IoT-Based CO2 Monitoring and Forecasting System to Prevent Transmission of COVID-19

Abstract

This paper presents the development of low-cost CO2 remote monitoring devices based on NDIR sensors, and the design of a Nonlinear Autoregressive Neural Network (NAR) that forecasts the indoor CO2 concentration in the short and medium term to avoid risks of SARS-CoV-2 contagion due to the accumulation of poor quality air previously breathed by other people. Different configurations of the NAR were analyzed, varying the number of layers, the number of neurons per layer and the number of input delays. The best network configurations predicted changes in CO2 concentration in an academic office up to a four-hour horizon with an RMS error around 30 ppm.