No abstract

No abstract

No abstract

The paper investigates the use of surrogate models for probabilistic building performance simulation that can be used for multiple applications across life cycle phases. The workflow presented aims to highlight a possible continuity among design and operation phase practices, in order to contribute to the reduction of the gap between simulated and measured performance, considering in particular the uncertainties caused by users' behaviour. Design phase simulation work is generally affected by relevant temporal and economic constraints and, consequently, a successful approach should enhance current design practices and implement new features which have to be automated, to decrease additional modelling effort. The parametric data obtained in the initial design phase by means of a detailed model are used to train an Artificial Neural Network model. The results obtained by this model are the compared with the ones obtained with a Resistance-Capacitance model. The approach is automated and tested for robustness using Monte Carlo simulation technique. This technique is used to identify, already in the design phase, probabilistic performance boundaries. The case study chosen is the eLUX Lab building at the Smart Campus of University of Brescia, in which highly variable occupancy patterns are present.