Assistant Professor Ali Ghahramani and PhD student Chen Kai from the Department of the Built Environment, College of Design and Engineering, National University of Singapore, have developed a smart thermostat algorithm called ComfortGPT.
ComfortGPT is a novel machine learning architecture trained on historical home thermostat interaction data from over 100,000 buildings spanning several years. Unlike traditional machine learning algorithms, ComfortGPT does not require training from scratch. Instead, it utilizes a transformer architecture to identify the best representative model from a library of pre-trained archetype models. As users interact with their thermostats, the system continuously updates itself, automatically preselecting new setpoints as outdoor conditions change. This enables ComfortGPT to be significantly more accurate than legacy models, accurately predicting individual comfort preferences with minimal user interactions and adapting to small changes in preferences over time.
By adjusting temperatures to align with individual comfort levels, ComfortGPT enhances personal comfort while reducing energy consumption in buildings by minimizing overheating and overcooling, supporting sustainable building management through efficient energy use.
For more information, refer to the following link: https://cde.nus.edu.sg/modelling-personal-comfort-a-smart-thermostat-that-learns-what-temperature-you-like/.
The department looks forward to their continued contributions in the years ahead!!!
