Data-driven thermostats : for feedback, comfort, and reliability

Abstract

In buildings, air conditioning consumes a significant proportion of the aggregate electricity bill. Specifically, in residential and small-scale commercial buildings, people prefer to use room-level ACs (AC stands for Air Conditioning in this thesis) with an in-built thermostat. As thermostat settings conceptually govern AC energy consumption and user comfort; over the years, researchers spent a significant amount of time and effort in enhancing thermostats to ensure optimal usage of ACs. By analyzing occupants' behaviour, today, smart thermostats, with multiple sensory abilities, can automatically adjust the set temperature to maximize both - energy savings and user comfort.

While thermostats are smart and ubiquitous, they often rely on occupants' dynamic schedule for the automated control of the set-point temperature. For a typical home, where everyone follows a particular routine, any deviation in daily schedule often leads to user discomfort. In addition to that, smart thermostats neither consider spatial variations across the buildings, nor temporal variations, such as climate change, while changing the set temperature. Subsequently, even today, the expensive thermostats are confined to automated temperature variation, with a limited scope of boosting the energy savings and enhancing the user comfort. In this dissertation, we address these concerns and introduce Data-Driven Thermostats to make AC experience efficient and comfortable for the users.

First, we propose PACMAN that monitors room temperature to ensure tenants' participation during AC usage by providing actionable energy-feedback. Next, we recommend a Comfort-Energy Trade-off (CET) knob, realized through an optimization framework, to allow users to balance their comfort and savings without worrying about the right set temperature. Our study indicates that such a knob can reduce residents’ discomfort by 23% and save 26% energy. Third, we investigate the impact of occupancy prediction errors on occupants' comfort and total energy consumption of a building. Finally, we propose Greina - to continuously monitor the readily available ambient information from the thermostat and timely report refrigerant leaks through the coils (or valves) of a refrigeration unit. Such leaks waste significant energy, risk occupants' health, and affect user comfort. Our methods are novel, scalable, and more effective than the state-of-the-art smart thermostats.