New optimisation methodology to uncover robust low energy designs that accounts for occupant behaviour or other unknowns

The use of software to aid in the design of buildings or to show compliance is now commonplace. This has led several authors to investigate the potential for using such software to automatically optimise a design, or to generate a variety of near-optimal designs. One area where this approach has been found useful is in minimising annual energy demand. It is known that any estimate of demand will depend not only on the architecture and constructions used, but on the preferences and behaviours of the occupants. This suggests which design is truly optimal will also depend on occupant behaviour. In this paper optimisation is carried out for an array of different occupant behaviours based on real records. It is found that the resultant designs are more robust in terms of predicted heating energy use and overheating than when only a single behaviour is considered. It is recommended that in future all such optimisations are made using a realistic spectrum of behaviours, and that the approach is expanded to include other elements of design that might show variance during construction, for example, U-values and air tightness. This, it is hoped, will reduce some of the risks of designing and asking people to occupy very low energy buildings. Importantly, it is found that the near-optimal building designs found under variable occupancy present different characteristics than when only a single statement of occupancy is used. Being cognisant of this reduces the potential for inappropriate designs to be created that rely on a serendipitous arrangement of design and occupancy parameters that might not be met on site or by the occupants.