Uncertainty and Sensitivity Analysis
As building energy modeling becomes more sophisticated, the amount of user input and the number of parameters used to define the models are growing. There are numerous sources of uncertainty in these parameters especially when the modeling process is being performed prior to final construction and commissioning. Past efforts to perform sensitivity and uncertainty analysis have focused on tens of parameters. In this work, we increase the size of analysis by two orders of magnitude (by studying the influence of a few thousand parameters). In addition to this, once uncertainty is quantified in the output of a building model with uncertain input parameters, sensitivity analysis is typically performed to identify which of these input parameters are most influential. We extend this analysis even further to decompose the influence pathway through the dynamics, which identifies which internal or intermediate variables induce the most influence on the output.
B. Eisenhower, Z. O'Neill, V. Fobonorov, and I. Mezic Uncertainty and Sensitivity Decomposition of Building Energy Models, Journal of Building Performance Simulation May 2011
|Schematic of the uncertainty process. Model parameters (~1000) with various distributions are the source of uncertainty in the model. Through multiple simulations (~5000) uncertainty of the output parameters is quantified.|
Building Model Zoning
When creating an energy model, a building is divided into a collection of zones; zones are regions in the building model where properties are lumped and assumed uniform over an area. Unfortunately, there is little consensus as to how a model should be zoned. Few industrial standards exist to control the quality of models that are made professionally. This can lead to the production of oversimplified models that miscalculate the characteristics of a building while still meeting all necessary industrial standards.
To understand how zoning effects a model’s predictions, the Student Resources Building at UCSB was investigated. A library of models was created with various amounts of refinement in how the building was zoned. The work shows that a simplified model will always under-predict a building’s operating characteristics despite meeting all modeling standards.
|Error in predicted energy usage between various models of the Student Resources Building. Models are shown in series of increasing zoning refinement illustrating that oversimplified models under-predict consumption by as much as 20%.|
|Floor plan view of two EnergyPlus models of the Student Resources Building’s second floor|