SolarDecomp: A Web App for Decomposing Solar Data for Spectrally Selective Building Simulation

Solar radiation plays an important role in solar architecture design. It not only determines the optical regime of the building envelope, but also influences the heating and cooling loads of the building. To understand the impacts of solar radiation on building thermal and lighting performance, computational analysis is essential. However, conventional modeling tools only take broadband solar radiation into consideration, which limits the modeling accuracy since window glazing is spectrally dependent. In this sense, we developed a new solar decomposing tool that can separate major solar irradiation components such as VIS, NIR in GHI, DNI and DHI, by using machine learning algorithms such as extreme boosting regression tree (XGBoost). The predictors are meteorological parameters that already exist or can be easily derived from traditional weather files (such as TMY3). Model performance is validated by NREL Solar Radiation Research Laboratory’s spectral solar dataset. Integrating these decomposing models, a web app (SolarDecomp) has been developed to enable researchers and designers import solar spectral data into existing building simulation programs for specific simulations in terms of spectrally selective design components.