Optimization Design of Miniature Air Quality Monitoring System Based on Multi-Sensor Fusion Technology

(1 National Engineering Lab of Special Display Technology,Academy of Opto-Electronic Technology,Hefei University of Technology,Hefei230009,China;2 State Key Laboratory of Transducer Technology,Institute of Intelligent Machines,HFIPS,Chinese Academy of Sciences,Hefei 230031,China;3 Key Laboratory of Environmental Optics and Technology,Anhui Institute of Optics and Fine Mechanics,HFIPS,Chinese Academy of Sciences,Hefei 230031,China)

Abstract:In view of the challenge that the National Ambient Air Quality Monitoring Station(NAAQMS)is not suitable for large area distribution due to its large volume and high cost,a miniature air quality monitoring system is developed to monitor the concentrations of CO,NO2,O3,SO2,PM2.5and PM10in atmosphere.Electrochemical gas sensors and optical particle counter are adopted to measure the concentration of air pollutants and particles respectively in the system.Considering that the sensor is vulnerable to the influcence of temperature and humidity of atmosphere when measuring the concentration of pollutants,the measurement results are modified through the temperature and humidity compensation algorithm based on the multi-sensor fusion technology.Then the modified measurement results are compared with the data released by NAAQMS,and it is shown that there is a strong correlation between the two data.Compared with the data obtained from NAAQMS,the correlation coefficient for 10 days monitoring data obtained from the developled system is better than 0.7.After the algorithm correction,the correlation coefficient of some pollutants,such as PM2.5,can even be improved to 0.9.The air quality monitoring system is robust and small in size,which is suitable for long-term and large area distributed network monitoring of environmental air pollutants.

Key words:air quality monitoring;optimization design;multi-sensor fusion;miniaturization

0 Introduction

Air is a necessary condition for human beings to survive.The quality of air has a great impact on human health and living environment.When the air quality becomes very poor,people living in it are prone to get sick and their body resistance will decline[1].In today′s society,the rapid development of industrialization,coal,dust,automobile exhaust,industrial emissions,domestic sewage and other factors cause a sharp increase in toxic and harmful gas emissions,resulting in the continuous deterioration of air quality.At the same time,people in the world suffer from air pollution and various respiratory diseases[2-4].

Haze weather and photochemical smog pollution is becoming more and more serious,which occurs frequently in Beijing and Yangtze River Delta region.In haze weather,PM2.5(fine particles with diameter less than or equal to 2.5 μm)in the atmosphere enter the human body with breathing,which is easy to cause lung cancer,bronchitis and other diseases[5-7].Moreover,the haze weather will reduce visibility and bring inconvenience to daily life.Among the gaseous pollutants discharged,nitrogen oxides will destroy the ozone layer of the atmosphere,and sulfur dioxide will combine with water vapor to form acid rain,causing pollution of rivers and land[8,9].

In China,the evaluation of ambient air quality index is mainly based on the concentration of the following gas pollutants,namely sulfur dioxide(SO2),carbon monoxide(CO),ozone(O3),nitrogen dioxide(NO2),fine particles(PM2.5)and inhalable particles(PM10).At present,China has established more than 1000 National Ambient Air Quality Monitoring Stations(NAAQMS)nationwide to cope with the increasingly severe air pollution situation.Air quality detection station can monitor the air quality change in a certain range near the station in real time,but it has many limitations[2].The cost of NAAQMS is high,whose price is as high as millions of Chinese Yuan(CNY)for a single site,and the overall distribution is also relatively scarce,which can only respond to large environmental pollution problems,and cannot meet the needs of the current stage.In addition,the monitoring station covers a large area,which requires special personnel to maintain regularly,so it is impossible to realize large-scale distribution.Therefore,a new type of ambient air quality monitoring system with advantages of low-cost,small size,highresolution and easy to be installed is urgently needed.The micro ambient air quality monitoring system proposed in this paper can provide real-time,accurate and intuitive data information,which is suitable for long-term monitoring of ambient air quality concentration.It can be widely used in outdoor monitoring,industrial area and residential air quality monitoring,etc[2].

1 Principle and experimental setup

1.1 Principle and calibration model

In order to measure the concentration of six kinds of pollutants in ambient air,the appropriate particle and gas sensors are selected in the system.The automatic monitoring methods for the mass concentration of particulate matter include gravimetric method,micro oscillation balance method,light scattering method,β-ray method and piezoelectric crystal method[10].In this paper,PM measurement uses the particle sensor OPC-N2 produced by Alphasense Company.OPC-N2 is an optical particle counter based on the principle of light scattering.It can calculate the particle number concentration and particle size by measuring the single scattered light particles carried by the laser beam in the sampled air stream.This method is based on the Mie scattering theory to measure the concentration of particles.When the light is irradiated on the particulate matter,it will generate scattered light.Under the condition of a certain particulate matter,the intensity of the scattered light is proportional to the concentration of the particulate matter.By measuring the intensity of the scattered light,combined with the corresponding conversion factor,the concentration of particulate matter can be obtained[11].Since the measurement of particulate matter is susceptible to humidity,the mass of the particle is related to the hygroscopicity through the κ-Köhler theory[12,13],as shown in Eq.(1)

whereahis relative humidity,also known as water activity;mis the mass of wet aerosol andm0is the mass of dry aerosol.In this paper,mis also the particulate mass measured by OPC-N2,andm0is the particulate mass measured by the NAAQMS;ρwis the density of water,and ρpis the density of dry particles,which is 1.65 g·m-3by default;kis the correction factor,which can be obtained by nonlinear fitting the humidity curve.So Eq.(1)can be converted to Eq.(2)

Taking the mass concentration ratio of OPC-N2 and NAAQMS as dependent variable and relative humidity as independent variable,the correction factorkcan be obtained by nonlinear curve fitting according to Eq.(2).The mass of aerosolm0can be calculated when the mass of wet aerosolmis known,which is the particulate matter mass after humidity correction.

According to different working principles, gas sensors can be divided into optical, electrical and electrochemical types.Since the toxic and harmful gases that the designed ambient air quality monitoring system needs to monitor are oxygen-containing elements(CO,NO2,O3,SO2),and considering the comprehensive consideration of cost and performance,we choose the electrochemical gas sensor,when the gas electrochemical reaction occurs on the electrode,there will be a current.The current is directly proportional to the gas concentration,and can be converted into gas concentration information by measuring the current size[14-16].This type of gas sensor has advantages of low cost,linear output,high sensitivity and low power consumption.However,the problems is that measurement results from electrochemical gas sensor are greatly affected by temperature.By establishing multiple linear regression model,the measurement results can be corrected.The multiple linear regression model is shown in Eq.(3)

whereZis the concentration of gas pollutants measured by NAAQMS,Xis the measurement result of the electrochemical gas sensor,andYis the ambient measurement temperature.Constant valuesA,BandCcan be determined by multiple linear regression,and the corrected concentrationZof gas pollutants can be calculated by taking the measuredXandYinto the formula.

1.2 Experimental setup

The main control chip of the system is stm32f103vet6,the PM sensor is OPC-N2,and the SPI communication mode is adopted.The analog output voltage signal of electrochemical sensor is sampled by 8-channel ADC chip(ADS1256).RS-485 communication mode is adopted for wind speed,wind direction,temperature,humidity and sensor.GPS module uses RS-232 serial communication to output positioning information.The hardware system design block diagram of the system is shown in Fig.1,and the air quality monitoring system is shown in Fig.2.

图1 硬件系统设计框图Fig.1 Hardware system design block diagram

图2 微型空气质量监测系统实物Fig.2 Picture of a miniature air quality monitoring system

2 Experimental results and discussion

2.1 Experimental results

The site of the first experiment is about 300 meters away from NAAQMS,and the instrument is installed on a column with 1.5 m above from the ground,as shown in Fig.3.The site of the second experiment is selected to be about 20 meters away from NAAQMS,which is located on a roof of building,as shown in Fig.4.

图3 第一次户外实验场地Fig.3 Site of the first outdoor experiment

图4 第二次户外实验场地Fig.4 Site of the second outdoor experiment

Since the data released by NAAQMS is an hourly average,the system built in this paper collects data every minute.The bad points are removed from the raw data of outdoor experiments,and the average value is calculated.The hourly average of outdoor experimental data was compared with the data released by NAAQMS.The comparison results between the first experimental data and NAAQMS are shown in Fig.5,and Fig.6 shows the comparison of the first experimental data after correction.In Fig.5 and Fig.6,Ccois the concentration of CO;CN-COis the concentration of CO measured by NQQMS;CM-COis the concentration of CO measured by MAQMS;CM-N-PM2.5is the ratio of the concentration of PM2.5measured by MAQMS and NAAQMS(abbreviations in the Fig.5-Fig.8 have similar meanings).

图5 第一次实验数据对比。(a)CO折线图;(b)CO散点图;(c)NO2折线图;(d)NO2散点图;(e)O3折线图;(f)O3散点图;(g)SO2折线图;(h)SO2散点图;(i)PM2.5折线图;(j)PM2.5散点图;(k)PM10折线图;(l)PM10散点图Fig.5 Comparison results of the first experimental data.(a)CO line chart,(b)CO scatter plot,(c)NO2 line chart,(d)NO2 scatter plot,(e)O3 line chart,(f)O3 scatter plot,(g)SO2 line chart,(h)SO2 scatter plot,(i)PM2.5 line chart,(j)PM2.5 scatter plot,(k)PM10 line chart,(l)PM10scatter plot

图6 校正后的第一次实验数据对比。(a)CO折线图;(b)CO散点图;(c)NO2折线图;(d)NO2散点图;(e)O3折线图;(f)O3散点图;(g)PM2.5非线性曲线拟合图;(h)PM10非线性曲线拟合图Fig.6 Comparison of the first experimental data after correction.(a)CO line chart,(b)CO scatter plot,(c)NO2 line chart,(d)NO2 scatter plot,(e)O3 line chart,(f)O3 scatter plot,(g)PM2.5 nonlinear curve fitting diagram,(h)PM10 nonlinear curve fitting diagram

图7 第二次实验数据对比。(a)PM2.5折线图;(b)PM2.5散点图;(c)PM10折线图;(d)PM10散点图;(e)SO2折线图;(f)SO2散点图;(g)NO2折线图;(h)NO2散点图Fig.7 Comparison results of the second experient data.(a)PM2.5 line chart,(b)PM2.5 scatter plot,(c)PM10 line chart,(d)PM10 scatter plot,(e)SO2 line chart,(f)SO2 scatter plot,(g)NO2 line chart,(h)NO2 scatter plot

图8 校正后的第二次实验数据对比。(a)PM2.5非线性曲线拟合图;(b)PM10非线性曲线拟合图;(c)PM2.5折线图;(d)PM2.5散点图;(e)PM10折线图;(f)PM10散点图Fig.8 Comparison of the second experimental data after correction.(a)PM2.5 nonlinear curve fitting diagramt,(b)PM10 nonlinear curve fitting diagramt,(c)PM2.5 line chart,(d) PM2.5 scatter plot,(e)PM10 line chart,(f)PM10 scatter plot

According to the calculation,the hourly average value of outdoor experimental results is compared with NAAQMS.The correlation coefficients of CO,NO2,O3and PM2.5were 0.72,0.5,0.88 and 0.32,respectively.The correlation coefficients of SO2and PM10were less than 0.1.Considering the influence of temperature and humidity,the first experimental data is corrected and shown in Fig.6.The total measurement data of the first experiment is 780 groups.By establishing a multiple linear regression model of the first 200 groups of gas pollutant measurement data and temperature values,the last 580 groups of data are corrected.The corrected results are shown in Fig.6(a)-(f).Through temperature correction,the measurement results of CO,NO2,and O3have a smaller difference compared to NAAQMS.The measurement data of SO2are not fitted by multiple linear regression,because the correlation of measurement data is very poor.Fig.6(g)-(h)is the result of non-linear fitting of PM2.5and PM10according to Eq.(2).Due to a large number of data deviating from the fitting red curve,it can be seen from Fig.5(i)and Fig.5(k)that the measurement results of OPC-N2are lower than those of NAAQMS,which is inconsistent with Eq.(2).Because the particles collected by the particle sensor OPC-N2contain a lot of water,the concentration of particles in wet environment will be greater than that in dry gas[2].

Since the first experiment,the correlation of other parameters is very poor except O3and CO.In the second experiment,the experimental site was changed,focusing on the measurement of PM2.5,PM10,SO2and NO2.The measurement results of the second experiment are shown in Fig.7,in which the correlation coefficients of PM2.5,PM10and NO2are 0.68,0.71 and 0.91 respectively,and the correlation coefficient of SO2is still very poor,less than 0.1.Fig.8 shows a comparison of the data of the second experiment after correction.Due to the short sampling period of the second experiment,the measurement results of SO2and NO2were not corrected by multiple linear regression fitting.Through humidity correction,the accuracy of PM measurement results is improved,the correlation coefficient is increased to more than 0.8,and the difference between PM measurement results and NAAQMS measurement results is reduced.

2.2 Discussion

The correlation coefficients calculated before and after the two experiments are shown in the Table 1.The blank part is due to the problems of the experimental results or the sampling time is too short to make correction analysis.

表1 相关系数比较Table 1 Comparison of correlation coefficient

Analysis of the data from the two experiments shows that,except SO2,the collected data of CO,NO2,O3,SO2,PM2.5and PM10have better correlation with NAAQMS data,and the change trend is basically the same.Due to the low concentration of SO2in the atmosphere,the SO2concentration of the selected comparison station(NAAQMS)is less than 10 μg·m-3most of the time.Meanwhile,the measurement of SO2sensor is easily affected by the cross interference of other gases,especially NO2and CO,so the accuracy of measurement in low concentration environment cannot be guaranteed[17,18].The difference between PM2.5and PM10in the measurement results of the two experiments is large,which is due to the great influence of the experimental site on PM measurement.The site of the first experiment is located on the ground with shielding around,while the site of the second experiment is located on the roof of a building,which is relatively open and more closer to NAAQMS.

Since the selected sensor module is different from NAAQMS,it is impossible to completely match the data of national control station.Only relying on the algorithm correction and factory calibration provided by sensor manufacturers cannot achieve high accuracy of gas monitoring.The starting point of our design of outdoor multi parameter ambient air quality monitoring system is to supplement the short board of NAAQMS,which is sparsely distributed and unable to make timely response to the high concentration of pollutants in a small area.From the experimental data,the air quality monitoring system can real-time monitor the concentration changes of pollutants in the coverage area.Compared with NAAQMS,it has good consistency,and can meet the needs of real-time monitoring and timely warning of pollutant emission in factories and other areas[4].

3 Conclusion

A new micro air quality monitoring system based on multi-sensor fusion technology is developed to accurately measure the concentration of CO,NO2,O3,PM2.5and PM10in the ambient air.Based on the multi-sensor fusion technology,the algorithm correction is made to the mass concentration monitoring of particulate matter and gas pollutants,which improves the accuracy of measurement data to a certain extent.At the same time,the micro air quality monitoring system built in this paper is small in size and low in cost,which is suitable for large-scale grid layout in towns,factories and other regions.In the future,the work in this paper needs to be improved,especially for the measurement of SO2.By building a more accurate calibration platform for electrochemical gas sensors,simulating the environment with different temperature and humidity and gas pollutant concentration in the laboratory,and combining curve fitting,linear regression,neural network and other methods to perform secondary correction on the electrochemical sensor,we can further improve the measurement accuracy of this system.