Prediction Model for Personal Thermal Comfort for Naturally Ventilated Smart Buildings

Abstract

Smart City concept can be realized by smart buildings. For creating smart buildings smart home systems need to be built. Smart home systems should have appliances which can adjust their settings according to the real time conditions. This research paper aims to improve the design of non-air conditioned or naturally ventilated (NV) buildings by determining the conditions for improving the satisfaction in thermal comfort levels of building occupants. For this study various datasets from ASHRAE RP-884 database have been taken from different climate zones and different seasons. For this purpose an optimum feature set is identified which indicates the parameters which impact most on personal thermal comfort level. Supervised machine learning techniques such as Support Vector Machines (SVM) and Naïve Bayes Classifier have been used. For feature selection Boruta Feature Selection Method has been used. The experimental results show that in any climate zone or season, indoor and outdoor temperature and humidity are most important factors in determining thermal comfort. For hot humid climate, air speed is another important factor. The research paper presents a low cost solution for improving thermal comfort of building occupants by determining the minimum number of features.