Uncertainty is a situation which involves imperfect and/or unknown information. As such, it is intrinsically integrated into any simulation process, including building performance simulation. However assessing, mitigating, and managing uncertainty in building performance simulation demands developing special tools that are able to integrate stochastic aspect of uncertainty into simulation.
The idea of a special issue devoted to the important topic of uncertainty emerged during the Building Simulation Conference in 2013 in France, where about 40 papers with the keyword uncertainty were presented. Assessing and managing uncertainty appeared then as an emerging and very important topic.
It was a correct assumption: in our Journal of Building Performance Simulation, papers integrating uncertainty aspects have become more and more numerous. Between 2008 and 2012, each volume of the journal contained from 7 to 11 articles related to uncertainty, whereas in the past year more than half the published articles address uncertainty in some respect.
Why this topic is important?
Firstly, even deterministic parameters are somehow uncertain. For example, when conducting simulations of laboratory experiments, measurement uncertainty can impact results. And the precise physical properties and geometry are to some extent unknown in the case of simulations supporting building design or operations.
Secondly, many boundary conditions and operational parameters are stochastic by nature: occupants’ presence and released heat, outdoor temperature one year from now, wind speed and direction, are only a few examples.
And last, but not least, speculations about future building performance: building use, building component ageing, climate change, energy prices, etc. cannot be assessed precisely.
How it can be treated?
Assessing, mitigating, and managing uncertainty in building performance simulation may be performed using a very large panel of methods and tools.
One of the primary aspects involves quantifying the uncertainty on input parameters, static and dynamic, as well as on models used for the simulation, and to evaluate its impact on building performance. Frequently, sensitivity analyses (SAs) are also used in order to assess the most important parameters influencing results.
Then, a variety of methods can be used in order to propagate the uncertainty within building performance simulation tools. Undeniably, the general issue is to limit the uncertainty on the results.
What you will find in the present journal issue?
To identify variables and their uncertainty impacting the uncertainty of outputs, SA is often a first step. The study by Monari and Strachan proposes a rigorous approach to SA based on a global SA method with three stages: factor screening, factor prioritizing and fixing, and factor mapping. The method is applied to a detailed empirical validation data set, with the focus of the study on the airflow network, a simulation programme sub-model which is subject to large uncertainties in its inputs.
The paper by Goffart et al. presents a statistical approach for uncertainty and SAs applied to hygrothermal properties of a brick material and their impact on air conditioning demand in hot and humid climates. A variance-based global SA is used, with an interesting practical implementation of the method providing reliable information with low simulation cost. Moreover, the authors also investigate the uncertainty related to the choice of the model.
A different approach is proposed by Kneifel and O’Rear whose study addresses a very important issue of the robustness of the design against varying weather conditions. Weather variability is here assessed in a very realistic way: using recorded weather files from 34 years in the same location. Given the unpredictability of weather conditions, such an approach enables realistic assessment of different energy savings methods. The proposed methodology contributes to limit the uncertainty of the predicted performance.
A distinctive study about the optimization in the design phase is proposed by Das et al. As it is clear that there are several choices to make when designing an uncertain optimization scheme, the authors explore the impacts of several of these choices to create an uncertain optimization scheme, applied here to economical optimization of Swedish attic retrofit. As the stochastic approach requires numerous simulations, the authors also explore another tool widely used in uncertainty analyses – metamodelling: constructed here using a neural network.
It is hoped that readers of this special issue will enjoy the mix of methods and results presented here and perhaps find the inspiration to help evolve this field further. The topic is far from being completed.