A Decomposition Method for Energy Prediction Metamodels and Surrogate Models in Early Design Stage of Buildings

Abstract

Early design is an important design phase for energy efficient design, as decisions about buildings have the highest impact on final performance at the lowest cost. A judicious selection of building form and variables in the early stages of design, especially in the conceptual design phase, can help to improve design performance early in the design process at minimal cost. In this study, a decomposition method is developed that can be used in building energy per formance evaluation by converting the building form into simple basic forms. The aim of the study is to develop a geometry-based energy estimation method for surrogate and metamodels to be used in the early design phase of buildings. The developed method is applied to certain cases of design variation under spec ified boundary conditions and the accuracy of heating and cooling energy loads are calculated with simulated energy models of these cases. As a result of the calculation, accuracy rates between 92.64% and 99.74% founded. This paper pro poses a prediction model with geometric identification method for an innovative geometry-based surrogate modelling method. This method also provides a way for artificial intelligence-based prediction models used in surrogate models to create a data set and can be used in the training in future works.