment collection sensor. Machine learning QC for generating learning data is performed by selecting one or more of them and performing quality control (QC) (S140).

For example, if an IoT meteorological environment collecting sensor installed in a certain area has only a temperature sensor, and there is only temperature observation data among the meteorological data, first quality control (QC) using only individual factors for each object item can be performed, but other than temperature Since there is no meteorological data, second quality control (QC) using a linkage factor linking different meteorological factors for each object item will not be possible.

However, if there is a temperature sensor and a humidity sensor in the IoT meteorological environment collecting sensor installed in another area away from the corresponding area, the temperature data collected by the IoT meteorological environment collecting sensor installed in the area and the IoT weather environment collecting sensor installed in the other area are collected. Spatial quality control (QC) can also be performed using one temperature data.

As another example, if an IoT meteorological environment collecting sensor installed in a certain area has a temperature sensor and a humidity sensor to collect temperature data and humidity data, the first quality control (QC) using only individual factors for each object item using temperature data ), and second quality control (QC) using a linkage factor by linking temperature and humidity.

Of course, if there is a temperature sensor and a humidity sensor in the IoT meteorological environment collecting sensor installed in another area away from the corresponding area, the temperature data collected from the IoT meteorological environment collecting sensor installed in that area and the IoT weather environment collecting sensor installed in the other area are collected. It is natural that spatial quality control (QC) can also be performed using one temperature data.

In this way, depending on the type of meteorological sensor installed in the IoT meteorological environment collection sensor, first quality control (QC), second quality control (QC), and spatial quality control (QC) are applied one by one, or one or two of them. By selecting and applying, machine learning QC to generate learning data is performed (S140).

3, a first quality control (QC) (S310) using only individual factors for each object item and a second quality control (QC) (S320) using a linkage factor linking different meteorological factors for each object item Quality control is performed by selecting one or more of spatial quality management (QC) (S330) using not only the meteorological data of the IoT meteorological environment collecting sensor but also the meteorological data of several neighboring IoT meteorological environment collecting sensors. The operator can choose any one or more of these three quality control (QC), or the system can choose automatically based on existing histories. When learning by selecting two or more quality management algorithms, the data learned first are used and then applied to the quality management algorithm.

The first quality control (QC) (S310) applies quality control using only individual factors. For example, it is possible to create learning data that predicts the current temperature value by inputting and learning the temperature value 10 minutes ago in the support vector regression (SMOreg, math-based) algorithm, which is a machine learning algorithm.

The second quality control (QC) (S320) applies quality control using not only individual factors but linked factors. For example, it is possible to create learning data that predicts the current temperature value by inputting and learning the temperature value 10 minutes ago, the current wind direction, and wind speed values in the support vector regression (SMOreg, math-based) algorithm, which is a machine learning algorithm. .

Spatial quality management (QC) (S330) applies quality management using not only the meteorological data of one IoT meteorological environment collecting sensor, but also meteorological data of several neighboring IoT meteorological environment collecting sensors. As a result of performing the basic QC (S130), it is preferable to select and apply spatial quality management (QC) when a large number of error values occur.

The verification unit 430 checks the error rate in the training data that has gone through the basic QC step (S130), and if it is determined that it exceeds the preset error rate, select and perform spatial quality management (QC) (S330). I can. That is, as a result of performing basic quality control on the meteorological data collected by the IoT meteorological environment collection sensor, if the error value described as -999, -998, -997 in temperature and humidity values, etc. exceeds 10%, Learning data can be obtained by learning using meteorological data measured by IoT weather environment collection sensors.

Specifically, spatial quality control (QC) is performed using weather data from other IoT weather environment collection sensors belonging to a certain distance based on the location of the IoT weather environment collection sensor, but the terrain near the IoT weather environment collection sensor Quality management is performed by considering the characteristics and the terrain characteristics of the IoT weather environment collection sensor as a variable. Among the terrain characteristics of the IoT weather environment collection sensor, the geographical similarity to the terrain characteristics near the IoT weather environment collection sensor IoT meteorological environment collecting sensor meteorological data showing a difference of more than a certain number are excluded from performing spatial quality control (QC).

Terrain including mountains, rivers, fields, and buildings near each IoT meteorological environment collection sensor is scored using distance and height information away from the IoT meteorological environment collection sensor, and topographical similarity is achieved by comparing scores for geographical characteristics. Calculate whether the difference is more than a certain number.

The terrain similarity calculation unit (not shown) calculates the score using the presence or absence of mountains, rivers, fields, and buildings within a radius of 1 to 2 km of the IoT meteorological environment collection sensor, and the height of the mountains and buildings. Of course, there are no high terrain conditions such as mountains and buildings, but when there are fields and rivers, different terrain characteristics must be considered as variables when performing spatial quality control (QC).

For example, if there are mainly roads and buildings within 1-2km radius of the IoT weather environment collection sensor, score these, and if there are buildings and mountains within 1~2km radius of other adjacent IoT weather environment collection sensors, score them. Another adjacent IoT weather environment collection sensor is scored when there is a field within a radius of 1 to 2 km. It is possible to score by using the presence and height information of buildings, mountains, rivers, etc. that belong to each grid of topographic data received from public institutions.

The score for the geographical feature near the IoT meteorological environment collection sensor is 82 points, the score for the geographical feature near another IoT weather environment collection sensor is 70, and the geographical feature near another adjacent IoT meteorological environment collection sensor Assuming that the score for is 48 points, compare 82 points and 70 points, and compare 82 points and 48 points. Since there is a difference of 34 points for 82 points and 48 points, it is excluded from performing spatial quality control (QC) in the case where the similarity differs by more than 15-20 points.

Spatial quality control (QC) is performed using the meteorological data of another IoT meteorological environment collecting sensor corresponding to 70 points, but the distribution and height of the roads and buildings, which are geographical features near the IoT meteorological environment collecting sensor, are variables In addition, it is possible to reduce errors in basic QC by generating learning data by taking into account the distribution and height of buildings and mountains, which are the topographic characteristics of nearby other IoT weather environment collection sensors, as variables.

As described above, learning data is generated by selectively performing at least one of individual QC for each entity item (S310), linkage QC for each entity item (S320), and spatial QC (S330) (S340).

The verification unit S430 calculates a standard deviation (σ) using meteorological data measured in the past for a certain period of time, and uses the standard deviation (σ) for the training data value generated in the machine learning QC step (S140). The error is determined and corrected (S150).

Referring to FIG. 3, the verification unit (S430) calculates the standard deviation (σ) using meteorological data measured for a certain time in the past (S350), and subtracts 3σ from the average value of the meteorological data measured for a certain time. It is checked whether the learning data value generated in the machine learning QC step (S140) exists between the value obtained by adding 3σ from the average value of the meteorological data measured for a predetermined time (S360).

If the learning data value S340 generated in the machine learning QC step S140 exists between the corresponding range, it is regarded as normal (S370), and out of this range, it is determined as an error (S380), and the learning data value determined as an error With respect to the value of the nearest time and corrected (S390). Here, the value of the nearest time may correspond to a value 1 minute before or 1 minute later.

The prediction unit (S440) applies the learning data generated through the learning data correction step (S150) to a machine learning algorithm to detect an error in the raw weather data obtained from the IoT meteorological environment collection sensor, and machine the weather data determined as an error. A prediction step of correcting the estimated value using learning is performed (S160).

As the machine learning algorithm, any one of support vector regression (SMOreg, math-based), decision table (DecisionTable, rule-based), and multilayer perceptron (neural network) can be used, and in the test of the present invention, support vector regression ( It was confirmed that the SMOreg, math-based) algorithm showed the most suitable result.

According to the present invention, one of the first quality control (QC) using only individual factors for each object item and the second quality control (QC) using a linkage factor linking different meteorological factors for each object item is selected according to the characteristics of the meteorological factors. Alternatively, or both can be selected and applied, and if it is determined that the error rate is high through the basic QC, the weather of several IoT meteorological environment collection sensors located in the vicinity rather than the first quality control and the second quality control. Since spatial quality control (QC) using data can be selected, optimal learning data can be generated through quality control that comprehensively considers the characteristics of meteorological factors and the ratio of errors.

The present invention is not limited to the specific preferred embodiments described above, and any person having ordinary knowledge in the technical field to which the present invention pertains without departing from the gist of the present invention claimed in the claims can implement various modifications Of course, it is obvious that such a change will fall within the scope of the description of the claims.

410: collection unit
420: pretreatment unit
430: verification unit
440: prediction unit

Claims (6)

A collecting step of obtaining meteorological data from an IoT meteorological environment collecting sensor by a collecting unit;
A preprocessing step of processing and converting the meteorological data collected by the preprocessor according to the type of the collection sensor;
A basic QC step of checking whether or not there is an error by performing basic quality control (QC) on the meteorological data preprocessed by the verification unit for each meteorological factor;
The verification unit uses meteorological data whose quality has been improved through the basic QC step as learning data, and the first quality control (QC) using only individual factors for each object item and a related factor linking different meteorological factors for each object item. 2 Quality control by selecting one or more of quality control (QC) and spatial quality control (QC) using meteorological data obtained from several neighboring IoT meteorological environment collection sensors, as well as meteorological data obtained from one IoT meteorological environment collection sensor. Machine learning QC step of generating learning data by performing (QC)-If there is only temperature observation data among the meteorological data collected from the IoT meteorological environment collecting sensor installed in a certain area, the first quality control (QC) using only individual factors for each object item ), but since there is no other weather data other than temperature, second quality control (QC) using a linkage factor that links different meteorological factors for each object item is not possible, and an IoT weather environment installed in a different area away from the corresponding area If the collection sensor has a temperature sensor and a humidity sensor, spatial quality control (QC) can also be performed using the temperature data collected from the IoT meteorological environment collection sensor installed in the area and the temperature data collected from the IoT weather environment collection sensor installed in another area. Can be
If the IoT meteorological environment collection sensor installed in a certain area has a temperature sensor and a humidity sensor, so that temperature data and humidity data can be collected, first quality control (QC) can be performed using only individual factors for each object item using temperature data. If there is a temperature sensor and a humidity sensor in the IoT meteorological environment collection sensor installed in an IoT weather environment collection sensor installed in another area away from the area, IoT installed in the area can also perform second quality control (QC) using a linkage factor by linking temperature and humidity. Spatial quality control (QC) can also be performed by using the temperature data collected by the meteorological environment collecting sensor and the temperature data collected by the IoT meteorological environment collecting sensor installed in another area
The verification unit calculates the standard deviation (σ) using meteorological data measured in the past for a certain period of time, and uses the standard deviation (σ) to determine and correct the error in the learning data value generated in the machine learning QC step. Correcting the learning data; And
The prediction unit applies the learning data generated through the learning data correction step to a machine learning algorithm to detect errors in raw weather data obtained from the IoT meteorological environment collection sensor, and corrects the weather data determined as errors to an estimate using machine learning. Including the predicting step,
The machine learning QC step,
The verification unit checks the error rate in the training data that has gone through the basic QC step, and if it is determined that it exceeds the preset error rate, selects and performs spatial quality control (QC), but is collected by the IoT weather environment collection sensor. As a result of performing basic quality control on one meteorological data, if the error value in the temperature and humidity values exceeds a certain percentage, learning data is generated by learning using the meteorological data measured by other IoT meteorological environment collecting sensors nearby,
The machine learning QC step,
Spatial quality control (QC) is performed using weather data from other IoT weather environment collecting sensors belonging to a certain distance based on the location of the IoT weather environment collecting sensor, but different from the terrain characteristics of the IoT weather environment collecting sensor. Quality control is performed by considering the terrain characteristics near the meteorological environment collection sensor as a variable, and among the terrain characteristics near the IoT meteorological environment collection sensor, the difference between the terrain characteristics and the terrain similarity near the IoT meteorological environment collection sensor is more than a certain number. The meteorological data of the IoT meteorological environment collecting sensor that shows are excluded from performing spatial quality control (QC),
Terrain including mountains, rivers, fields, and buildings near each IoT meteorological environment collection sensor is scored using distance and height information away from the IoT meteorological environment collection sensor, and topographical similarity is achieved by comparing scores for geographical characteristics. Calculate whether the difference is more than a certain number,
The terrain similarity calculation unit scores roads and buildings within a certain range of the IoT meteorological environment collection sensor radius, scores buildings and mountains within a certain range of other adjacent IoT weather environment collection sensors, and scores other adjacent IoT weather environment collection sensors. If there is a field within a certain radius of the IoT meteorological environment collection sensor, it is scored, and it is scored using the presence and height information of buildings, mountains, rivers, etc. that fall within each grid of terrain data received from public institutions.
The score for the geographical feature near the IoT meteorological environment collection sensor is A, and the score for the geographical feature near the other adjacent IoT weather environment collection sensor is less than the above A point, and another adjacent IoT weather environment collecting sensor Assuming that the score for the nearby topographical feature is less than the point B, points A and B are compared, points A and C are compared, and the difference between points A and C is a predetermined score. Excluding from performing spatial quality control (QC), when the degree of similarity differs by more than a certain score because of the above difference,
Spatial quality control (QC) is performed using the weather data of another IoT weather environment collection sensor adjacent to point B, but the distribution and height of roads and buildings, which are the topographical characteristics of the IoT weather environment collection sensor, are used as variables. In addition, by generating learning data by considering the distribution and height of buildings and mountains, which are the topographic characteristics of nearby other IoT weather environment collection sensors, as variables, the error of basic QC is reduced. How to ensure stability.
The method of claim 1,
The basic QC step,
Perform basic quality control including physical limit inspection, time variability/step inspection, continuity inspection, and internal consistency inspection,
In the physical limit test, if it exceeds the upper and lower limit values set for each meteorological factor, it is judged as an error.
The time variability/step test is judged as an error if the difference between the value 1 minute ago and the current value for each meteorological factor exceeds the threshold value,
In the persistence test, if the variation of the target factor for more than 60 minutes for each meteorological factor is less than a certain threshold, it is determined as an error,
In the internal consistency test, an error is determined based on a match for a pair of weather factors that are related to each other.
The physical limit test, time variability/step test, persistence test, and internal consistency test are sequentially performed, and if all are satisfied, it is determined as normal, and if any step is an error in the sequential process, it is determined as an error value. The method of securing the stability of the collected data of the IoT meteorological environment, characterized in that described as and corrected in a later step.
delete delete delete The method of claim 1,
The learning data correction step,
The verification unit calculates the standard deviation (σ) using the meteorological data measured for a certain time in the past, subtracts 3σ from the average value of the meteorological data measured for a certain time, and 3σ from the average of the meteorological data measured for a certain time. If the learning data value generated in the machine learning QC step exists between the added values, it is regarded as normal, and if it is out of this range, it is determined as an error, and the learning data value determined as an error is compared with the value of the nearest time and corrected. A method of securing the stability of data collected in an IoT meteorological environment, characterized in that.

Images