[BLDG-SIM] Calibrated Models

[Glenn Haynes]

Mark,

At the outset, I think the term "calibration" should be limited to a simple 
matching (within pre-approved tolerances) of the target data one may have 
to work with.  If the target data are monthly utility bills alone, then 
calibration is fairly simple, but will necessarily leave unanswered 
questions regarding the individual variables that were "tweaked" to achieve 
calibration.  I would leave control of those input variables within the 
context of modeling detail or quality rather than confuse the issue of 
calibration with them.

A more stringent calibration project may require matching of both monthly 
kWh and monthly peak demand kW (which may need to be converted to 
hourly).  In this case the calibration process will necessarily require 
more detail in the model just to achieve the required tolerances around the 
12 additional target points.  The more target data you have, the more 
demanding the calibration procedure will be, and the more exact the model 
will necessarily have to be.  The term "calibration" could still be limited 
to the context of matching any number of target data points within an 
acceptable set of tolerances.  Issues of modeling accuracy and detail 
maintain their inherent importance both with and without calibration.

Glenn Haynes,
RLW Analytics, Inc.

At 11:30 AM 3/12/2003 -0700, you wrote:
>In Reference to the recent `Weather Normalization Question' discussion line
>which evolved into a discussion of calibration, and to many years of
>discussions of calibrated modeling, I think perhaps we are lacking a
>definition(s) of what a `calibrated' model is. There are many out there but
>there are no standard definitions that I've ever seen, beyond `matching
>utility bills within x %' that appears in some DSM program guidelines.
>I believe this must be done in the context of `What are we going to use the
>model for?"  I've seen the term used in reference to everything from
>aggregated load research work to the building on the corner that is
>considering a HVAC upgrade, from the Texas A&M shoot-out predictive
>statistical routines to detailed end-use monitoring and M&V approaches.
>Often it is used to as a surrogate for `expensive' or for `high quality'.
>Questions that should be answered in a definition:
>Calibrated to what standard measure? And why that particular standard?
>The entire model or just pieces?
>How closely does it have to match the defined standard? Possibly including
>various statistical measures.
>
>
>Mark E. Case, President
>etc Group, Inc.
>3481 South 2300 East
>Salt Lake City, UT 84109
>801-278-1927, 801-278-1942(F)
>
>
>
>
>======================================================
>You received this e-mail because you are subscribed
>to the BLDG-SIM at GARD.COM mailing list.  To unsubscribe
>from this mailing list send a blank message to
>BLDG-SIM-UNSUBSCRIBE at GARD.COM


======================================================
You received this e-mail because you are subscribed 
to the BLDG-SIM at GARD.COM mailing list.  To unsubscribe 
from this mailing list send a blank message to 
BLDG-SIM-UNSUBSCRIBE at GARD.COM