Probabilistic behavioral modeling in building performance simulation: A Monte Carlo approach

• Published in: Energy and buildings 2017-08, Vol.148, p.128-141
• Main Authors: Cecconi, Fulvio Re, Manfren, Massimiliano, Tagliabue, Lavinia Chiara, Ciribini, Angelo Luigi Camillo, De Angelis, Enrico
• Format: Article
• Language: English
• Subjects: Behavioral learning; Behavioral modeling; Building performance simulation; Energy efficiency; Energy management; Probabilistic modeling; Uncertainty propagation
• ISSN: 0378-7788
• DOI: 10.1016/j.enbuild.2017.05.013

•Occupant's behavior causes a relevant gap between simulated and measured performance.•Probabilistic modeling of occupancy can improve simulation reliability.•It is possible to establish a continuity among design and operation phase practices.•A reduced-order model is trained on design phase parametric simulation data.•The model trained is used for probabilistic simulation using Monte Carlo method.•Probabilistic scenarios can be used for different practices across building life-cycle. The increased awareness on sustainability matters is contributing to the evolution of energy and environmental policies for the building sector at the EU level, oriented toward resource efficiency. There exist today several possible strategies to model building performance through the life cycle. The increase of available computational capacity and of data acquisition capability is opening new scenarios for practical applications, which can contribute to the reduction of the gap usually encountered between simulated and measured energy performance. This article aims to investigate an approach for probabilistic building performance simulation to be used across life cycle phases, employing reduced-order models for performance monitoring and energy management. The workflow proposed aims to establish a continuity among design and operation phases. Design phase simulation is generally subject to relevant temporal and economic constraints and a successful workflow should incorporate elements from current design practices but should also add new features, which have to be reasonably automated to reduce additional effort. Therefore, the workflow proposed is automated and tested for robustness using Monte Carlo technique. In the design phase, the approach can be used for identifying probabilistic performance bounds suitable for risk analysis in energy efficiency investments, employing cost-optimal or life cycle cost accounting methodologies. In the operation phase, it can be used for performance monitoring and energy management based on daily energy consumption analysis, similarly to other multivariate regression-based methods at the state of the art, addressing the problem of maintaining energy consumption and related costs constantly under control.