Computational time reduction using detailed building models with Typical Short Sequences

Comprehensive studies of detailed dynamic building models, which take into consideration both the envelope and the connected systems, yield more precise results compared with simplified ones, but at considerable computational expense. Aside from classical approaches that work on the model itself to accelerate the simulation process such as model reduction or metamodels, this paper focuses on the concept of applying reduced simulation sequences directly to detailed models to calculate annual results. The objective is to quickly and precisely reproduce the integrated annual profiles of predefined criteria of a computationally expensive reference model. After presenting and analyzing methods used in the literature to reduce weather data, we categorize the methods based on the type of data used and the nature of the process for selecting the typical days. Analysis of these methods led to the development of a new iterative approach with an embedded grouping algorithm. The method creates and iteratively enhances a short simulation sequence of typical days based on data reflecting the integrated annual profiles calculated using the detailed model. The reduced sequence led to much faster simulations while achieving profiles highly correlated with the reference integrated annual profiles. In addition, the last annual value, i.e., final annual sum, of each criterion extrapolated from a typical 12-day simulation differs little from the reference values (errors less than 1%). Moreover, the method was compared to two other clustering methods based on different types of selection data and an iterative method used in the literature. The results show that the classical method of day selection based only on weather data, typically used to generate Short Reference Years (SRYs), is in fact unable to accurately reproduce the annual reference profiles. Finally, the approach was also efficient when generalized, demonstrating its applicability to future optimization studies.