Parametrized Modelling and Automatic Simulation-Driven Building Energy Prediction in Early Design

Building energy prediction, a critical status in the energy saving process, attracts much attentions in academia. In this study, an efficient building energy prediction approach for early design is proposed to enlarge the scope of application, accelerate the speed of energy prediction result feedback and reduce the operational complexity. The approach follows a two-stage framework. The Stage 1 is to establish mappings between building information model and the energy consumption level calculated by the EnergyPlus. A rapid 3D building modelling is also developed to satisfy the requirement of wide applicability in construction project. In light of the thermal simulation by EnergyPlus, a large variety of parameters in these model samples can be extracted as the database of the energy consumption prediction. With the help of GA-NN (Genetic Algorithm-Neural Network), a prediction model, in which the building design parameters are the inputs only needed, is developed in Stage 2 to predict the energy consumption level. A case study is carried out to verify the availability of the proposed two-stage approach. This study addresses the improvement in both generalization and efficiency and urges a wide diffusion of passive design approach, which can provide designers a reliable building energy consumption prediction tool.