Refinement of dynamic non-residential building archetypes using measurement data and Bayesian calibration

Abstract

Dynamic simulation and optimization of energy systems at urban level becomes increasingly important as an additional decision criteria for planning and operation. Essential in this context is the dynamic heating and cooling demand of buildings. Often building archetypes are used to depict the individual characteristics of buildings. While in average the demand is accurately reproduced by archetype buildings, individual buildings deviate from the statistical average. This paper presents a methodology to refine building archetypes using real measurement data and Bayesian Calibration. The calibration uses statistical indices instead of whole time series or yearly cumulated energy demands. We apply the methodology to four real buildings, the results demonstrate the potential of the methodology by reducing the RMSE between measured and simulated heating demand up to 57 %.