SBTP 2012 - Oct, 2012

Title of the Conference/Workshop: Simulace budov a techniky prostředí 2012

Location: Brno, Czech Republic

Dates: 23rd-26th October, 2012


Campus21 publications



This paper examines the potential of sensitivity analysis-assisted optimization-based simulation model calibration. Toward this end, a university campus office area was selected and equipped with a monitoring infrastructure. Various streams of data were collected, including indoor climate, outdoor weather conditions, energy delivery via the building's heating system, and energy use for lighting and equipment. One of the primary objectives of this comprehensive monitoring campaign was to use monitored data to both populate the initial thermal simulation model of the office  area and to maintain its fidelity through a calibration process. The present  contribution also addresses a specific problem faced by an optimization-based  simulation calibration approach: In many realistic circumstances, a large number of model input variables could be subjected to the optimization process. This large number of candidate input variables can be reduced to a certain extent via  heuristically based considerations pertaining, for example, to the knowledge domain captured in building physics. We argue, however, that this process could be further rationalized, if we make use of sensitivity analysis to identify a subset of the input variables most likely to influence the simulation results. Distinguishing this subset from the entire set of input variables will reduce the computational cost of the subsequent calibration process.