KR20170122043A - Real-time indoor air quality outlier smoothing method and apparatus - Google Patents

Abstract

A method of removing real-time indoor air quality outlier according to an embodiment of the present invention includes: (a) sensing indoor air pollutants using at least one sensor; and (b) removing an outlier by applying an exponential moving average method to data that is output from each of the at least one sensor. Accordingly, due to a less complex algorithm for removing an outlier, the use of less memory, a quickly converging result, and reasonable accuracy, the present invention is able to be applied to a low-cost and microminiature indoor air quality monitoring system, and when sensing values are used for various analyses and predictions, reliably perform the analyses and predictions.

Description

TECHNICAL FIELD [0001] The present invention relates to indoor air quality abnormalities,

The present invention relates to a real-time indoor air quality abnormal removal method and apparatus.

Interest in Indoor Air Quality (IAQ) has also increased as interest in home and indoor health and comfort has increased. This is because indoor air quality is one of the important factors affecting our health and life. Indoor air quality refers to the quality of air in buildings and structures. IAQ can be used for any gas, particulate matter, microbial contaminants (fungi, bacteria) or any mass or energy stress that can lead to harmful health conditions, including carbon monoxide, radon and volatile organic compounds (VOC) It can be influenced by factors. Using source control, filtration and ventilation to dilute contaminants are the main methods for improving indoor air quality in most buildings.

The main source of indoor air pollution in developing countries is the combustion of biomass for heating and cooking, for example, wood, coal, manure, or crop debris. As a result of exposure to high levels of particulate matter, 1.5 million to 2 million people died in 2000.

Until now, indoor air quality measurement has relied on measurement and analysis by environmental experts using professional equipment. However, due to the recent development of low-cost sensors and embedded systems, it is becoming possible to develop ultra low-cost indoor air quality (IAQ) monitoring systems. A major feature of these systems, which can have various low cost environmental monitoring sensors, is that they can inform the users of the indoor air quality in real time and can observe the variation of the indoor air quality in a specific place for a certain time.

Although these compact indoor air quality monitoring systems have a number of advantages, they suffer from low sensor accuracy and low reliability due to frequent outliers. Moreover, although a number of studies have been conducted to correct sensor values and outliers, most studies have been focused on different sensors and applications. Moreover, some of them require complex computation based on abundant resources in the processor and memory. Therefore, it is difficult to apply these algorithms to a small indoor air quality monitoring system.

Since the micro indoor air quality monitoring system is based on a low-cost low-performance sensor, improving the reliability of the sensors and coping with the outliers is one of the important challenges in the design of the indoor air quality monitoring system. Moreover, existing smoothing algorithms can not be applied directly to a micro indoor air quality monitoring system, since existing smoothing algorithms require a great deal of overhead to be processed in the system.

Thus, there is a need for an outlier removal method and apparatus for a low cost, ultra-small indoor air quality monitoring system, wherein the algorithm is less complex, uses less memory, results quickly converge, and has reasonable accuracy.

KR 10-1202783 B1

It is an object of the present invention to provide a real-time indoor air quality abnormality elimination method for an inexpensive ultra-small indoor air quality monitoring system in which algorithms are less complicated, less memory is used, results are converged quickly, and reasonable accuracy will be.

Another object of the present invention is to provide a real-time indoor air quality abnormality elimination device for a low cost indoor indoor air quality monitoring system with less complicated algorithm, less memory use, quick results converged, and reasonable accuracy .

According to an embodiment of the present invention, there is provided a method for removing real-time indoor air quality abnormalities,

(a) sensing indoor air pollutants using at least one sensor; And

(b) removing an ideal value by applying an exponential moving average method to the data output from each of the at least one sensor.

The method may further include the step of determining and displaying the indoor air quality based on the data to which the exponential moving average method is applied after the step (b).

In the step (b), S _{t, alternate} = α · Y _t + (1-α) · S _t -1, ₁ , the exponential moving average method is applied to the data output from each of the at least one sensor, where S _{t, alternate} is data to which an exponential moving average method at time t is applied, and Y _t denotes data data at time t, S _t-1 may be an exponential moving average of data immediately before the time t, and? may be an exponential smoothing coefficient.

Further, in the real-time indoor air quality abnormal value removal method according to an embodiment of the present invention, the exponential smoothing coefficient? May be 0.3.

The at least one sensor may be a volatile organic compound sensor (VOC), a carbon monoxide sensor, a carbon dioxide sensor, a fine dust sensor, a temperature sensor, and a humidity sensor And at least one sensor selected from the group consisting of:

According to another aspect of the present invention, there is provided a real-

At least one sensor for sensing indoor air pollutants; And

And a controller for applying an exponential moving average method to the data output from each of the at least one sensor to remove an abnormal value.

According to an embodiment of the present invention, there is provided a real-time indoor air quality abnormality eliminating apparatus, further comprising a display unit for displaying indoor air quality, wherein the controller determines an indoor air quality based on the data to which the exponential moving average method is applied, Can be displayed on the display unit.

Further, in the real-time indoor air quality abnormality removing apparatus according to an embodiment of the present invention, the control unit may be configured to perform the following steps based on Equation 1: S _{t, alternate} = α · Y _t + (1-α) · S _t-1 Wherein S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data of the time point of the data output to the respective sensors, The data, S _t-1 , may be the exponential moving average of the data immediately before the point in time t, and? May be the exponential smoothing coefficient.

Further, in the real-time indoor air quality abnormality removing apparatus according to an embodiment of the present invention, the exponential smoothing coefficient? May be 0.3.

According to an embodiment of the present invention, the at least one sensor may include a volatile organic compound sensor (VOC), a carbon monoxide sensor, a carbon dioxide sensor, a fine dust sensor, a temperature sensor, and a humidity sensor And at least one sensor selected from the group consisting of:

According to the method and apparatus for removing real-time indoor air quality abnormalities according to an embodiment of the present invention, since algorithms for eliminating anomalous values are less complex, less memory is used, results are quickly converged, and reasonable accuracy, Of indoor air quality monitoring system.

In addition, according to the method and apparatus for removing real-time indoor air quality abnormalities according to an embodiment of the present invention, since abnormal abnormal values are smoothed and removed from data output from each sensor, when the sensing values are used for various analysis and prediction, Analysis and forecasting can be done.

In particular, in the case of the server collection model in which the indoor air quality data is collected from the server, the burden of the server to process the data can be considerably reduced since each indoor air quality monitoring system itself has the ability to process outliers.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a view showing a micro indoor air quality monitoring system and a test-bed to which a real-time indoor air quality abnormal-value removal method and apparatus according to an embodiment of the present invention can be applied.
2A to 2D are graphs showing the results of data collection from each sensor.
FIGS. 3A to 3D are graphs showing the result of applying a simple moving average (SMA) to data output from each sensor. FIG.
FIGS. 4A to 4D are graphs showing the results of applying the exponential moving average (EMA) to the data output from each sensor.
5A to 5D are graphs for comparing an exponential moving average and a simple moving average.
6 is a block diagram illustrating a real-time indoor air quality monitoring system including a real-time indoor air quality abnormal-value removing apparatus according to an embodiment of the present invention.
7 is a flowchart of a real-time indoor air quality abnormal-value removal method according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The objectives, specific advantages and novel features of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG.

Prior to that, terms and words used in the present specification and claims should not be construed in a conventional and dictionary sense, and the inventor may properly define the concept of the term in order to best explain its invention Should be construed in accordance with the principles and the meanings and concepts consistent with the technical idea of the present invention.

It should be noted that, in the present specification, the reference numerals are added to the constituent elements of the drawings, and the same constituent elements are assigned the same number as much as possible even if they are displayed on different drawings.

Also, the terms "first", "second", "one side", "other side", etc. are used to distinguish one element from another, It is not.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following description of the present invention, a detailed description of known arts which may unnecessarily obscure the gist of the present invention will be omitted.

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

First of all, we have identified some outliers by experiments with small indoor air quality monitoring system. Based on this, a lightweight in-system smoothing algorithm based on an exponential moving average algorithm is proposed in an embodiment of the present invention. do.

Specifically, in one embodiment of the present invention, a plurality of indoor air quality sensors including a VOC sensor, a CO sensor, a CO2 sensor, a PM10 sensor, a temperature sensor, and a humidity sensor are introduced into a real-time indoor air quality monitoring system, Experimental results showing some problems caused by Based on the results, design requirements for a real-time indoor air quality monitoring system are defined, and two smoothing methods (simple moving average and exponential moving average) are applied to the system, and the results are analyzed. Through additional important measurement experiments, the most suitable in-system real-time outlier removal methods and parameters for a compact indoor air quality monitoring system are finally proposed.

Small interior Air quality monitoring  System overview

FIG. 1 shows an ultra-small indoor air quality monitoring system and a test-bed capable of sensing VOC, CO, CO2, PM10, temperature and humidity, respectively, in order to collect indoor air quality information. The collected data sets are periodically transmitted to a server (not shown) via Wi-Fi. Using this system, indoor air quality data was collected and the distribution of each sensor data was analyzed based on the data.

The experimental environment for monitoring real-time indoor air quality is as shown in Table 1.

Under the experimental environment shown in Table 1, data of each sensor was collected and indoor air quality value distribution was observed.

Experiment environment location Embedded Network System Architecture Lab, INU Sensor type VOC, Co, PM10, Co2, Temperature, Humidity Collection period 5 minutes Collection period 355 hours (about 2 weeks) Number of data collected 4006 Data analysis tools R Studio Plotting Tools Plotly Database MySQL Workbench 5.1 CE

Figures 2a to 2d show the variation of each sensor collected during the experimental period. Sensors are inexpensive and compact, but they can identify changes and trends in indoor air quality. However, it should be noted that the results show some outliers (red boxes) or missing values in some cases. Since most of the outliers can be recovered again, it can be considered less important when intuitively identifying the overall variation of the sensing values using the graphs as in Figs. 2A-2D.

However, when the sensing values are used for various analyzes and predictions, the outliers can cause unreliable results. In particular, in the case of the server collection model, the overhead of the server will increase significantly because all the processing overhead is concentrated on the server. Therefore, it is desirable that each indoor air quality monitoring system itself has the ability to process outliers.

Each sensor in the indoor air quality monitoring system has several outliers and unfortunately they occur unexpectedly. Therefore, there is a need to develop an in-system outlier removal method for a compact indoor air quality monitoring system. First, representative smoothing algorithms applicable to the micro indoor air quality monitoring system are presented, and other outlier detection algorithms are briefly introduced. Furthermore, an important requirement is presented for in-system out-of-order smoothing designs for low cost indoor air quality monitoring systems.

K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and K-Means method are generally used for detection of outliers. KNN is a classification method that calculates and classifies distances from each existing object when a new object appears in the nearest neighbor approach. SVM is a model for finding a straight line that separates two different classes when an arbitrary object appears. K-Means is a distance-based clustering technique that decomposes a set of N objects into K clusters.

Although the above outlier detection algorithms exhibit good performance, most of them have unique assumptions and are limited in the type of sensor they use, so they can be used in a wide variety of sensor types and can be implemented in servers Smoothing methods that can be used at low overhead should be developed.

Although there are a number of outliers detection algorithms, we focus on three moving averages known as representative methods for a simple smoothing algorithm: Simple Moving Average (SMA), Exponential Moving Average (EMA), and Weighted Moving Average (WMA).

The SMA described in Equation (1) provides a simple but robust function of predicting the next data based on analyzing the trend from the historical data set by calculating the average value dynamically using the moving window. Unlike SMAs that use a certain period average, EMA and WMA calculate the average for all data, but recent data is more weighted in the time series. Unlike WMA, where all data in the time series should have different weights as shown in equation (2), EMA can be obtained more simply using equation (3).

In Equation (3), S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data at time t among the data outputted to each sensor, S _t-1 is exponential moving average , and? is the exponential smoothing coefficient.

The important requirements for designing a new in-system outlier smoothing for a micro indoor air quality monitoring system based on existing smoothing and outlier detection methods are as follows.

- Low algorithm complexity

- Low memory usage

- Fast convergence

- Reasonable accuracy

Experiment result

To test the effectiveness of the proposed smoothing algorithms, experiments were carried out with SMA and EMA applied to the micro indoor air quality monitoring system. Figures 3a-3d show the observation results of the SMA applied to raw sensing data using different window sizes 12-26, Figures 4a-4d illustrate the observation results using different exponential smoothing coefficients 0.1 and 0.3 EMA < / RTI >

Experimental results show that the best window size for SMA is 12 and the best exponential smoothing factor for EMA is 0.3. Therefore, to directly compare the two smoothing methods, Figure 5 shows the comparison of SMA (window size = 12) and EMA (alpha = 0.3).

Performance evaluation

In fact, it is difficult to conclude only with the comparison result shown in Fig. 5, which smoothing method shows better performance. Therefore, in order to evaluate more specific and objective performance, additional metrics are adopted: Root Mean Square Error (RMSE), Variability and Memory Usage. These evaluation factors will be useful in assessing how much each algorithm meets design requirements.

The square root mean square error (RMSE) described in Equation (4) is the most intuitive and meaningful accuracy evaluation method used to measure the accuracy of smoothed data compared to raw data. The root mean square error (RMSE) is measured by averaging over the square of the difference between the actual data and the modeling data. The greater the number of data, the more accurate the evaluation of accuracy is possible, and the closer to 0, the more accurate it is. ,

The variability described in Equation (5) is a method of measuring how far apart the average value of the data is and is used to evaluate the reliability of the data. The variance of the difference between the actual data and the modeling data is divided by the variance of the actual data, and then measured from 1 to 뺌. The closer to 1, the better the reliability. In general, it is judged to be reliable if it is 0.8 or more.

Additionally, the required memory usage of each algorithm is checked during execution time.

Table 2 shows the comparison of SMA and EMA for each evaluation factor. The results show that EMA has better performance over SMA for all sensing values. In particular, for RMSE, EMA shows about 0.5 times better SMA, and variability of EMA shows better reliability than SMA. Moreover, the memory usage of the EMA is six times better than the SMA.

SMA (Windows = 12) EMA (alpha = 0.3) Evaluation factor
(Metric) TVOC Co Dust Average TVOC Co Dust Average RMSE 12.2403 0.6664 12.0798 8.3288 4.3033 0.3777 8.6568 4.4459 Variability 0.9801 0.9338 0.8141 0.9093 0.9975 0.9788 0.9045 0.9603 Memory Usage
(bytes) 180 180 180 540
(Total
Required Memory) 30 30 30 90
(Total
Required Memory)

conclusion

Indoor air quality is one of the important factors affecting our health and life. Recently, a compact low cost indoor air quality monitoring system has been developed. In particular, since the ultra-small indoor air quality monitoring system is based on a low-cost low-performance sensor, improving the reliability of sensors and coping with abnormalities is one of the important challenges in designing an indoor air quality monitoring system. Moreover, existing outlier detection algorithms can not be directly applied to the ultra-small indoor air quality monitoring system. This is because they require enormous overhead to be processed in the system. Therefore, some outliers problems have been identified, and a lightweight in-system smoothing algorithm has been proposed based on an exponential moving average algorithm. Experimental results show that the proposed in-system outlier smoothing algorithm can smooth out outliers with high accuracy, reliability and low memory usage. Moreover, through evaluation, it has been demonstrated that the EMA-based in-system smoothing algorithm outperforms the SMA-based method.

Based on the experimental results, the method and apparatus for removing real-time indoor air quality abnormalities according to an embodiment of the present invention apply an exponential moving average to data output from each sensor to remove the outliers by smoothing the outliers.

6 is a block diagram illustrating a real-time indoor air quality monitoring system including a real-time indoor air quality abnormal-value removing apparatus according to an embodiment of the present invention.

Referring to FIG. 6, the real-time indoor air quality abnormalities removal device according to an embodiment of the present invention, which is a real-time indoor air quality abnormalities removal device in the object-internet-based environmental information collection platform, includes a plurality of sensors A control unit for applying an exponential moving average method to data outputted from each of the plurality of sensors 600, 602, 604, 606, 608, and 609 and the plurality of sensors 600, 602, 604, 606, 608, (610).

The plurality of sensors 600, 602, 604, 606, 608, and 609 may include a volatile organic compound sensor (VOC) 600, a carbon monoxide sensor 602, a carbon dioxide sensor 604, a fine dust sensor 606, A sensor 608, and a humidity sensor 609.

In addition, the real-time indoor air quality abnormality eliminating apparatus according to an embodiment of the present invention includes a display unit 614 for displaying indoor air quality, a communication unit 612 for communicating with a server (not shown) and transmitting indoor air quality- .

The operation of the real-time indoor air quality abnormal-value removal apparatus according to the embodiment of the present invention will now be described.

The control unit 610 receives data output from each of the plurality of sensors 600, 602, 604, 606, 608, and 609, and applies an exponential moving average method (EMA) to the received data.

That is, the control unit _{610, S t, alternate = α ·} Y t + (1-α) · S t-1 wherein the plurality of sensors based on a (600, 602, 604, 606, 608, 609) respectively, The exponential moving average method is applied to the data output from the memory.

In the above, S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data at time t among the data outputted to each sensor, S _t-1 is the exponential moving average for data just before time t, Is the exponential smoothing coefficient.

Since the RMSE, the variability, and the memory use are superior to the SMA when the exponential smoothing coefficient is 0.3, the controller 610 in the indoor air quality abnormalities removal apparatus according to an embodiment of the present invention has an exponential smoothing coefficient of 0.3 The exponential moving average method is applied to the data output from each of the plurality of sensors 600, 602, 604, 606, 608, and 609 in the set state.

Since the exponential moving average method is applied to the data output from each of the plurality of sensors 600, 602, 604, 606, 608 and 609, the plurality of sensors 600, 602, 604, 606, 608 and 609 Even if the outliers are output, the outliers are smoothed out.

As described above, since the control unit 610 smoothes and removes the outliers by applying the exponential moving average method to the data output from each of the plurality of sensors 600, 602, 604, 606, 608, and 609, The indoor air quality can be determined accurately and the indoor air quality determined accurately can be displayed on the display unit 614. [

7 is a flowchart of a method for removing indoor air quality abnormalities according to an embodiment of the present invention.

6 and 7, in step S700, the controller 610 senses indoor air pollutants using the plurality of sensors 600, 602, 604, 606, 608, and 609.

Next, in step S702, the control unit 610 applies the exponential moving average method to the output data of each of the sensors 600, 602, 604, 606, 608, and 609.

Control unit 610 exponential moving the data output from the _{S t, alternate = α · Y} t + (1-α) · on the basis of S _t-1, each sensor (600, 602, 604, 606, 608, 609) The averaging method is applied.

In the above, S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data at time t among the data outputted to each sensor, S _t-1 is the exponential moving average for data just before time t, Is the exponential smoothing coefficient. In one embodiment of the present invention, the exponential smoothing coefficient alpha is 0.3.

Next, in step S704, the controller 610 accurately determines the indoor air quality based on the data to which the exponential moving average method is applied, and displays the determined indoor air quality on the display unit 614. [

According to the method and apparatus for removing real-time indoor air quality abnormalities according to an embodiment of the present invention, since algorithms for eliminating anomalous values are less complex, less memory is used, results are quickly converged, and reasonable accuracy, Of indoor air quality monitoring system.

In addition, according to the method and apparatus for removing real-time indoor air quality abnormalities according to an embodiment of the present invention, since abnormal abnormal values are smoothed and removed from data output from each sensor, when the sensing values are used for various analysis and prediction, Analysis and forecasting can be done.

In particular, in the case of the server collection model in which the indoor air quality data is collected from the server, the burden of the server to process the data can be considerably reduced since each indoor air quality monitoring system itself has the ability to process outliers.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. It is clear that the present invention can be modified or improved.

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

600: VOC sensor 602: CO sensor
604: CO2 sensor 606: fine dust sensor
608: Temperature sensor 609: Humidity sensor
610: Control unit 612:
614:

Claims (10)

(a) sensing indoor air pollutants using at least one sensor; And
(b) removing exponential values by applying an exponential moving average method to the data output from each of the at least one sensor. The method according to claim 1,
After the step (b)
And determining and displaying the indoor air quality based on the data to which the exponential moving average method is applied. The method according to claim 1,
In the step (b)
[Equation 1]
And applying an exponential moving average method to the data output from each of the at least one sensor based on S _{t, alternate} = α · Y _t + (1-α) · S _t-1 ,
In the above, S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data at time t among the data outputted to each sensor, S _t-1 is the exponential moving average for data just before time t, Is an exponential smoothing coefficient. The method of claim 3,
Wherein the exponential smoothing coefficient (alpha) is 0.3. The method of claim 4,
Wherein the at least one sensor comprises at least one sensor selected from the group consisting of a volatile organic compound sensor (VOC), a carbon monoxide sensor, a carbon dioxide sensor, a fine dust sensor, a temperature sensor and a humidity sensor. At least one sensor for sensing indoor air pollutants; And
And a control unit for applying an exponential moving average method to the data output from each of the at least one sensor to remove an abnormal value. The method of claim 6,
Further comprising a display unit for displaying indoor air quality,
Wherein the control unit determines the indoor air quality based on the data to which the exponential moving average method is applied and displays the determined indoor air quality on the display unit. The method of claim 6,
Wherein,
[Equation 1]
And applying an exponential moving average method to the data output from each of the at least one sensor based on S _{t, alternate} = α · Y _t + (1-α) · S _t-1 ,
In the above, S _{t, alternate} is the data to which the exponential moving average method at time t is applied, Y _t is data at time t among the data outputted to each sensor, S _t-1 is the exponential moving average for data just before time t, Is an exponential smoothing coefficient. The method of claim 8,
Wherein the exponential smoothing coefficient (alpha) is 0.3. The method of claim 9,
Wherein the at least one sensor comprises at least one sensor selected from the group consisting of a volatile organic compound sensor (VOC), a carbon monoxide sensor, a carbon dioxide sensor, a fine dust sensor, a temperature sensor and a humidity sensor.

Images