Urban physics and outdoor thermal comfort for sustainable street canyons using ANN models for composite climate

Abstract

This study is focused on the development and validation of the Artificial Neural Network (ANN) model for the prediction of hourly values of thermal comfort indices, namely – Physiological Equivalent Temperature (PET) and the Universal Thermal Climate Index (UTCI). Two ANN models are developed – one using air temperature, wind speed, relative humidity and mean radiant temperature as meteorological input parameters along with date and time. Another uses air temperature as only meteorological input parameters and date and time. The applicability of ANN model was validated for peak winter (December – February) and summer months (March-May). The ANN model showed remarkable predictive ability through a high coefficient of correlation, low root mean square error (RMSE) and fitting values of true predicted rate (TPR), false alarm rate (FAR) and success index (SI). The predicted output from the validated ANN models is used to discuss the influence of optimum urban physics parameters on outdoor thermal comfort. This study establishes that air temperature can be used as the only microclimatic input parameter to predict PET and UTCI using ANN when recording other microclimatic parameters is difficult.