Empowering citizens : a comparative study of low-cost sensors for particulate matter monitoring

Details

Supervisors

Raskin, Jean-Pierre ; Toussaint, Sébastien

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil électromécanicien, à finalité spécialisée: mécatronique

Abstract

This thesis investigates the potential of low-cost air quality sensors for monitoring Particulate Matter (PM2.5 and PM10) concentrations, aiming to complement the high-cost, limited-coverage Reference Stations (RS) currently in use. The study explores the viability of these sensors by employing various calibration and validation techniques to enhance their accuracy and reliability, ultimately providing data of significant scientific and legal value. Air quality monitoring is crucial due to its profound impact on public health and the environment. Traditional high-cost Reference Stations (RS), while accurate, are sparse, highlighting the need for alternative solutions. Low-cost sensors offer a promising approach, capable of widespread deployment and continuous monitoring. This research evaluates the performance of several low-cost sensors, including AirGradient Outdoor monitors, Antilopinae MiniStations, and Sensor Community monitors, against a high-accuracy Reference Station (RS 41R001) in Belgium. The findings reveal that low-cost sensors tend to overestimate PM concentrations. However, their performance improves significantly with effective calibration methods such as Linear Regression (LR) and Random Forest Regression (RFR) models, which adjust sensor readings to align with established standards and account for environmental factors like Relative Humidity (RH). Data collected over a month shows high correlations between the same type of low-cost sensors, validating their potential to provide consistent data. This underscores the importance of cross-sensor calibration, which uses data from a master sensor to calibrate other sensors (slaves), thereby improving overall data accuracy.