Multi-criteria robustness assessment of a sequential whole-building design optimization

This paper assesses the reliability of a sequential approach, derived from a conventional design process, in the optimization of building geometry, fabric, HVAC system and its controls for building performance. In this study, the reliability is defined as the ability of the approach to find all Pareto-optimal design solutions under scenario-based uncertainties on occupant behaviour and weather conditions. The optimization criteria are the minimization of energy demand and warm discomfort hours. A full-factorial (exhaustive) search method has been adopted to explore the design space, therefore ensuring that all Pareto-optimal solutions are identified and used a benchmark for the sequential search. Eight combinations of tests have been performed based on different initial values and groups of the design parameters used at the start of the sequential search. It concluded that a sequential search, across scenarios, identified, on average, 70% of the Pareto-optimal solutions obtained from a full-factorial search, with a computational saving of function evaluations of 99.99%. Additionally, the ability to obtain solutions displaying medium and high robustness is not affected by the lower frequency of optima found across scenarios.