Energy Performance Gap of an Office Building: The influence of modelling assumptions and level of detail

Abstract

Building simulation software can be used to predict the energy use of buildings in the design phase. However, due to the importance of time efficiency in building projects today, energy models are often simplified. Numerous cases have been reported where the energy models are not performing in accordance with the actual building. One of these cases is an office building in Trollhättan, Sweden, where the predicted energy use was 37% lower than the average of the yearly total energy use during operation. The difference between a building’s predicted energy performance and the actual energy performance is called energy performance gap. This study investigated whether a more detailed energy model could decrease the energy performance gap of the building. The level of detail was increased by implementing the building’s actual geometry and solar shading system, among other factors. Furthermore, the magnitude of various modelling assumptions was assessed by a sensitivity analysis of the model, where uncertain input data were varied one factor at a time. The original design model led to a performance gap of 37%. This however includes post-simulation addition of template values for energy use due to heat losses not accounted for in the simplified model, e.g. for the hot water circuits. The performance gap based on the result from the simulation alone was 53%. The performance gap derived from the detailed energy model was 47%, i.e. only slightly decreased compared to the original model result, and even increased compared to the original model with post-simulation addition of template energy use values. Consequently, the energy performance gap of the building was not decreased by implementing a more detailed model. Furthermore, the sensitivity analysis showed that incorrect modelling assumptions can be a reason for the performance gap. The chosen setpoints for heating and cooling had the largest impact on the energy use among the four parameters tested. At last, important input data for energy modelling was emphasised, and proposals for increasing the level of detail further was presented.