Parallel object-oriented algorithms for building performance simulation. Application to an existing dwelling

Abstract

In the present work an existing dwelling, situated in the Netherlands, has been modeled by means of a parallel object-oriented simulation tool, called NEST-Buildings. The model is based on a pre-defined collection of elements (e.g., walls, rooms, openings, outdoors, occupants, ventilation tubes and boxes, solar radiation distributors, HVAC equipment, etc.) that are connected to each other conforming a dynamic thermal system. New configurations can be easily handled by adding or removing elements. Moreover, the building elements can be modeled at distinct levels of accuracy ranging from lumped volumes mixed with one-dimensional to detailed CFD&HT models. This approach makes possible the assessment of general-type buildings (residential, services, old, modern, etc.) using the appropriate modeling level at each component. The work is one more step in the improvement of this computer simulation tool. So far, the full simulation of the overall building model is based on block-Jacobi and Gauss-Seidel algorithms. With the current implementation, the computational time for performing practical simulations may become an important impediment as the size of the building increases. For instance, the computational expenses of a family house are far larger than those in a single apartment since the number of rooms, walls, events and so on is bigger. The last advances in this research line, including the use of optimum time stepping, proper mesh sizes, convergence criteria, loop control strategies and the use of other non-linear solvers based on the Newton method, are presented and discussed through comparative analysis of the simulated dwelling. The advances in this direction will help first to better understand the behavior of the already available algorithms and later to speed up the simulations. The second is important in the attainment of optimal designs of dwellings or other type of buildings.