Choosing the appropriate sensitivity analysis method for building energy model-based investigations

Abstract Current literature holds various examples of investigations that make use of a building energy model (BEM) combined with a sensitivity analysis (SA) technique to identify and rank the BEM input parameters that the model output is most sensitive to. However, a sound argumentation that vouches for the reliability, validity and necessary complexity of the chosen SA method for the specific purpose of the BEM-based analysis, is rare. This paper reports on an investigation of how two different levels of a-priori information about input parameters, applied to three different SA methods (Local, Morris and Sobol), influenced the identification and ranking of the input parameters that the annual energy need output of a quasi-steady-state BEM using monthly time steps, and a simple dynamic BEM using hourly time steps, is most sensitive to. It was found that the three SA methods, to a great extent, were able to identify the same cluster of most sensitive input parameters, independent of the level of a-priori input parameter information and BEM. However, the ranking of most sensitive input parameters varied with the applied SA method, BEM, and level of a-priori input parameter information. From a practical point of view, the choice of appropriate SA method is concluded to depend on the purpose of the SA analysis.