Performance Comparison of Nonlinear Pre–Calibrate Low–Cost PM2.5 Sensors Using an SPS30 Reference

Abstract

This research presents a performance comparison of low–cost particulate matter (PM2.5) sensors, widely used in Internet of Things (IoT) applications for air quality monitoring. Since sensor calibration is often costly, this study proposes a cost–reduction strategy by applying pre–calibration before full calibration. The SPS30 was selected as the primary reference device due to its combination of low cost and near–regulatory–grade performance. Unlike other low–cost sensors, the SPS30 benefits from factory calibration against reference instruments (e.g., TSI DustTrak DRX 8533, OPS 3330), and it has demonstrated very low intra–model variability (<1.5% for PM2.5) and strong correlations across all concentrations with Federal Equivalent Method (FEM) instruments. It is also MCERTS–certified (UK Environment Agency), confirming its compliance with PM2.5 monitoring standards. To validate the methodology, the SPS30’s accuracy was additionally examined using an air purifier in the test setup. A nonlinear mathematical model was then applied to calibrate commonly used sensors, including the Plantower PMS series (PMS7003, PMS5003, PMS3003) and SDS011. Experiments were conducted in an indoor environment at 33 ± 1°C and 69 ± 4% relative humidity. The results showed coefficient of determination values of 0.98, 0.98, 0.96, and 0.88, with root mean square error values of 1.2, 1.47, 1.84, and 3.26 for the PMS7003, PMS5003, PMS3003, and SDS011, respectively. The findings indicate that low–cost sensors, particularly the PMS7003 and PMS5003, can achieve high measurement accuracy when combined with appropriate pre–calibration and a suitable reference device. The SDS011 also demonstrated consistent performance. In addition, applying a nonlinear model reduces costs and enhances sensor reliability. For initial deployment, pre–calibration lowers expenses by approximately one–third compared to full calibration, while pairwise pre–calibration for recalibration can substantially reduce or even eliminate recurring calibration costs during long–term operation and maintenance. These results highlight the practicality of deploying low–cost sensors in air quality monitoring applications.