2024

An AIoT-based Air Quality Monitoring System for Real-time PM2.5 Prediction in Urban Environments

ASEAN Journal of Scientific and Technological Reports

2024-12-14 | Journal article | Writing - original draft, Conceptualization, Formal analysis, Project administration, Methodology, Supervision, Visualization, Writing - review & editing

DOI: 10.55164/ajstr.v28i1.255168

Part of ISSN: 2773-8752

Contributors: Nuth Otanasap; Siriwich Tadsuan; Chanintorn Chalermsuk

URL

http://dx.doi.org/10.55164/ajstr.v28i1.255168

Contributors

Nuth Otanasap (Writing - original draft, Conceptualization, Formal analysis, Project administration, Methodology, Supervision, Visualization, Writing - review & editing, 0000-0002-8713-3951)

Siriwich Tadsuan (Author)

Chanintorn Chalermsuk (Author, 0009-0006-6589-0552)

Citation (formatted-apa)

Show citation

Abstract

This study aimed to develop an air quality monitoring and forecasting system focusing on PM2.5 using a combination of AI of Things (AIoT) technology. The system was designed to provide warnings of PM2.5 levels through a mobile application. Air pollution, particularly PM2.5, is a significant health concern globally, with Southeast Asia being heavily affected. Bangkok, Thailand, experiences high PM2.5 concentrations during cool weather. Existing research explores short-term PM2.5 prediction using AIoT. Still, there is a need for improved software, hardware, and ML algorithms for user-friendly mobile applications with real-time data access and health advisories. The system was installed on a building next to a main road in Bangkok. It collected data on PM2.5. The Air Quality Index (AQI) was used to categorize PM2.5 levels and their health impacts. Time series analysis with moving averages and the Random Forest algorithm were employed in advance for PM2.5 forecasting. A mobile application was developed to provide a user interface and data visualization. The MARF (Moving Average and Random Forest) model emerged as a success, achieving higher accuracy (average of 92.59%) for 1-hour advance forecasts compared to the Moving Average (MA) model (average of 84.16%). The developed system demonstrates the potential of AIoT for accurate PM2.5 monitoring and forecasting. Future research could explore more advanced ML algorithms and integrate additional environmental factors for enhanced forecasting accuracy.