RECOMMENDATION SYSTEMS FOR THE BUILT ENVIRONMENT: PREDICTING AND MANAGING THERMAL COMFORT & ENERGY EFFICIENCY

Abstract

Recommendation systems are a form of Artificial Intelligence and Machine Learning (AI/ML) that are able to learn the preferences of users or pre-emptively guess the preferences of new users who have never used the application before and recommend an item or a service to the users. AI/ML are increasingly commonplace tools that are able to benefit users of applications that they are implemented in, ranging from e-commerce to video and music entertainment services.

This paper investigates the possibilities of applying AI/ML recommendation systems to the context of the built environment and the advantages that would occur from such implementation, focusing on determining and improving the thermal comfort of occupants in a building and estimating the improved energy efficiency in a building when there is a higher overall thermal comfort for occupants.

When applied to a dataset consisting of 30 participants and benchmarked against the most commonly used method of evaluating thermal comfort (PMV/PPD method), there is evidence that recommendation systems assisted by AI/ML were able to accurately determine the thermal comfort levels of occupants by at least 50% or more on average. Recommendation systems were also able to determine the change in thermal environment required (prefer warmer, no change, prefer cooler) to achieve thermal comfort.

These results illustrate that with further refinement, there is great feasibility for recommendation systems in the built environment for the purpose of increasing the comfort of occupants. A discussion is provided about the potential increase of thermal comfort and energy efficiency from implementing recommendation systems in buildings, an outlook on other practical uses other than thermal comfort, and future research that could be done to refine the study of recommendation systems for the built environment.