Simulation Speedup Techniques For Computationally Demanding Tasks

Abstract

The computational cost for the repeated evaluation of zonal-type building simulation models can be prohibitive especially in contexts, such as Building Optimization and Control Design, where repeated evaluation of the models -for different initial and boundary conditions -is required. In the present paper, two techniques to reduce simulation time are investigated: (i) geometry simplification for periodic geometries; and, (ii) the use of co-simulation to split a building into simpler sub-buildings, that can be evaluated in parallel, exploiting the resources of multi-core computational architectures. These simulation speed-up approaches are evaluated, with respect to accuracy and computational effort, against the validated full-scale models of two real buildings.