Field comparison of electrochemical gas sensor data correction algorithms for ambient air measurements

SENSORS AND ACTUATORS B-CHEMICAL

卷:327

Article

Liang, Yue^1,2,3,4  Wu, Cheng^1,2,3,4  Jiang, Shutong⁵  Li, Yong Jie⁶  Wu, Dui^1,3,4,7  Li, Mei^1,3,4  Cheng, Peng^1,3,4  Yang, Wenda^1,3,4  Cheng, Chunlei^1,3,4  Li, Lei^1,3,4  Deng, Tao⁷  Sun, Jia Yin^1,3,4  He, Guowen^1,3,4  Liu, Ben⁶  Yao, Teng^4,8  Wu, Manman⁹  Zhou, Zhen^1,3,4 

[1] Jinan Univ, Inst Mass Spectrometry & Atmospher Environm, Guangzhou 510632, Peoples R China

[2] Chinese Acad Sci, Guangzhou Inst Geochem, State Key Lab Organ Geochem, Guangzhou 510640, Peoples R China

[3] Guangdong Prov Engn Res Ctr Line Source Apportion, Guangzhou 510632, Peoples R China

[4] Guangdong Hongkong Macau Joint Lab Collaborat Inn, Guangzhou 510632, Peoples R China

[5] Shenzhen Soarabil Technol Co Ltd, Shenzhen 518057, Peoples R China

[6] Univ Macau, Fac Sci & Technol, Dept Civil & Environm Engn, Taipa, Macao, Peoples R China

[7] CMA, Inst Trop & Marine Meteorol, Guangzhou 510080, Peoples R China

[8] Hong Kong Univ Sci & Technol, Div Environm & Sustainabil, Hong Kong, Peoples R China

[9] Guangzhou Hexin Analyt Instrument Ltd Co, Guangzhou 510530, Peoples R China

关键词: Low-cost sensors;    Electrochemical gas sensors;    Air pollution measurements;    Data correction algorithm evaluation;   

DOI  :  10.1016/j.snb.2020.128897

来源: Elsevier

【 摘 要 】

Electrochemical gas sensors (ECGS) have gained substantial popularity in ambient measurements. Several data correction algorithms had been proposed to tackle the drifting response of ECGS due to environmental factors, but there is a lack of performance evaluation of these data correction schemes. To fill this knowledge gap, we conduct a comprehensive evaluation of these data correction algorithms using a large dataset from field comparisons. The dataset covered three commonly used gas pollutants, including CO, NO2 and O-3 measured by both ECGS and reference instruments, with a time resolution of 1 min and a duration of 6 months. Taking advantage of this large dataset, the performance of 8 different data correction schemes (2 new algorithms and 6 algorithms from the literature) was benchmarked by a set of evaluation metrics using raw signals from ECGS (nA level currents from the working and auxiliary electrodes). Eight scenarios were considered to examine the robustness of correction algorithms in response to different training and evaluation data period configurations. In addition, the bias dependence on temperature, RH, target gas levels and cross-sensitivity by different correction algorithms was investigated. Recommendations on data correction scheme selection are provided based on the comparison results.

【 授权许可】

Free