Improving the way we model people's actions in energy models of buildings

What is it about?

We suggest a model for describing people's actions (opening and closing windows) for use in models of buildings that are used to predict the building's energy use. We developed this behavior model statistically, i.e. based on gathered data from real buildings. It is stochastic, meaning probability-based. Under certain conditions, we assign a probability that a building occupant will open or close a window. This probability is calculated and used in each time interval in the building energy model. This is a more sophisticated approach than the what is currently used in building energy simulation by most people. We compared the energy predictions from this simulation to measured data (hourly steam usage). We then ran building energy models that used other forms of occupant behavior models, including the conventional method and other empirical (experiment based) models suggested by other researchers. Our behavior model worked better than the others in producing accurate energy use predictions. However, we were also able to show that models from other researchers worked reasonably well. This shows that these empirically-based behavior models can be generalizable from building to building under certain conditions. In addition to this, we also described a method for predicting a window's state (either open or closed) based on other data, including indoor/outdoor temperatures and CO2 levels and their changes over time. This is important, because window state sensors are very rare in buildings, yet temperatures and CO2 levels are commonly measured in buildings with Building Automation Systems (i.e. many buildings). Therefore, the pool of buildings available for studying window operation could be exponentially larger than most researchers expect.

Featured Image

Why is it important?

See above.