ASHRAE Journal - September 2021 - 10

LETTERS
consumption of the completed
school to reduce electricity costs
and provide a building with a net
positive production of energy yearly.
Geothermal heat pumps with
decoupled DOAS for ventilation is
the most efficient system that met
the project's design criteria.
This overall reduction in energy
and removal of combustion from the
site helps reduce the building's lifetime
operational carbon footprint by
reducing the total amount of energy
used from the power grid. The best
way to reduce operational carbon
emission is to not consume power
to begin with, thus the use of one
of the most efficient HVAC systems
available.
Lee Harrelson, P.E., Associate Member ASHRAE,
Richmond, Va.; Ray Beaufait, Member ASHRAE;
Brice Watson, P.E., Associate Member ASHRAE;
Brian Turner, P.E., Member ASHRAE, Louisville, Ky.
Modeled Energy
Vs. Actual
Performance
I strongly disagree with the conclusion
of " Modeled Energy vs. Actual
Performance: Does the Lab Building
Follow the Model? " by Kelley
Cramm, P.E., Member ASHRAE, in
the July issue of ASHRAE Journal.
Cramm states, " Many factors make
predicting actual laboratory energy
performance difficult, if not impossible. "
This reads like a denouncement
of energy models as viable
predictive tools. It stems from a false
dilemma. We act as if accurately predicting
energy is a yes/no question.
It is not; it is a continuum.
There are better questions than yes
or no. How accurately can a model
predict energy? Which energy parts
10
ASHRAE JOURNAL ashrae.o rg
can a model predict with higher or
lower accuracy? How accurately can
a model predict cooling energy, and
under what conditions? How can
we improve predictive accuracy?
How accurately can a model predict
cooling energy, given outdoor
temperature?
Weather variation itself should not
be insurmountable. Figure 1 and
Figure 2 are reductio ad absurdum. To
show a 20% to 40% spread in cooling
and heating energy, Cramm
compares a year full of record low
temperatures to a year full of record
high temperatures. Both years are
absurd. The point is valid, albeit
to a much lesser degree. A model's
prediction of cooling energy won't
be dead-on accurate; we should
expect a range. I think 20% to 40% is
too broad; a competent professional
should be able to do better than that.
(And, If the variance is based on
weather, it can be stated as a sensitivity,
not an error.)
I challenge Cramm to find fault
with the following logic. At a given
outdoor temperature and time of
day, a model can predict cooling
energy to about 15%. Once built, the
owner can validate cooling energy,
hour by hour, comparing conditions
to the model. If the measured cooling
energy doesn't fall in the control
range predicted by the model for at
least 80% of observed hours, then
either the model or the installation
is a failure.
Used correctly, a model should be
categorically invaluable during commissioning.
Think about it: if the
model says chiller energy should be
250 kW at 10 a.m. with it 80 degrees
outside, and, during commissioning,
we measure 750 kW at those conditions,
something has clearly gone
S E PTEM B E R 2021
wrong! Heck, if it's 300 kW, something
has gone wrong, and everyone
should be working on fixing it.
Likewise, occupant behavior also
shouldn't kill viable accuracy. Of
course, density, activity and equipment
diversities are variables, as
Cramm rightly points out. But they
don't vary wildly or unpredictably.
Figure 4 shows variability of about
20%. Moreover, in Figure 3, the plug
loads are 23% of the overall energy
signature. So, even if they vary by
20%, the overall impact is only about
5%.
As an energy professional, my
advice is to state the accuracy of
models. For example, instead of telling
a client cooling energy will be
231,254.7 kWh/yr, tell them it will
be between 210,000 kWh/yr and
265,000 kWh/yr. Tell them what
factors affect the range, and by
how much. No investment should
be made without knowing (and,
frankly, beating) the margin of error
on the model.
Concluding energy models are poor
predictive tools is self-defeating. As
a buyer, if you tell me a model " ...
cannot be used to accurately predict
future use or cost " that literally makes
it sans value. Also, if a model is a poor
predictive tool, logically, it cannot be
" an effective design tool when used
for comparative analysis. " What's the
use of comparing two possibly not
credible models to each other? Only a
comparison between credible models
can be valuable.
Rather than writing a column
complaining how energy models are
poor predictive tools, I wish Cramm
had written one telling us how to
make and use them better.
Travis R. English, P.E., Member ASHRAE,
Fountain Valley, Calif.
http://www.ashrae.org/
ASHRAE Journal - September 2021

Table of Contents for the Digital Edition of ASHRAE Journal - September 2021
