Occupancy modelling using data driven models

• Main Author: Lee, Gabriel Hanjie
• Other Authors: Soh Yeng Chai
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: https://hdl.handle.net/10356/140536
• Institution: Nanyang Technological University

Indoor occupancy information is key to office and home automation systems. It is used as an input for the control of indoor lighting systems [1] and Heat, Ventilation and Air-conditioning (HVAC) systems [2]. HVAC technology ensures constant supply of good quality air and thermal comfort for occupants to live and work using designed heating, filtration and ventilation systems. As our society steadily progresses towards a sustainable future by reducing ecological footprints, more emphasis and attention has been given to the issue of building energy optimization. Studies have also shown that around one-third of the energy consumed in buildings can be saved using occupancy-based control [3]. As such, a great amount of attention has been given to energy efficiency issues in designing and improving our buildings today. A conventional way to estimate the occupancy level in a particular room is to employ numerous sensors in order to completely capture the occupancy profile of the entire environment. Data collected from multi-camera videos coupled with pattern recognition technology can accurately estimate the number of indoor occupants, however, these methods require expensive hardware and are not often used due to their intrusive nature which brings privacy concerns. Thus, many non-intrusive and non-terminal-based types of sensors have been used for indoor occupancy estimation, such as pyro-electric infrared (PIR) sensors [4], ultrasonic sensors [5], and microphones [6]. The author will work on the collected data from surrounding parameters, such as temperature, humidity, air pressure and CO2 levels, and present a performance analysis on the models trained on these parameters.