ASHRAE Journal - September 2014 - 39

TECHNICAL FEATURE

Model Calibration and Uncertainty
We then compared initial results from the 24
energy models to the actual monitored energy usage.
Discrepancies between the two were assumed to be the
result of uncertainty in model inputs such as envelope
properties, lighting power, plug load equipment power,

100

Percent 2011 Source Energy

predominately set to perimeter-core with specific zoning only occurring for areas with loads significantly different than the building as a whole (i.e. warehouse adjacent to an office). Windows were modeled as approximated window-to-wall ratios taken from site photos.
Because of the age of many of the buildings, we could not
determine precise assembly properties for roofs, walls,
and windows. For these cases, we assumed the roof to
initially have R-10 insulation. We assumed the walls to be
12 in. (300 mm) medium weight concrete with minimal
insulation, and the windows to be single-paned with clear
glazing. For the handful of newer buildings in our study,
we assumed code required minimum values of insulation and window properties from ASHRAE Standard
90.1-2004, as required under federal regulations near the
time of construction. Occupancy density was provided
by SSC facility staff. The buildings were predominately
considered occupied between 6:00 am and 6:00 pm as
corroborated by facility staff and the electric interval data.
We preliminarily set lighting to code required values
from Standard 90.1-2004 for the building's predominant
use type (i.e., 1.0 W/ft2 [11 W/m2] for buildings that were
mostly office). No daylighting controls were reported for
the modeled buildings. We initially set miscellaneous
loads to default values outlined in COMNET's Commercial
Buildings Energy Modeling Guidelines and Procedures6
for a given building's predominant use type. Infiltration
flow rates were approximated according to guidelines
published by Pacific Northwest National Laboratory.7
We modeled HVAC system types according to input
from SSC facility staff. The majority of primary HVAC
systems for the modeled buildings were variable air
volume with hot water reheat. Cooling was provided
by water-cooled chillers, while heating was provided
by atmospheric boilers. The efficiencies for the HVAC
equipment were preliminarily set to code-required
minimum values as outlined in Standard 90.1-2004.
No demand control ventilation controls were found in
the modeled buildings, and only one instance of energy
recovery ventilation was found.

80
60
40
20
0

11 21 31 41 51 61 71 81 91 101 111 121 131 141
Building Count

FIGURE 2: SSC cumulative source energy consumption by number of buildings
largest energy users first.

infiltration flow rates, outdoor air flow rates and HVAC
equipment efficiencies.
We used the Nelder-Mead simplex optimization algorithm8 to calibrate each of the 24 energy models to
actual monthly energy use data. The algorithm searches
for the energy model input parameter set that minimizes an objective function comparing modeled energy
use to actual energy use.
Our choice for objective function follows ASHRAE
Standard 1051 and Guideline 14 for energy model
calibration and evaluation. We used Goodness of Fit
(GOF) as our objective function, which is based on the
coefficient of variation of the root mean squared error
between modeled and measured monthly energy consumption, and weighted by the annual cost of each fuel
type.
The convergence criteria for the objective function was
set to 15% for each model (i.e., GOF <15% for each building model). We inspected all calibrated model parameters to ensure values fell within acceptable ranges based
on our understanding of the building and our engineering experience. Quality checks were also performed on
model results. Cooling load, economizer operation, and
reheat controls were each rigorously explored to determine proper performance.
Once the calibration algorithm had been applied to
each building energy model, we had a set of models
that represented SSC energy use under current climate
conditions. Modeled total annual energy use was within
5.5 % and 2.1 % of measured 2011 data for electricity and
SEPTEM BER 2014

ashrae.org

ASHRAE JOURNAL

ASHRAE Journal - September 2014

Table of Contents for the Digital Edition of ASHRAE Journal - September 2014