CFD simulation of indoor climate in low energy buildings

Abstract: In this thesis computational fluid dynamics (CFD) was used for simulation of the indoor climate of low-energy buildings in cold climate. The heat consumption in newly built houses was reduced drastically. Along with the different classification systems for low-energy buildings the demand for the indoor climate has increased, which causes a need to investigate buildings even before they are built. Than CFD is of importance in studies of different heating systems and how new construction solutions can affect the indoor environment. The work focus was on investigating the computational setup, such as grid size and boundary conditions in order to solve the indoor climate problems in an accurate way and compare different heating systems. A limited number of grid elements and knowledge of boundary settings is therefore essential in order to obtain reasonable calculation time.The models show that radiation between building surfaces has a large impact on the temperature field inside the building, with the largest differences at the floor level. An accurate grid edge size of around 0.1 m was enough to predict the climate. Different turbulence models were compared with only small differences in the indoor air velocities and temperatures. To explore the viability of this approach, the indoor climate in a building was studied considering three different heating systems: an underfloor heating system, air heating through the ventilation system and an air heat pump installation. The underfloor heating system provided the most uniform operative temperature distribution and was the only heating system that fully satisfies the recommendations to achieve tolerable indoor climate set by the Swedish authorities. On the contrary, air heating and the air heat pump created a relatively uneven distribution of air velocities and temperatures, and none of them fulfils the specified recommendations. From an economic point of view, the air heat pump system is cheaper to be installed but produces a less pleasant indoor environment then distributed heating systems. The most widely used turbulence model for indoor CFD simulations, the k-ε model, has exhibited problems with treating natural convective heat transfer, while other turbulence models have shown to be too computationally demanding. One paper therefore studies how to deal with natural convective heat transfer for a radiator in order to simplify the simulations, reduce the numbers of cells and the simulation time. By adding user-defined wall functions, to the k-ε model the number of cells can be reduced considerably compared with the k-ω SST turbulence model. The user-defined wall function proposed can also be used with a correction factor for different radiator types without the need to resolve the radiator surface in detail. Compared to manufacturer data the error was less than 0.2% for the investigated radiator height and temperature.