Machine learning-based Sobol sensitivity analysis for building energy assessment without Monte Carlo integral Article Swipe
YOU?
·
· 2023
· Open Access
·
· DOI: https://doi.org/10.26868/25222708.2023.1311
· OA: W4392628863
Sensitivity analysis would be very useful to prioritize energy-saving measures for both new and existing buildings. The variance-based Sobol sensitivity analysis can provide reliable sensitivity results at the expense of high computational costs. The BASS (Bayesian Adaptive Spline Surfaces) method is one of the efficient sensitivity analysis approaches to reduce computational time without Monte Carlo integrals. The performance of this type of sensitivity analysis is much less studied in building energy assessment. An office building is used to demonstrate the performance of BASS sensitivity analysis by combining the EnergyPlus program. Results indicate that the BASS method can provide robust and reliable sensitivity results compared to the conventional sampling-based Sobol sensitivity analysis. Sufficient data is required to obtain stable results of the BASS sensitivity analysis and the variations of total effects using the bootstrap technique.
