Heru Nurwarsito, Taufik, Rekyan Regasari Mardi Putri, Muhammad Azhar Hibatulloh
This article describes a new air quality monitoring and forecasting system based on the Internet of Things (IoT) technology that uses the Random Forest algorithm to forecast pollutants from waste temporary storage. While previous IoT monitoring systems focused on methane (CH4) and carbon dioxide (CO2), the system under development also measures ammonia (NH3), enabling more comprehensive evaluation of landfill gas emissions (LFG). The system hardware also includes MQ-9, MQ-137, and MQ-135, which are used to detect CH4, NH3, and CO2, and is combined with the ESP32 and Wemos D1 Mini microcontrollers for data acquisition and transmission. The data collected by the sensors is uploaded to Google Sheets, which are later used to train the forecasting model. The collected dataset comprised 1300 assigned sensor readings over 30 days. The system was assessed by comparing the models’ performance against the general fundamentals of regression. The reviewed MAE of methane was 9.2769 ppm. The reviewed MAE of ammonia gas was 0.0000005 ppm, and the reviewed MAE of carbon dioxide was 0.8227 ppm. RMSE values were 15.9674 ppm for methane, 0.00000079 ppm for ammonia, and 1.2783 ppm for carbon dioxide. MAPE values were 4.48% for methane, 2.20% for ammonia, and 2.33% for carbon dioxide, while R2 values reached 0.9649, 0.9899, and 0.9752, respectively. The system achieved prediction accuracies above 95% for all pollutants, with an average data transmission delay of 287.31 ms, demonstrating strong potential for real-time landfill air quality monitoring. IJASEIT is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.
Faculty of Computer Sciences, University of Brawijaya, East Java, Malang, Indonesia; California Polytechnic State University, CA, United States
