KR102493171B1 - Method and system for anomaly detection of IoT meteorological sensor based on integration of sensor data - Google Patents

Abstract

The present invention relates to a method and system for detecting an abnormality in a sensor data fusion-based IoT weather sensor device, and more particularly, by convergence of various weather sensor data received from IoT (Internet of Things) devices in a smart city environment, IoT weather sensor It relates to a method and system for detecting whether a device is abnormal.
According to the present invention, specific weather sensor data of an IoT device performing abnormal detection, other weather sensor data of the corresponding IoT device, weather sensor data of other nearby IoT devices, and past weather sensor data previously generated and stored in a server In combination, it provides a method for optimally determining whether the weather sensor device is abnormal by calculating and collecting abnormal numerical values of the weather sensor data without being affected by the specific value according to the location where the IoT device is installed and the time when the sensor data is measured. do.

Description

Method and system for anomaly detection of IoT weather sensor device based on sensor data convergence {Method and system for anomaly detection of IoT meteorological sensor based on integration of sensor data}

The present invention relates to a method and system for detecting an abnormality in a sensor data fusion-based IoT weather sensor device, and more particularly, by convergence of various weather sensor data received from IoT (Internet of Things) devices in a smart city environment, IoT weather sensor It relates to a method and system for detecting whether a device is abnormal.

The IoT device contains sensors that can measure five meteorological elements (temperature, air pressure, wind direction, wind speed and humidity). IoT devices are installed at fixed locations that are already known at regular distances. Anomaly detection of a weather sensor is to determine whether a specific value measured by each weather sensor is a normal value or an abnormal value. The abnormality detection of the weather sensor device is to determine whether or not the weather sensor device is abnormal using values measured by each weather sensor.

In this way, when determining whether the meteorological sensor data measured by the IoT device is abnormal, it may include a singular value according to the location where the IoT device is installed and the time when the sensor data is measured, so a general filter cannot be used. For this reason, there is a need for a new technique to determine the abnormal state of a weather sensor device by calculating and collecting anomalies through weather sensor data convergence.

KR 10-1874085 B1

The present invention was conceived to solve this problem, and the specific weather sensor data of an IoT device that performs anomaly detection, other weather sensor data of the IoT device, and weather sensor data of other nearby IoT devices and previously generated server By combining the past weather sensor data stored in the IoT device, it is not affected by the specific value according to the place where the IoT device is installed and the time when the sensor data is measured, and the abnormal numerical value of the weather sensor data is calculated and collected to determine the abnormality of the weather sensor device. Its purpose is to provide a method for optimally determining whether

In order to achieve the above object, a sensor data convergence-based IoT weather sensor device anomaly detection method according to the present invention includes: (a) receiving weather sensor data measured by a plurality of weather sensors from each IoT device; (b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; And, (c) including the step of storing in a database whether there is an abnormality for each weather sensor of each IoT device, and before step (b), from (b0) the database, the weather sensor received from each IoT device in the past Further comprising the step of reading the data, wherein step (b) is, (b3) the IoT device that is currently performing the anomaly detection (hereinafter referred to as the 'corresponding IoT device') for each IoT device performing the anomaly detection. Find a time point in the past data at which the meteorological elements currently performing anomaly detection (hereinafter referred to as 'corresponding weather elements') of nearby IoT devices (hereinafter referred to as 'surrounding IoT devices') coincide, and Performing an abnormal detection of a corresponding meteorological element value of the IoT device from the corresponding meteorological element value of the IoT device; (b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device; (b5) Find the point in time when the current value and the past value of other meteorological factors of the surrounding IoT device coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and, (b6) find a point in time when the current value and past value of other meteorological factors of the IoT device and the current value and past value of all meteorological factors of the surrounding IoT device all match from the past data, and the corresponding IoT device at that time Including one or more steps of performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the corresponding meteorological element value, and between the step (b) and the step (c), (b7) the step (b3), Based on the abnormality detection results in the steps included among steps (b4), (b5), and (b6), the step of finally determining whether or not the corresponding meteorological element of the corresponding IoT device is abnormal according to a predetermined criterion contains more

delete

delete

If the IoT device currently performing anomaly detection is IoT k and the meteorological factor currently performing anomaly detection in the IoT k is A, in step (b3), for the meteorological element A measured by the IoT k Anomaly detection is performed when (b31) the value of the meteorological factor A currently measured by each IoT device around the IoT k and the value of the meteorological factor A measured in the past by each IoT device around the IoT k match for each IoT device. Searching for a past point in time; (b32) extracting the values of the meteorological factor A at the time points retrieved in the step (b31) as sample data; and (b33) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data.

If the IoT device currently performing anomaly detection is IoT k and the meteorological element currently performing anomaly detection in the IoT k is A, in step (b4), for the meteorological element A measured by the IoT k Anomaly detection is performed by (b41) the values of other meteorological factors other than the meteorological factor A currently measured by the IoT k and the values of other meteorological factors other than the meteorological factor A measured in the past by the IoT k, Searching for past points in time that match each element; (b42) extracting the values of the meteorological factor A at the time points retrieved in the step (b41) as sample data; and (b43) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data.

If the IoT device currently performing anomaly detection is IoT k and the meteorological element currently performing anomaly detection in IoT k is A, in the step (b5), for the meteorological element A measured by the IoT k Anomaly detection is performed by (b51) the values of other meteorological factors other than the meteorological factor A, currently measured by each IoT device around the IoT k, and the meteorological factors measured in the past by each IoT device around the IoT k. Retrieving a past point in time at which values of other meteorological factors, except for factor A, match each meteorological factor of each neighboring IoT device; (b52) extracting, as sample data, the values of the meteorological factor A measured by the IoT k at the time points retrieved in the step (b51); and (b53) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data.

If the IoT device currently performing anomaly detection is IoT k and the meteorological element currently performing anomaly detection in the IoT k is A, in step (b6), for the meteorological element A measured by the IoT k Anomaly detection is performed by (b61) values of other meteorological factors other than the meteorological factor A currently measured by the IoT k and all meteorological factors including the meteorological factor A currently measured by each IoT device around the IoT k. values, values of other meteorological factors other than the meteorological factor A, measured in the past by the IoT k, and values of all meteorological factors, including the meteorological factor A, measured in the past by each IoT device around the IoT k , Searching for a past point in time that matches each meteorological element of IoT k and each neighboring IoT device; (b62) extracting, as sample data, the values of the meteorological factor A measured by the IoT k at the time points retrieved in the step (b61); and (b62) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data.

According to another aspect of the present invention, a sensor data fusion-based IoT weather sensor device anomaly detection system includes at least one processor; And at least one memory for storing computer-executable instructions, wherein the computer-executable instructions stored in the at least one memory are, by the at least one processor, (a) from each IoT device, a plurality of receiving meteorological sensor data measured by the meteorological sensor; (b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; And, (c) a step of storing whether or not there is an abnormality in the database for each weather sensor of each IoT device is executed, and before step (b), (b0) the weather received from each IoT device in the past from the database. The step of reading the sensor data is further executed, and in the step (b), for each IoT device performing abnormality detection, (b3) the IoT device currently performing abnormality detection (hereinafter referred to as 'corresponding IoT device') The point in time at which the meteorological elements (hereinafter referred to as the “corresponding meteorological element”) of the nearby IoT devices (hereinafter referred to as “surrounding IoT devices”) currently performing anomaly detection is matched in the past data, and at that point Performing an anomaly detection of a corresponding meteorological element value of a corresponding IoT device from a corresponding meteorological element value of a corresponding IoT device; (b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device; (b5) Find the point in time when the current value and the past value of other meteorological factors of the surrounding IoT device coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and, (b6) find a point in time when the current value and past value of other meteorological factors of the IoT device and the current value and past value of all meteorological factors of the surrounding IoT device all match from the past data, and the corresponding IoT device at that time Including one or more steps of performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the corresponding meteorological element value, and between the step (b) and the step (c), (b7) the step (b3), The step of finally determining whether the corresponding meteorological element of the corresponding IoT device is abnormal based on the abnormality detection results in the steps included in steps (b4), (b5), and (b6) is based on a predetermined criterion. make it run more.

According to another aspect of the present invention, a computer program for detecting an anomaly in a sensor data fusion-based IoT weather sensor device is stored in a non-temporary storage medium, and by a processor, (a) from each IoT device, a plurality of weather sensors Receiving weather sensor data measured by; (b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; And, (c) a command for executing the step of storing whether or not each meteorological sensor of each IoT device is abnormal in a database, and prior to step (b), from (b0) the database, each IoT device in the past. It further includes a command for executing the step of reading the meteorological sensor data received from the step (b), for each IoT device that detects anomaly, (b3) the IoT device that is currently performing anomaly detection (hereinafter referred to as 'corresponding IoT device') and the weather elements currently performing anomaly detection (hereinafter referred to as 'corresponding weather elements') of nearby IoT devices (hereinafter referred to as 'surrounding IoT devices') coincide. Searching for in past data, and performing an anomaly detection of a corresponding meteorological element value of the corresponding IoT device from the corresponding meteorological element value of the corresponding IoT device at that time; (b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device; (b5) Find the point in time when the current value and the past value of other meteorological factors of the surrounding IoT device coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and, (b6) find a point in time when the current value and past value of other meteorological factors of the IoT device and the current value and past value of all meteorological factors of the surrounding IoT device all match from the past data, and the corresponding IoT device at that time Including one or more steps of performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the corresponding meteorological element value, and between the step (b) and the step (c), (b7) the step (b3), The step of finally determining whether the corresponding meteorological element of the corresponding IoT device is abnormal based on the abnormality detection results in the steps included in steps (b4), (b5), and (b6) is based on a predetermined criterion. Contains more commands to be executed.

According to the present invention, specific weather sensor data of an IoT device performing abnormal detection, other weather sensor data of the corresponding IoT device, weather sensor data of other nearby IoT devices, and past weather sensor data previously generated and stored in a server In combination, it provides a method for optimally determining whether the weather sensor device is abnormal by calculating and collecting abnormal numerical values of the weather sensor data without being affected by the specific value according to the location where the IoT device is installed and the time when the sensor data is measured. has the effect of

1 is a diagram showing a network configuration for performing abnormal detection of a weather sensor according to the present invention;
2 is a flow chart showing an overall process in which the sensor data fusion-based IoT weather sensor device anomaly detection system of the present invention performs anomaly detection for each weather sensor.
3 is a flowchart illustrating an algorithm for detecting an anomaly for a specific weather sensor by the sensor data fusion-based IoT weather sensor device anomaly detection system according to the present invention.
FIG. 4 is a diagram for explaining a 6-step algorithm for performing abnormal detection on a specific weather sensor of FIG. 3;
5 is a diagram showing the configuration of the sensor data fusion-based IoT weather sensor device anomaly detection system of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, the terms or words used in this specification and claims should not be construed as being limited to the usual or dictionary meaning, and the inventor appropriately uses the concept of the term in order to explain his/her invention in the best way. It should be interpreted as a meaning and concept consistent with the technical idea of the present invention based on the principle that it can be defined. Therefore, since the embodiments described in this specification and the configurations shown in the drawings are only one of the most preferred embodiments of the present invention and do not represent all of the technical ideas of the present invention, various alternatives may be used at the time of this application. It should be understood that there may be equivalents and variations.

1 is a diagram showing a network configuration for performing abnormal detection of a weather sensor according to the present invention, and FIG. 2 is a sensor data convergence-based IoT weather sensor device abnormality detection system 100 of the present invention for each weather sensor. It is a flowchart showing the entire process of performing anomaly detection.

An Internet of Things (IoT) device 200 includes a plurality of sensors capable of measuring a plurality of meteorological factors. Examples of such weather elements include temperature, atmospheric pressure, wind direction, wind speed, and humidity, and may further include other weather elements.

The IoT devices 200 are installed at fixed locations at regular distances, and the locations are stored in the sensor data convergence-based IoT weather sensor device anomaly detection system 100 .

In the sensor data convergence-based IoT weather sensor device anomaly detection system 100 of the present invention, when trying to determine whether the meteorological sensor data of the IoT device 200 is anomaly, it does not judge an anomaly by using general filtering with only a single weather sensor data. Instead, it determines whether the optimal weather sensor data is abnormal by combining other weather sensor data of the corresponding IoT device, weather sensor data of other nearby IoT devices, and past weather sensor data previously generated and stored in the server.

The sensor data convergence-based IoT weather sensor device anomaly detection system 100 waits to receive various weather sensor data from each IoT device 200 (S202) and receives weather sensor data obtained by measuring current weather element values (S203). ). The sensor data convergence-based IoT weather sensor device anomaly detection system 100 then detects an anomaly in each weather sensor data, and if necessary, the past IoT devices 200 stored on the weather sensor database (105, see FIG. 5) The meteorological sensor data received from and stored can be read and used (S201). As shown in FIG. 2, it can be read in advance, but it can also be read while starting to detect abnormalities.

Thereafter, abnormality detection is performed on the temperature sensor, air pressure sensor, wind direction sensor, wind speed sensor, and humidity sensor in each of the IoT devices 200 (S204 to S208). As described above, other weather sensors are further provided. If it is detected, anomaly detection may be performed therefor, or only anomaly detection may be performed by using only some of the above-mentioned 5 weather sensors.

After detecting an abnormality for each weather sensor, the result is stored and updated in the weather sensor database (105, see FIG. 5) (S209).

Although omitted in FIG. 2, steps S204 to S209 are repeatedly performed for each data of each IoT device.

In addition, the above-described abnormality detection for the temperature sensor, air pressure sensor, wind direction sensor, wind speed sensor, and humidity sensor is performed through the same algorithm, and such an algorithm may include six steps as an embodiment, which will be described below. It will be described in detail with reference to FIGS. 3 and 4 .

Thereafter, all sensor data abnormal detection results of the IoT device 200 are collected (S210), and the abnormal detection collection result of the IoT device 200 is updated in the DB (S211).

3 is a flow chart showing an algorithm for performing abnormal detection of a specific weather sensor by the sensor data convergence-based IoT weather sensor device anomaly detection system 100 of the present invention, and FIG. 4 is an anomaly detection of a specific weather sensor of FIG. 3 This is a diagram to explain the six-step algorithm that performs

In order to determine the final anomaly detection of one meteorological element of one IoT device, six anomaly detection steps may be performed. Here, the meteorological factor is, for example, temperature as described with reference to FIG. 2 . Air pressure, wind direction, wind speed, humidity, etc. As described above, in the flowchart of FIG. 2, all of these five anomalies may be detected, only some of them may be detected, or anomalies of other meteorological factors may be further included. there is. i.e. temperature. Air pressure, wind direction, wind speed, and humidity are only examples of meteorological factors that detect abnormalities in the present invention. In addition, since detecting an abnormality of a 'weather element' is the same as detecting an abnormality of a 'weather sensor' that detects the weather element, in the present invention, the expression 'weather element' abnormal detection and abnormal detection of 'weather sensor' shall be used interchangeably with the same meaning.

In order to detect anomalies, multiple meteorological sensor data samples must be collected for learning, and since anomalies are detected for one meteorological element of one device, the total number of IoT devices X all meteorological elements (temperature, air pressure, wind direction, wind speed) and humidity) is performed as many times as possible.

For example, if there are 10 IoT devices received at one time and anomaly detection is performed for 5 meteorological elements, a total of 50 times ( 10 X 5 = 50 ) will be performed.

Hereinafter, a six-step anomaly detection algorithm for each weather element is described. At this time, it is preferable to perform all six steps for more precise abnormality detection, but it is not necessary to perform all six steps, and steps may be selectively performed among them.

In addition, for convenience, the following description will describe a case in which there are three IoT devices in one region, and the first IoT device performs abnormal detection of the temperature T1 of the meteorological elements measured in six steps. do.

In step 1, the same meteorological factors of nearby IoT devices within a certain distance from the corresponding device are used as samples at the present time (S301). That is, referring to FIG. 4, when detecting whether or not the temperature T1 measured by IoT device 1 (IoT 1) is abnormal, a number of samples, that is, the temperature T2 measured by IoT 2 and measured by IoT 3 It is to detect whether or not T1 is abnormal from the temperature (T3) obtained. In this way, there may be various methods of determining whether a specific weather element is abnormal from a plurality of samples, and how far the corresponding weather element T1 is out of the samples is determined. In FIG. 4 , the derived samples are marked in red, and it is determined whether or not T1 is abnormal.

Step 2 is the corresponding meteorological element (temperature) value of the equipment (IoT 1) between a fixed point in time (for example, from November 1 to December 31 every year if the current date is December 1, 2020) from historical data is extracted as a sample (S302). That is, in FIG. 4, T1_01 to T1_61 which is the temperature data from 11.1 to 12.31 in 2019, T1_01 to T1_61 which is the temperature data from 11.1 to 12.31 in 2018, T1_01 to T1_61 which is the temperature data from 11.1 to 12.31 in 2017, etc. are samples. it will be data.

Step 3 finds the point in time at which the same meteorological element (temperature) of the surrounding IoT devices (IoT 2 and IoT 3) coincides with the past data, and the corresponding meteorological element (temperature) value (T11) of the corresponding IoT device (IoT 1) at that time , T12, T13 ...) are extracted as samples (S303).

Here, the meaning that the temperature 'matches' or that the temperature at the time of 'past 1' is 'same' as T2 as shown in FIG. 4 means that the temperature of IoT 2 at the time of 'past 1' is This means that the temperature is close to the temperature T2 of the IoT 2, that is, within the range of T2±ΔT, and the same applies to other cases below.

For example, when trying to detect an abnormal temperature value of IoT 1, when the temperature value of nearby IoT 2 is 4.8 degrees and IoT 3 is 4.9 degrees, in the past data, the temperature value of IoT 2 is 4.1 to 5.5 degrees and IoT 3 is 4.2 degrees. It is to find points at ~5.6 degrees, find the temperature values (T11, T12, T13) of IoT 1 at those points and use them as one of the sample data.

In this way, if the number of matching samples is insufficient, it is detected by making ΔT larger in the past data. That is, by finding points at which the temperature value of IoT 2 is 4 to 5.6 degrees and that of IoT 3 is 4.1 to 5.7 degrees, the temperature values (T11, T12, T13) of IoT 1 are found and used as one of the sample data.

Step 4 finds a point in time at which other meteorological factors of the corresponding device (IoT 1) match in past data, and extracts the corresponding meteorological factor (temperature) value of the IoT 1 at that point as a sample (S304).

Referring to FIG. 4, a past point in time (past 1, past 2, past 3. ..) to extract the temperature values T11, T12, T13, ... as samples.

For example, if you want to detect an abnormal temperature value of IoT 1, when the current humidity value is 47%, the wind direction is 315', the wind speed is 17m/s, and the atmospheric pressure is 1049mb, the humidity value of IoT 1 in the past data is 40~ 54%, the wind direction is 300 ~ 330 ', the wind speed is 14 ~ 20 m/s, and the air pressure is 1019 ~ 1079mb, and the temperature value of IoT 1 (T11, T12, T13,... ) is found and used as sample data.

Again, if the number is insufficient here, the same value range (ΔW, △D, △V, △P) is extended and searched in the past data, that is, the humidity value of IoT 1 is 39 to 55% and the wind direction is 295 to 335' , Find the point at which the wind speed is 13 ~ 21 m/s and the atmospheric pressure is 1014 ~ 1084mb, and find the temperature value (T11, T12, T13,...) of No. 1 equipment at that point and use it as sample data.

Step 5 is W21, D21, V21, P21 (IoT 2), W31, D31, V31, P31 (IoT 3), which are the current values of humidity, wind direction, wind speed, and air pressure, which are other meteorological factors of IoT 2 and IoT 3, which are peripheral devices. (Past 1, Past 2, Past 3...) are found in the past data, and T11, T12, T13, ..., which are the corresponding meteorological element values of the IoT at that time, are used as sample data (S305). .

For example, if the temperature value of IoT 1 is abnormally detected, the humidity value of IoT 2 is 42%, the wind direction is 321', the wind speed is 17 m/s, and the air pressure is 1049 mb, and the humidity value of IoT 3 is 40% and the wind direction is When the wind speed is 16 m/s and the barometric pressure is 1051 mb at 308', the humidity value of IoT 2 is 35 to 49% in the past data, the wind direction is 306 to 336', the wind speed is 14 to 20 m/s, and the barometric pressure is 1019 to 1079 mb. , find the point where the humidity value of IoT 3 is 33~47%, the wind direction is 293 ~ 323', the wind speed is 13 ~ 17 m/s, and the atmospheric pressure is 1021 ~ 1081mb, and find the temperature value of device 1 at that point It is used as one of the sample values.)

Again, if the number of samples is insufficient here, the same value range (ΔW, △D, △V, △P) is increased and searched. ', wind speed 13 ~ 21 m/s, air pressure 1014 ~ 1084mb, IoT 3 humidity value 32 ~ 48%, wind direction 288 ~ 328', wind speed 12 ~ 18 m/s, air pressure 1016 ~ 1086mb Find the point in time and find the temperature value (T11, T12, T13, ...) of IoT 1 at that point and use it as sample data.

Step 6 is W11, D11, V11, P11, which are the current values of humidity, wind direction, wind speed, and air pressure, which are other meteorological factors of IoT 1, and all meteorological factors of IoT 2 and IoT 3, which are peripheral devices, such as temperature, humidity, The point at which the present values of wind direction, wind speed, and air pressure, T21, W21, D21, V21, P21 (IoT 2), T31, W31, D31, V31, P31 (IoT 3) all coincide (past 1, past 2, past 3). ...) is found in the past data, and T11, T12, T13, ... which are corresponding meteorological element values of the IoT at that time are used as sample data (S306).

For example, if the temperature value of IoT 1 is abnormally detected, the temperature value of IoT 1 is 5.2 degrees, the humidity value is 47%, the wind direction is 315', the wind speed is 17 m/s, the atmospheric pressure is 1049 mb, and the temperature value of IoT 2 is This 4.8 degrees, humidity value 42%, wind direction 321', wind speed 17m/s, air pressure 1049mb, temperature value of IoT 3 is 4.9 degrees, humidity value 40%, wind direction 308', wind speed 16m/s s, when the air pressure is 1051mb, in the past data, the humidity value of IoT 1 is 39~55%, the wind direction is 295 ~ 335', the wind speed is 13 ~ 21 m/s, the air pressure is 1014 ~ 1084mb, and the temperature of IoT 2 The value is 4~5.6 degrees, the humidity value is 35~49%, the wind direction is 306 ~ 336', the wind speed is 14 ~ 20 m/s, the air pressure is 1019 ~ 1079mb, the temperature value of IoT 3 is 4.1 ~ 5.7 degrees, humidity Find the point where the value is 33~47%, the wind direction is 293 ~ 323', the wind speed is 13 ~ 17 m/s, and the air pressure is 1021 ~ 1081mb, and the temperature values of equipment No. 1 at that point are T11, T12, T13,... . is used as sample data.

Again, if the number of samples is insufficient here, the same value range (ΔT, ΔW, ΔD, ΔV, ΔP) is extended and searched.

In addition, extraction conditions can be modified so that the number of samples is the optimal number for anomaly detection. That is, as an embodiment, it is possible to repeat the readjustment of extraction conditions so that 1 to 2% of the total is obtained. For example, the humidity value of IoT 1 is 39 to 55%, the wind direction is 0 to 360 ', and the wind speed is 13 to 21 m /s, air pressure is 1014 ~ 1084mb, temperature value of IoT 2 is 4 ~ 5.6 degrees, humidity value is 34 ~ 50%, wind direction is 0 ~ 360', wind speed is 13 ~ 21 m/s, air pressure is 1014 ~ 1084mb , the temperature value of IoT 3 is 4.1 ~ 5.7 degrees, the humidity value is 32 ~ 48%, the wind direction is 0 ~ 360 ', the wind speed is 12 ~ 18 m/s, and the air pressure is 1016 ~ 1086mb. If it is not 1-2% of the number, it can be readjusted again and repeated counting.

After performing steps 1 to 6 (S301 to S306) in this way, the anomaly detection result of each step is combined to determine whether it is an anomaly for the corresponding meteorological element (temperature above) (S307). Various standards may be applied to this, and as an example, if four or more abnormalities are determined to be normal as a result of detecting abnormality in steps 1 to 6, it may be determined that the abnormality is normal.

In addition, as described above, only some of the steps 1 to 6 may be selectively performed, and even in such a case, it is possible to finally determine whether the corresponding meteorological element is abnormal by setting a certain criterion.

When all 6 steps are performed for each weather element, and an anomaly is detected for all meteorological elements for all IoT devices, anomaly detection is performed for a total of 6 steps X the number of all IoT devices X the number of all meteorological elements.

5 is a diagram showing the configuration of the sensor data fusion-based IoT weather sensor device anomaly detection system 100 of the present invention.

Since the method for detecting an anomaly in a weather sensor by sensor fusion of the present invention has already been described in detail with reference to FIGS. The functions of the components will be briefly summarized and explained.

The control unit 101 controls each of the following components of the sensor data convergence-based IoT weather sensor device anomaly detection system 100 to perform a series of processes related to weather sensor anomaly detection.

The IoT device data receiving unit 102 receives various meteorological factor data measured by the IoT devices 200 .

The anomaly detection unit 103 performs anomaly detection for each meteorological element as described with reference to FIGS. 2 to 4 .

The anomaly detection collection unit 104 collects anomaly numerical values of weather sensor data of an IoT weather sensor device.

The database access unit 105 is a weather sensor database for storing various weather element data received from the IoT device 200 and data such as abnormal numerical values for each weather sensor data and records of abnormal numerical values for the weather sensor device ( 106) and also serves to store the data in the weather sensor database 106.

100: sensor data convergence-based IoT weather sensor device anomaly detection system
200: IoT (Internet of Things) device

Claims (10)

Sensor data convergence-based IoT weather sensor device anomaly detection method,
(a) receiving weather sensor data measured by a plurality of weather sensors from each IoT device;
(b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; and,
(c) Step of storing in the database whether there is an abnormality for each weather sensor of each IoT device
including,
Before the step (b),
(b0) Step of reading weather sensor data received from each IoT device in the past from the database
Including more,
In the step (b), for each IoT device performing abnormality detection,
(b3) Weather elements (hereinafter referred to as 'corresponding IoT devices') currently detecting anomalies of nearby IoT devices (hereinafter referred to as 'surrounding IoT devices') of the IoT device currently performing anomaly detection (hereinafter referred to as 'corresponding IoT device') (referred to as 'corresponding meteorological element') finding a coincidence point in past data, and detecting an anomaly of a corresponding meteorological element value of a corresponding IoT device from a corresponding meteorological element value of a corresponding IoT device at that time;
(b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device;
(b5) Find the point in time when the current value and the past value of other meteorological factors of the surrounding IoT device coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and,
(b6) Find a point in time when the present and past values of other meteorological elements of the IoT device and the present and past values of all meteorological elements of the nearby IoT device all match from past data, and the corresponding weather of the IoT device at that time Performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the element value
Including one or more steps of
Between step (b) and step (c),
(b7) Based on the abnormality detection results in the steps included among the steps (b3), (b4), (b5), and (b6), the corresponding meteorological element of the corresponding IoT device according to a predetermined criterion. The final decision on whether or not the
Including more,
Sensor data convergence-based IoT meteorological sensor device anomaly detection method.
delete delete delete The method of claim 1,
If the IoT device currently performing anomaly detection is IoT k, and the meteorological element currently performing anomaly detection in the IoT k is A,
In the step (b3), the abnormal detection of the meteorological factor A measured by the IoT k,
(b31) the value of the meteorological factor A currently measured in each of the IoT devices around the IoT k;
The value of the meteorological factor A measured in the past by each of the surrounding IoT devices,
Searching for a matched past point of view for each nearby IoT device;
(b32) extracting the values of the meteorological factor A at the time points retrieved in the step (b31) as sample data; and,
(b33) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data;
Sensor data fusion-based IoT weather sensor device anomaly detection method comprising a. The method of claim 1,
If the IoT device currently performing anomaly detection is IoT k, and the meteorological element currently performing anomaly detection in the IoT k is A,
In the step (b4), the abnormal detection of the meteorological factor A measured by the IoT k,
(b41) values of other meteorological factors other than the meteorological factor A currently measured by the IoT k;
Values of other meteorological factors other than the meteorological factor A, measured in the past by the IoT k,
Searching for a matched past point in time for each weather element;
(b42) extracting the values of the meteorological factor A at the time points retrieved in the step (b41) as sample data; and,
(b43) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data;
Sensor data fusion-based IoT weather sensor device anomaly detection method comprising a. The method of claim 1,
If the IoT device currently performing anomaly detection is IoT k, and the meteorological element currently performing anomaly detection in the IoT k is A,
In the step (b5), the abnormal detection of the meteorological factor A measured by the IoT k,
(b51) values of other meteorological factors other than the meteorological factor A currently measured by each IoT device around the IoT k;
Values of other meteorological factors, except for the meteorological factor A, measured in the past by each IoT device around the IoT k,
Retrieving a past point in time that matches each meteorological element of each neighboring IoT device;
(b52) extracting, as sample data, the values of the meteorological factor A measured by the IoT k at the time points retrieved in the step (b51); and,
(b53) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data;
Sensor data fusion-based IoT weather sensor device anomaly detection method comprising a. The method of claim 1,
If the IoT device currently performing anomaly detection is IoT k, and the meteorological element currently performing anomaly detection in the IoT k is A,
In the step (b6), the abnormal detection of the meteorological factor A measured by the IoT k,
(b61) Values of other meteorological factors other than the meteorological factor A, currently measured by the IoT k, and values of all meteorological factors, including the meteorological factor A, currently measured by each IoT device around the IoT k;
Values of other meteorological factors other than the meteorological factor A, measured in the past by the IoT k, and values of all meteorological factors, including the meteorological factor A, measured in the past by each IoT device around the IoT k, IoT k and retrieving a past point in time that matches each meteorological element of each neighboring IoT device;
(b62) extracting, as sample data, the values of the meteorological factor A measured by the IoT k at the time points retrieved in the step (b61); and,
(b62) performing anomaly detection on the value of the currently measured meteorological factor A of the IoT k from the sample data;
Sensor data fusion-based IoT weather sensor device anomaly detection method comprising a. As a sensor data convergence-based IoT weather sensor device anomaly detection system,
at least one processor; and
including at least one memory for storing computer-executable instructions;
The computer-executable instructions stored in the at least one memory are, by the at least one processor,
(a) receiving weather sensor data measured by a plurality of weather sensors from each IoT device;
(b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; and,
(c) Step of storing in the database whether there is an abnormality for each weather sensor of each IoT device
to run,
Before the step (b),
(b0) Step of reading weather sensor data received from each IoT device in the past from the database
to run more,
In the step (b), for each IoT device performing abnormality detection,
(b3) Weather elements (hereinafter referred to as 'corresponding IoT devices') currently detecting anomalies of nearby IoT devices (hereinafter referred to as 'surrounding IoT devices') of the IoT device currently performing anomaly detection (hereinafter referred to as 'corresponding IoT device') (referred to as 'corresponding meteorological element') finding a coincidence point in past data, and detecting an anomaly of a corresponding meteorological element value of a corresponding IoT device from a corresponding meteorological element value of a corresponding IoT device at that time;
(b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device;
(b5) Find the point in time when the current value and the past value of other meteorological factors of the surrounding IoT device coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and,
(b6) Find a point in time when the present and past values of other meteorological elements of the IoT device and the present and past values of all meteorological elements of the nearby IoT device all match from past data, and the corresponding weather of the IoT device at that time Performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the element value
Including one or more steps of
Between step (b) and step (c),
(b7) Based on the abnormality detection results in the steps included among the steps (b3), (b4), (b5), and (b6), the corresponding meteorological element of the corresponding IoT device according to a predetermined criterion. The final decision on whether or not the
to run further,
Sensor data convergence-based IoT weather sensor device anomaly detection system. As a computer program for detecting anomalies in sensor data fusion-based IoT weather sensor devices,
It is stored in a non-transitory storage medium, and by the processor,
(a) receiving weather sensor data measured by a plurality of weather sensors from each IoT device;
(b) for each IoT device, using weather sensor data received from the corresponding IoT device, detecting an anomaly in each weather sensor of the IoT device; and,
(c) Step of storing in the database whether there is an abnormality for each weather sensor of each IoT device
contains a command that causes the
Before the step (b),
(b0) Step of reading weather sensor data received from each IoT device in the past from the database
Further includes a command that causes the to be executed,
In the step (b), for each IoT device performing abnormality detection,
(b3) Weather elements (hereinafter referred to as 'corresponding IoT devices') currently detecting anomalies of nearby IoT devices (hereinafter referred to as 'surrounding IoT devices') of the IoT device currently performing anomaly detection (hereinafter referred to as 'corresponding IoT device') (referred to as 'corresponding meteorological element') finding a coincidence point in past data, and detecting an anomaly of a corresponding meteorological element value of a corresponding IoT device from a corresponding meteorological element value of a corresponding IoT device at that time;
(b4) Find a point in time at which meteorological elements other than the corresponding meteorological element (hereinafter referred to as 'other meteorological elements') of the IoT device coincide with past data, and from the value of the meteorological element of the corresponding IoT device at that time Performing an anomaly detection of a corresponding meteorological element value of the IoT device;
(b5) Find the point in time when the current value and the past value of other meteorological factors of nearby IoT devices coincide with past data, and detect an anomaly in the corresponding meteorological factor value of the IoT device from the corresponding meteorological factor value of the IoT device at that time performing; and,
(b6) Find a point in time when the present and past values of other meteorological elements of the IoT device and the present and past values of all meteorological elements of the surrounding IoT devices all match from the past data, and the corresponding weather of the IoT device at that time Performing an anomaly detection of the corresponding meteorological element value of the corresponding IoT device from the element value
Including one or more steps of
Between the steps (b) and (c),
(b7) From the abnormality detection results in the steps included among the steps (b3), (b4), (b5), and (b6), the corresponding meteorological element of the corresponding IoT device according to a preset criterion The final decision on whether or not the
Further comprising a command that causes the to be executed,
A computer program for detecting anomalies in sensor data fusion-based IoT weather sensor devices.

Images