Application of a New CO₂ Prediction Method Within Family House Occupancy Monitoring

Abstract

The article describes the application of Python for verification of a newly designed method of CO2 prediction from measurements of indoor parameters of temperature and relative humidity within occupancy monitoring in real conditions of a family home. The article describes the implementation of non-electric quantities (indoor CO2 concentration, indoor temperature, indoor relative humidity) measurement in five rooms of a family home (living room, kitchen, children’s room, bathroom, bedroom) using Loxone technology sensors. The IBM IoT (Internet Of Things) was used for storing and subsequent processing of the measured values within the time interval of December 22, 2018, to December 31, 2018. The devised method used radial basis function (artificial neural networks (ANN)) mathematical method (implementation in Python environment) to perform accurate predictions. For further increase of the accuracy and reduction of prediction noise from the obtained course of the predicted signal, multiple variations of the LMS adaptive filter algorithm (Sign, Sign-Sign, Sign-Regressor) were used (implemented in the MATLAB SW tool). The accuracy of the newly proposed CO2 concentration prediction method exceeds 95% in the selected experiments.

Keywords

• Artificial neural network (ANN)
• intelligent buildings (IB)
• Loxone
• monitoring
• occupancy
• prediction