LAUSR

Abstract Green roofs have been used widely in tropical countries due to their energy-saving and passive cooling effect. However, the model to calculate a thermal performance evaluation metric has yet to be established, which has hindered the inclusion of green roof in the envelope thermal performance standards in building regulations. This paper aims to develop a model to predict Roof Thermal Transfer Value (RTTV) of green roofs under a tropical climate. Firstly, the efficiency and accuracy of hygrothermal transfer model of green roof was validated against field data, and annual heat gain through four types of green roofs was simulated. Subsequently, the RTTV of four green roofs was calculated based on linear interpolation between RTTV of a common bare roof and corresponding annual heat gain, which ranged from 2.29 to 2.49 W/m2. Given the complexity and time-consuming of RTTV calculation by the hygrothermal transfer model, three types of regression models were developed based on the dataset of four significant design factors and corresponding values of RTTV. And it was found that multilayer perceptron (MLP) regression method had a better performance than multiple regression (MR) and random forest (RF) method. Finally, the equivalent thermal resistance of plant layer was calculated based on the RTTV model and other five simplified calculation methods, and the implicit reason that differences exist among them were discussed. The conclusions drawn in this paper can provide a methodology for a quick evaluation of thermal performance of green roof during the early stages of design.