Energy efficiency improvement with k-means approach to thermal comfort for ACMV systems of smart buildings

This paper proposes a novel methodology to predict thermal comfort states of occupants with k-means approach. The approach is embedded into an optimization problem, which is used to locate optimal operating conditions via Augmented Firefly Algorithm (AFA), for improving energy efficiency of buildings and maintaining satisfactory indoor thermal comfort states in the meantime. The neural networks models of energy, air temperature, skin temperature and skin temperature gradient have been implemented and verified. The prediction of thermal comfort states via k-means approach has been implemented and it is based on features of skin temperature and skin temperature gradient. The problem is formulated directly from the developed thermal comfort model and energy model. The formulated problem has been followed by optimizations of AFA approach, and the experimental results show that the energy efficiency can be improved by at least 21% while maintaining the indoor thermal comfort satisfaction of occupants, thus conforming to the objectives of a smart building.