| As one of the near-surface air pollutants,NO2 is a major source of pollutants such as ozone,secondary aerosols and acid rain,which seriously affects the health of human beings and the balance of the entire ecosystem.Therefore,in order to accurately analyze pollution sources and carry out pollution prevention and control work smoothly,it is urgent to monitor atmospheric NO2.Due to the high cost and large volume of traditional optical detection equipment,the new four-electrode NO2 electrochemical sensor has been valued by more and more researchers due to its low cost,small size,and portability.However,such sensors are seriously interfered by environmental factors when working,so it is necessary to study the calibration method of electrochemical sensors to detect atmospheric NO2 concentration.In this paper,an atmospheric trace NO2 detection system based on electrochemical sensor was constructed,and its sensitivity was measured to be 0.238mV/ppb,and the detection limit was lower than 2ppb.Based on the laboratory sample data and the first field measurement data,three calibration models of atmospheric NO2 concentration data were established.Firstly,differential correction algorithm and linear regression algorithm were used to calibrate atmospheric NO2 concentration.In view of the limited concentration calibration capability of these two methods,the nonlinear BP neural network algorithm came into being.The neural network model of atmospheric NO2 concentration prediction was established,including network structure,training function,model evaluation index setting,source data preprocessing and PCA analysis to select environmental variables.By changing the combination of four input variables and debugging the network model for several times,it was found that the model with working electrode,auxiliary electrode,temperature and relative humidity as input variables,training function as Trainlm and network structure as 4-7-1 had the best prediction effect,and the cross-validation determination coefficient reached 0.944 and the root-mean-square error was 2.2.In order to verify the calibration effect of three methods and the drift characteristics over time again,two atmospheric NO2 field experiments were carried out in the later stage,and the results were compared with the measurement results of CRDS system of blue laser light source at the same monitoring point.After a comprehensive comparison of the three calibration methods,it was found that the system data fitting effect using neural network model was the best,and the correlation of data sets during the two measurements reached 0.949 and 0.881,respectively.The calibration method showed that the sensor only showed a slight drift of about 5ppb with the change of time.The research results show that this method can effectively compensate the influence of environmental factors on the measurement of atmospheric NO2 by electrochemical sensors,and also provides a theoretical basis for concentration calibration of other electrochemical sensors.Figure[54]table[16]reference[65]... |
PDF Full Text Request