It seems to me that the whole building energy simulation community has become fixated on the topic of "How to calibrate an energy model." Calibration ensures long term growth of the energy modeling industry because a calibrated model commands respect across multiple disciplines. Calibration is nearly always useful, albeit unnecessary, in many circumstances. First things first:

What does it mean to calibrate an energy model?

In laymen's terms, calibrating an energy simulation means "Gathering the actual energy used by the REAL building, after the building has been fully operational for some time - usually about a year - and then adjusting energy model inputs so that the simulation output more closely approximates the energy consumed by the real building"

When can a simulator calibrate a model?

The simulator requires a completed initial model AND the actual building energy consumption after it has been in full operation for some time (ideally, a full year).  At that point, one can compare the model results to actual data determined by the unique condition called "reality". The simulator may subsequently adjust the model's inputs so that the modeled outputs align best with the building's energy consumption in real-world operation.

Do all models require calibration?

Not all models require calibration. Many simulators do not intend to predict the actual results. Instead, many simulators create models to yield a comparative product. For example, if a simulator calculates electric resistance heating compared to a gas boiler, the simulator expresses the energy cost in terms of a percent difference. In most cases, a calibrated model produces a similar percent difference, especially if the model has reasonable assumptions.

When is a model considered calibrated?

There is no universal definition for calibration. It is done when the person in charge says it is done. This often referred to as the rating authority. Thus, the rating authority defines the definition of a calibrated model in mathematical terms. ASHRAE Guidelines 14 speaks to many of the official mathematical measurements. In the most basic scenario, authorities such as utility companies use a statement such as, "The annual energy use of the model must match within 5% of the annual building energy use with no individual month varying beyond 10%". 

Example: A building operates for a year, and the records show the building consumed exactly 10,000 kWh each month for the entire year (note that this is unrealistic!). The actual building, therefore, consumes 120,000 kWh for the year.  By the definition mentioned above, the calibrated model must output an annual energy consumption within 5% of 12*10,000 = 120,000 kWh (between 114,000 and 126,000 kWh). Additionally, each month in the calibrated model must consume +/-10% of 10,000 kwh or between 9,000 and 11,000 kWh. 

Shameless self-promotion here: During a webinar in early 2013, an attendee asked, "What level of calibration can we consider possible?" 

I thought for a moment and answered on the spot, stating, "5% on the year with all months within 10%." Interestingly, several utility companies were watching that presentation. A year later, I completed some models and submitted them to one of these utility companies. The model was accurate, with one anomaly - there was a month where the boiler didn't operate the building required backup electric heat for a single month. The reviewer commented, "We require the model's results to match within 5% on the year all months match within 10% of the utility data". I wish I would have said "5% on the year, where 10 of the 12 months are within 10% of the corresponding utility data".  The reasoning is that many buildings have a month of abnormal operations in any given year. Still, it serves little value to adjust a model for one-time anomalies (such as in my case, where the boiler controller had trouble and they used backup electric heat for a single month). Nonetheless,  I was glad that I didn't suggest more precise numbers!

Hopefully, that provides ground-level insight as to what building energy model calibration is. Understand that is the most basic level of calibration, and some calibration methods require the model to match at various enduses, where the results are broken down into the total energy by lights, plug loads, cooling, heating, and so on.

We will discuss those requirements in upcoming posts.

The book to read in the meantime is ASHRAE G14, which contains numerous definitions, though it is challenging to read for a novice modeler. In fact, many experienced modelers have difficulty understanding ASHRAE 14, so don't let it intimidate you. We will attempt to break it down into simpler terms in upcoming blog posts.

Stay tuned for upcoming posts on calibration