Automated calibration of non-residential urban building energy modeling

  • Automatisierte Kalibrierung von Simulationsmodellen für Nichtwohngebäude im städtischen Maßstab

Remmen, Peter; Müller, Dirk (Thesis advisor); van Treeck, Christoph Alban (Thesis advisor)

1. Auflage. - Aachen : E.ON Energy Research Center, RWTH Aachen University (2022)
Book, Dissertation / PhD Thesis

In: E.ON Energy Research Center ; EBC, Energy efficient buildings and indoor climate 105
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme

Dissertation, RWTH Aachen University, 2021

Abstract

Building simulation supports the conceptual design, planning and operation of innovative and economical energy supply options for buildings and cities. Typical applications of Urban Building Energy Modeling are quantification of the impact of retrofits or the optimization of heat transfer of district heating networks. The necessary input data for design-driven building simulation are often not available for existing buildings. Archetypes are used to fill the missing information and thus enable a dynamic simulation of a large number of existing buildings. The simulation results of archetypes show differences to hourly measured data of individual buildings due to statistical and discrete assumptions. To improve the description and prediction of individual buildings, the calibration of simulation models is key. This thesis describes a method for automated calibration of individual urban building simulation models using hourly measurement data. The proposed framework automates all necessary steps from model generation, selection of sensitive parameters, calibration, and evaluation using statistical indices. The model generation uses the standardized information model CityGML and its extension EnergyADE and utilizes all available building data. In addition, further building parameters are identified with the help of the analysis of existing measurement data. The selection of sensitive parameters is done for each building and thus takes into account specific building characteristics. Bayesian calibration is used as the calibration method, which is particularly characterized by the determination of probability distributions. Hourly measurements are considered in calibration and evaluation of the simulation results with a combination of several statistical indices. The application of the framework to emulated and real buildings helps to better understand the strengths and weaknesses of automated calibrations for building models. It was shown that the reliability of automated calibration cannot be evaluated based on simulation results alone, but the calibrated parameter values have to be taken into account. In this context, compensation effects were described in detail. It could be shown that the more information is used for the parameterization of the initial building model and the more measurement data is available in hourly resolution, the better is the calibration of the models in comparison to the simulation results and the lower are compensation effects.

Identifier

Downloads