Meta-model based optimization of building thermal performance incorporating local comfort analysis

Good thermal control of the built environment is necessary to ensure comfort for occupants without excessive energy use. This is a particular concern in hospital environments where comfort needs may be more stringent and system design and control options are constrained by the need to consider infection control and clinical needs. In this study a building thermal optimisation methodology with the two contrasting objectives of optimizing thermal comfort and minimizing energy use is developed and applied to a generic "small room" model, representative of a typical single-bed hospital room. This model incorporates both a mechanical HVAC system and radiator heating, with selected design and operation parameters of both systems taken as design variables. The simulation is undertaken at a high level of detail, utilizing conflated dynamic thermal modelling (DTM) and computational fluid dynamics (CFD) to evaluate spatial variation in thermal conditions throughout the room. A moving least squares regression (MLSR) meta-model is applied to the problem and shown to model the problem to a very high degree of accuracy with a smaller sample set than required for complex methods such as neural networks and support vector regression. A parametric study shows how the room optimum changes when considering spatial variation of thermal comfort within the room, and variation in the time period of each optimization. As well as their own conclusions these analyses demonstrate the value of the "one sample many optimizations" approach, as they are all performed using only very few sample sets.