A Multifaceted Framework for Predicting Ambient Air Pollution in the National Capital Region

Abstract

Air quality forecasting models anticipate and control pollution concentrations quickly. In this research, two diversified datasets are analyzed and modeled using the proposed novel Hybrid Forecasting Model (HFM). PM10, PM2.5, NO, and NO2 pollutant concentrations for the National Capital Region, Delhi, were collected from the Central Pollution Control Board as Dataset 1. A Raspberry Pi hardware integrated with an MCP3008 Analog-to-Digital Converter and pollutant sensors like MQ135 and GP2Y1010AU0F was assembled to collect PM2.5, CO, and NH3 pollutant concentrations at the study site as Dataset 2. The results obtained for the proposed HFM are compared with three existing algorithms in terms of R-squared and Mean Squared Log Error. The novelty of this work is to test the performance of the algorithm by executing it on a computer with an Intel 8265U processor and also on a Raspberry Pi integrated with an MCP3008 Analog-to-Digital Converter. The experimental results indicate that the Raspberry Pi requires approximately 3.37% of the power consumed by the Intel 8265U processor. Conversely, the Intel 8265U processor achieves approximately 86.67% lower latency, thereby delivering significantly faster computational performance compared to the Raspberry Pi.