Improving Performance of Low-Cost Sensors Using Machine Learning Calibration with a 2-Step Model

Abstract

AbstractWith the advancement of air pollution management, low-cost sensors are increasingly being used in air quality monitoring, but the data quality of these sensors is still a major source of concern. In this paper, a two-step model was created to refine the calibration process for low-cost PM sensors using data from five standard air monitoring stations in Sofia. At first, we calibrated five sensors located next to the standard instruments using five different supervised machine learning models and then the ANN-final model with anomaly detection completed the results. The ANN-final model improved the \(R^2\) values of the PM10 determined by low-cost sensors from 0.62 to 0.95 as compared to standard instruments, which confirmed the effectiveness of the calibration. This calibration model was further applied and evaluated on other sensors of the network, located in a 500m radius near the standard instruments.KeywordsAir qualityCalibrationMachine learning