A Machine Learning Approach for Optimal Ventilation based on Data from CO2Sensors

ABSTRACT

Window ventilation is important in everyday life. The COVID-19 pandemic in particular has shown that air exchange is necessary to minimize the spread of viruses. Efficient ventilation can be supported with the help of sensors and intelligent data processing. A CO2 sensor, for example, can be used to measure CO2 levels and, together with IoT hardware, indicating when ventilation is needed. By combining these components together with algorithms, an assessment of such window ventilation can be made. This paper presents a measurement setup to measure CO2. The measured values in the form of time series are used in the setup to learn the time points ventilation. Two approaches are taken to analyze these time series. The first approach is based on the simple K-Means and K nearest neighbors algorithm, the second approach uses Dynamic Time Warp (DTW) Barycenter Averaging (DBA). Both approaches are compared in this work in the detection of ventilation events. © 2022 IEEE.