Calibration in Urban Building Energy Modeling

This thesis work tries to study the annual calibration process for building performance simulations. The bottom-up modelling is a tool to predict energy consumption in Urban Building Energy Modeling (UBEM) which is a common approach to find the most appropriate retrofitting strategies or to monitor energy use for a group of buildings (an archetype), a neighbourhood or a city. To do this, one needs to have local climate data for a site, building geometries, construction assemblies, usage schedules and HVAC system information for each building in that site. But many parameters can be uncertain or completely unknown and not all of them are equally influential. Therefore, the aim is to find representative values for affecting but unknown parameters. To do this, first a modeller needs to identify which parameters affect the thermal behaviour (energy usage) of buildings. Later instead of considering one value for each parameter, the modeller should define distributions for those parameters based on experience. The idea is to perform enough number of random simulations from parameter ranges to find the most representative areas of each influential parameter. In this study, two methods are identified to be useful for sensitivity analysis, Morris and RDB-FAST. Later a brute-force search calibration process is developed to find a good representative average set of parameters for a group of buildings, based on their annual energy consumption. In the end, to reduce the computational complexity of the calibration process an intuitive approach based on normal transformation of calibrated parameters is used to recalibrate the parameter ranges. The study shows that by starting from calibration with a low number of simulations, the sampling from the result of a recalibration can reduce the percentage of deviation between measurement and simulation and increase the number of matches for parameter combinations in comparison with the sampling from only one calibration process