Paper Conference

Proceedings of BauSim Conference 2008: 2nd Conference of IBPSA-Germany and Austria


Optimierung von linearen Regressionsmodellen für Energiesignaturen auf Basis von Gebäudesimulationen

Dirk Jacob, Sebastian Dietz, Christian Neumann, Sebastian Herkel

Abstract: Often building operation is only adjusted during initial commissioning, possibly leading to energyinefficient operation. Simple Black-Box models can be used to indicate energy savings achieved by changes to the operation of the building. Replacing real buildings by simulations is advantageous for systematic research. For this study a typical office building was modelled with EnergyPlus in 152 variations. Multi-Variable Linear Regression models with a change point were applied to the whole buildings heating, cooling and electricity consumption. By adapting the model a mean coefficient of determination R² above 95 % for daily, weekly and monthly values was achieved for all 152 variations. The quality dependency of the regression models on the time period used for training is also investigated.
Pages: 50 - 51