Volume 16: Low Carbon Cities and Urban Energy Systems: Part V

Deep Learning-based Occupancy Behaviour Approach Towards the Improvement of the Indoor Air Quality Within Building Spaces Paige Wenbin Tien, Murtaza Mohammadi, Fangliang Zhong, John Kaiser Calautit, Jo Darkwa, Christopher Wood

Abstract

This study presents a vision-based deep learning approach for detecting and recognising occupant’ activities and window opening behaviour to help control the heating, ventilation, and air-conditioning (HVAC) system according to space’s actual thermal and ventilation requirements. A convolutional neural network (CNN) model was developed, trained, and deployed to a camera for real-time detection. The results of an experimental test within the case study building indicated an overall detection accuracy of 92.72% for occupancy activities and 87.74% for window operations. Real-time detection and recognition provided the generation of the deep learning influenced profiles (DLIP) used as input for building energy simulation to evaluate the impact of the approach on energy demand and indoor air quality. The present work assesses the importance of the proposed approach for predicting indoor air quality and comfort while optimising building HVAC operations to provide an effective demand-controlled ventilation strategy.

Keywords Deep learning, Building energy management, Window and occupancy detection, HVAC systems, Building energy performance, Indoor air quality