Evaluation and calibration of a low-cost particle sensor in ambient conditions using machine-learning methods

Abstract

Abstract. Particle sensing technology has shown great potential for monitoring particulate matter (PM) with very few temporal and spatial restrictions because of its low cost, compact size, and easy operation. However, the performance of low-cost sensors for PM monitoring in ambient conditions has not been thoroughly evaluated. Monitoring results by low-cost sensors are often questionable. In this study, a low-cost fine particle monitor (Plantower PMS 5003) was colocated with a reference instrument, the Synchronized Hybrid Ambient Real-time Particulate (SHARP) monitor, at the Calgary Varsity air monitoring station from December 2018 to April 2019. The study evaluated the performance of this low-cost PM sensor in ambient conditions and calibrated its readings using simple linear regression (SLR), multiple linear regression (MLR), and two more powerful machine-learning algorithms using random search techniques for the best model architectures. The two machine-learning algorithms are XGBoost and a feedforward neural network (NN). Field evaluation showed that the Pearson correlation (r) between the low-cost sensor and the SHARP instrument was 0.78. The Fligner and Killeen (F–K) test indicated a statistically significant difference between the variances of the PM2.5 values by the low-cost sensor and the SHARP instrument. Large overestimations by the low-cost sensor before calibration were observed in the field and were believed to be caused by the variation of ambient relative humidity. The root mean square error (RMSE) was 9.93 when comparing the low-cost sensor with the SHARP instrument. The calibration by the feedforward NN had the smallest RMSE of 3.91 in the test dataset compared to the calibrations by SLR (4.91), MLR (4.65), and XGBoost (4.19). After calibrations, the F–K test using the test dataset showed that the variances of the PM2.5 values by the NN, XGBoost, and the reference method were not statistically significantly different. From this study, we conclude that a feedforward NN is a promising method to address the poor performance of low-cost sensors for PM2.5 monitoring. In addition, the random search method for hyperparameters was demonstrated to be an efficient approach for selecting the best model structure.

Evaluation of optical particulate matter sensors under realistic conditions of strong and mild urban pollution

“…As expected, 11-D has the ability to count particles below 300 nm, which appear in the greatest numbers. Counting efficiency of OPC-N2 is investigated in laboratory conditions using PSL particles in Sousan et al (2016a), and the results were good for particles larger than 0.8 µm, while for particles with a diameter of 0.5 µm OPC-N2 the device showed a lower detection efficiency (the detection limit of OPC-N2 is 0.38 µm). In our realistic scenario, the dominant contribution to the mass comes from particles much smaller than 0.8 µm (Figs.…”

“…Due to all the above-mentioned factors, it is always interesting to check how OPS perform in different realistic scenarios. Numerous papers deal with laboratory calibrations and outdoor evaluations of OPS (Karagulian et al, 2019;Borghi et al, 2018;Chatzidiakou et al, 2019;Magi et al, 2020;Sousan et al, 2016b;Malings et al, 2020;Kelly et al, 2017;Sayahi et al, 2019;Crilley et al, 2018;Zheng et al, 2018;Tasic et al, 2012;Cavaliere et al, 2018;Mukherjee et al, 2017;Sousan et al, 2016a;Zhang et al, 2018;Holstius et al, 2014;Badura et al, 2018). Reported results vary depending on the composition of particulate matter pollution, range of concentrations and meteorological factors.…”

Field Evaluation of Low-Cost Particulate Matter Sensors in Beijing

“…There are limited studies and analyses on the performance of such low-cost sensors and their long-term comparisons in China. Most of the studies on low-cost PM sensors are based on field and laboratory assessments, which were conducted in countries and regions with good air quality (PM 2.5 concentration < 50 µg/m 3 ), such as the United States and Europe [ 16 , 22 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 ]. Zheng et al [ 21 ] and Gao et al [ 36 ] tested the performance of low-cost sensors in urban environments with high PM concentrations.…”

A Review of Low-Cost Particulate Matter Sensors from the Developers’ Perspectives

“…Very recently, more advanced calibration models have been tested. A few months ago, Zaidan et al, have proposed and tested more modern approaches like NARX (nonlinear autoregressive with exogenous input neural networks) and even deep learning models (LSTM-Long Short Term Memory) comparing them to conventional approaches in a transfer learning setup [110]. However, they reported no performance advantage in using LSTM with respect to the NARX approach that scored the best.…”