HVAC energy use is typically the largest energy load in low-income households, which have fewer plug and appliance loads and suffer from substandard HVAC systems and poor building envelope. Connected thermostats, with features like remote control and machine learning, can be a cost-effective strategy to reduce HVAC energy use, but these devices are prohibitively expensive for low-income households. This research was part of a larger project to develop and test an affordable connected thermostat designed for low-income households. This paper presents results from usability testing of the thermostat mobile app and two iterations of the thermostat hardware with a total of 27 households from two low-income communities in Sacramento, California. We present findings regarding user preferences for using the app versus the hardware for different functionalities and the value of remote control via Bluetooth rather than Internet. Usability implications of trade-offs between functionality, cost, and design are discussed.

Key words: HVAC, low-income households, connected thermostats, machine learning, Bluetooth