A stochastic model of particulate matters with AI-enabled technique-based IoT gas detectors for air quality assessment

Abstract Monitoring air quality in urban and industrial environments and estimating exposure to particulate matter (PM) pollution concentrations are critical issues that affect human health. Because of aerosols (suspended particles), PM is mostly observed near the surface and thus can be inhaled. To predict the modeling of micro-to-nano-sized particle suspensions, this study presents a stochastic model in environmental dynamics with internet of things (IoT) gas detectors based on an artificial intelligence (AI)-enabled technique; the model can determine floating fine PM dispersion in a city to assess and monitor air quality. The factors that influence the prediction are weather- and air pollution-related data, such as humidity, temperature, wind, PM2.5, and PM10. In this study, these factors have been considered at 7 measuring stations across the urban region in Taipei City, Taiwan, from 2013 to 2018. A nonlinear autoregressive network with exogenous inputs model is constructed using estimated states to investigate approaches for identifying PM; the model can be a state–space self-tuning stochastic model for predicting unknown nonlinear sampled data. The results indicate that a satisfactory agreement was obtained using a normalized root mean square deviation, with small values of 0.0504 and 0.0802 for PM2.5 and PM10, respectively. Accordingly, this study presents that the time-domain causality between PM and the atmospheric environment can be constructed using discrete-time models that can be satisfactorily implemented in developing different air quality monitoring systems for the long-term prediction of air pollution.