Guidelines for Energy Simulation of Commercial

Buildings -

Final

ilk
PaWER ADMINBSIWAPION
GUIDELINES FOR ENERGY SIMULATION

OF

COMMERCIAL BUILDINGS

Prepared by:

Michael Kaplan
and Phoebe Caner
Kaplan Engineering

March 1992

Prepared for:

Bonneville Power Administration

Commercial Technology Section - RMCB
P.O. Box 3621
Portland, Oregon 97208
(503)230-7507

Under Cooperative Agreement #DE-FC79-85BP26683

 4.2.5 Daylighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2.6 Thermal Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2.7 Unconditioned Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2.8 Interior Walls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.9 Above-Ceiling Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.10 Heat Loss to Ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2.1 1 Weather . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3 INTERNAL LOADS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.1 General Plug Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.2 Computer Rooms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.3 Lighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.4 Refrigeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.5 Cooking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

HVAC SYSTEM SIMULATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.1 SYSTEM SELECTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2 HOW AIR DISTRIBUTION FITS INTO THE OVERALL SIMULATION . . . . . . . 62
5.3 CAPACITY. LOAD FACTORS AND PART LOAD EWICIENCIES . . . . . . . . . . . 63
5.3.1 User inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3.2 Default Capacities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.3 Sizing Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.4 Part Load Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.4 CONTROLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4.1 Sequence of Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4.2 Role of the Analyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.4.3 The Building Operator's Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.4.4 Specific Control Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.5 SIMULATION OF MULTIPLE ZONE SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.5.1 General System Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.5.2 Multi-Zone and Dual Duct Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5.3 Variable Air Volume Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.5.4 Water-Loop Heat Pump Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.6 VENTILATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.7 FAN SCHEDULES AND SUPPLY CFM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.8 FANHEAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . j . . . . . . . . . . . . . 78

OTHER ENERGY CONSUMPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

6.1 EXTERIOR AND "HIDDEN" ENERGY USERS . . . . . . . . . . . . . . . . . . . . . . . . . . 79

THE BASELINE BUILDING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

7.1 GENERAL DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.2 BASELINE ISSUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

ERROR CHECKING AND DEBUGGING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

8.1 GENERAL ERROR CHECKING APPROACH . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.2 INPUT CHECK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.3 OUTPUT CHECK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.4 EUICHECKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

9. DOCUMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
9.1 PURPOSE O F DOCUMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
9.2 DOCUMENTATION REQUIREMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
REFERENCES ............................................................ 92

SUMMARY OF MODELING RECOMMENDATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

APPENDICES .................................................................. 97
 INTRODUCTION
This report distills the experience gained from intensive computer building
simulation work for the Energy Edge project. The purpose of this report is
twofold: to use that experience to guide conservation program managers in
their use of modeling, and to improve the accuracy of design-phase
computer models. Though the main emphasis of the report is o n new
commercial construction, it also addresses modeling as it pertains to retrofit
construction. To achieve these purposes, this report will:

discuss the value of modeling for energy conservation programs

discuss strengths and weaknesses of computer models
provide specific guidelines for model input
discuss input topics that are unusually large drivers of energy use
and model inaccuracy
provide guidelines for developing baseline models
discuss types of energy conservatidn measures (ECMs) and
building operation that are not suitable to modeling and present
possible alternatives to modeling for analysis
provide basic requirements for model documentation.

Energy Edge is a large-scale research and demonstration project developed

by the Bonneville Power Administration, This project was initiated to
determine whether commercial buildings can be designed and constructed to
use at least 30% less energy than if they were designed and built to meet the
current regional model energy code, the Model Conservation Standards
(MCS)developed by the Pacific Northwest Electric Power and Conservation
Planning Council. Secondary objectives of the project are to determine the
incremental energy savings of a wide variety of ECMs and to compare the
predictive accuracy of design-phase models with models that are carefully
tuned to monitored building data.

Twenty eight commercial buildings were selected to participate in Energy

Edge. All but two were new construction. Building types include large,
medium, and small offices, a grocery, a warehouse, a convenience store,
restaurants, a nursing home, retail stores, medical offices, schools, a motel,
and a strip mall. Funded ECMs include a wide range of envelope, lighting,
HVAC, and controls measures.

In its demonstration aspects, Energy Edge is intended to encourage owners

and developers to exceed MCS requirements with enera-efficient
technologies. Four "Sponsors" -- the Oregon Department of Energy, the
Washington State Energy Office, Portland Energy conservation Inc., and
Pacific Power -- assist in administering the Energy Edge project.

In the area of research, Bonneville decided to pay spccial attention to the

methods for determining actual energy savings. Typically, design-phasc
computer analysis must deal with relatively little information about the
reality of a specific building and its operation. Energy Edge attempts to
expand the present limits of energy modeling by monitoring the selected
buildings in great detail and then using the monitored data to ground the
model in reality.

Though we doubt the reader will find better entertainment elsewhere, we

acknowledge that not all readers will want to read all of this report. For the
benefit of those semi-readers, we provide this guide to the report:

Conservation program managers and others with a program perspective

will find the Modeling for Conservation Programs chapter of most
interest.

Beginning and intermediate modelers should read the Philosophy of

Modeling chapter. We hope this will help deepen their understanding
of modeling assumptions and error.

All modelers should read all of Part 2, Technical Guidelines. This

section proposes default assumptions and modeling requirements for
Bonneville's Energy Smart project. It also proposes new requirements
for model documentation.

The last chapter of Part 1, Technical Overview of Energy Edge

Modeling, provides both program managers and modelers with
interesting, and perhaps, useful information about the role of modeling
in the Energy Edge project. The last part of this chapter summarizes
lessons we learned from the modeling. We hope that this section will
save interested parties the discomfort of first-hand experience.

In proof-reading our work, we have found that our tone often sounds
excessivelynegative when we discuss modeling error, accuracy, and reliability.
Yet we are avid modelers, and we firmly believe in the value of computer
simulation for energy conservation programs. This contradiction is due to
the fact that we intend this report, and especially part 2, to be an in depth
discussion of the more troublesome aspects of modeling. The less
troublesome aspects do not require extensive treatment.

We wish to acknowledge the invaluable assistance of Phoebe Caner of

Seattle City Light, who updated her work from an 1985 unpublished manual
for modelers. Much of that work has been incorporated in part 2 of this
report.

Note on gender usage: Whether to use "his", "her", "hisher", or "their" is a

common problem in report-writing. We are uncomfortable with the
grammatical error in using "their" to modify a singular subject; we are weary
of the historically-correct reversion to "his" to salve this discomfort; and we
would feel silly to litter this report with "hisher". Therefore, we have
elected to use "her".
Computer simulation of building energy performance can be an expensive
endeavor. For this reason, program managers wish to know when this
expense is warranted. In this report we address this and other questions
related to modeling. When can modeling provide useful information? Who
can make best use of that information? When are methods other than
modeling better suited to the task at hand? What can modelers do to
maximize the accuracy and reliability of their models? What can program
managers do to ensure quality in the modeling phase of their programs?
What other program activities are needed to increase the benefit of the
modeling?

Modeling is not necessarily the optimal tool for all energy analysis. It
should not be over-used. It should not be used when a more simple or less
costly method of analysis can yield adequate results. Similarly, a complex
software program should not be used when a simple one can adequately
address a specific building or ECM. Complex programs do not necessarily
yield more accurate results. (Conversely, a simple program should not be
used when a more complex one is needed to adequately address specific
ECMs or buildings.)

In general, modeling is best suited for the analysis of ECMs that either
interact with or directly affect HVAC performance. An important exception
to this generality is daylighting lighting controls. Modeling is overkill for
analysis of ECMs that yield to simple manual calculations.

Modeling can be no more accurate than the assumptions that lie behind
both the proposed building and the baseline building models. Even though --
the model performs complex calculations accurately on these assumptions,
the result will be misleading if the assumptions are faulty. This report
provides guidance for development of some of the more critical assumptions.
However, we cannot address all possible assumptions for all possible
buildings. Therefore much of the success of modeling necessarily rests on
the experience, skill, and integrity of the modeler.

ECM savings are estimated as the difference between two models--the as-
designed model and the baseline model. Modelers typically turn most of
their attention towards the as-designed model. But we have found that the
baseline model is one of the most significant sources of discrepancies
between the savings estimates of different modeling phases (or modelers).
We provide guidelines for baseline model development in part 2 of this
report.

Models are complex, and are very subject to error. Models are to error as
sponges are to water. We intend with this statement to inject realism, not
pessimism, into the modeling activity. Computer simulation is often the best
available method for estimating ECM savings. But it is a method that must
be subjected to rigorous quality control. We make several recommendations
concerning quality control:
 We recommend that the Energy Smart modelers' qualifications be
enforced. If the infrastructure cannot supply enough modelers who
have those qualifications, then supplemental required training should
be considered.

We strongly recommend that every model o n which important decisions

are to be based should be reviewed by a competent modeler. Review
should include the baseline model as well as the as-designed model.

The more complex the software program, the more likely input error
is to occur. Modeler experience and budget must increase as the
complexity of building and software increases. A carelessly or
inexpertly-used complex program may yield less accurate estimates than
an equally carelessly used less complex program.

The modeler should compare the end-use energy use indices (EUIs) of
every baseline model to statistical data for similar building types.
Significant discrepancies should be investigated. This report provides
the statistical data as well as guidelines for this comparison.

For retrofit projects, modelers should compare the baseline (the

building before retrofit) to historical billing data. Significant
discrepancies should be investigated.

The modeler should comprehensively document the model.

Much work remains on evaluating the reliability of model estimates. There

have been a number of studies comparing modeled end-use energy
predictions to monitored end-use data. But there have been few that
attempt the comparison of modeled ECM energy savings with monitored
data. But the latter comparison is the crux of evaluating modeling as a
program tool. Program managers also wish t o know what level of
monitoring is necessary. We know that models informed by monitored data
yield different results than uninformed models, but we cannot prove that the
former results are better.

An accurate model can be useful on several levels. Most commonly, it is

used to estimate the energy savings of specific ECMs. Financial decisions
are based on these estimates. Slightly less obviously, design decisions can
benefit from these estimates. The model can often illuminate obscure
aspects of the designed building performance. Sometimes the model can
uncover design weaknesses o r error.

Can modeling be used for program evaluation? Perhaps. More research is

needed on the statistical reliability of model predictions. The Energy Edge
project case studies indicate a strong variability in the results from different
models, modelers, and model-phases. But this variability may tend to
disappear and the estimates become more reliable when applied to a very
large number of buildings.

Modeling is only one aspect of the technical tasks in a successful

conservation program. Model estimates cannot be reliable for many ECMs
if the ECMs are not commissioned and maintained properly.
("Commissioning" can include many different tasks, but for our purposes
here we define it simply as ECM performance verification.) Energy Edge
provides many examples of the failure of non-commissioned ECMs.
PART 1 THE CONSERVATION PROGRAM PERSPECTIVE

1. MODELING FOR CONSERVATION PROGRAMS

A primary purpose of this report is to give energy conservation program

managers information that will better enable them to wisely use computer
simulation of buildings in the programs they create. To achieve this purpose
we will discuss the strengths and weaknesses of computer simulation, how
to maintain quality control over the modeling process, and what other work
should be done to increase our understanding of the value of modeling for
conservation programs.

1.1 THE STRENGTHS AND WEAKNESSES OF COMPUTER

SIMULATION

Computer simulation of building energy behavior is primarily useful for

understanding the operation of the building heating and cooling systems.
Energy end-uses that do not impact HVAC can be analyzed with a
spreadsheet, other software, or with a pencil and envelope as effectively as
with a building energy simulation software package. Outdoor lighting,
elevators, and domestic hot water are examples of such end-uses.

In evaluating ECMs, the modeler must judge whether the candidate ECMs
are likely to have significant interactive effects on HVAC performance. If
so, computer simulation can analyze these effects more efficiently,
comprehensively, and accurately than any other available method.
Simulation packages excel at handling complex and exhaustively repetitive
arithmetic calculations. Depending on the simulation package used,
simulation can also be an effective tool for dealing with multiple end-use
schedules--whether or not these can be expected to affect HVAC
performance.

Computer simulation also offers the advantage of standardizing the energy

calculations. Most simulation packages are based on well-researched
ASHRAE algorithms. (However, it is interesting, given the common basis
of most programs, that different programs can give such wildly different
estimates.)

When is it inappropriate to utilize computer simulation for energy analysis?

Certainly it is inappropriate when the only ECMs of concern are those that
have no effect on the HVAC systems. We suggest that there is another type
of ECM for which simulation analysis is inappropriate. This category
includes those ECMs about which the modeler can only guess baseline and
design behavior. We would include ECMs that affect infiltration and
occupancy sensor lighting controls in this category.

In the case of infiltration-related ECMs (e.g. vestibules, caulking, and vapor

barriers), HVAC theory is inadequate to give the modeler a firm basis for
input assumptions. In the Energy Edge program, we have taken the tack of
not modeling such ECMs, and noting that the fact that they haven't been
modeled adds a measure of reliability to the energy savings estimates for the
remaining ECMs.

In the case of occupancy sensor lighting controls, the problem lies in the fact
that both baseline and ECM lighting schedules depend entirely on modeler
judgement. The model doesn't really do anything but process the modeler's
assumptions. Secondary HVAC interactive effects are simulated, but the
primary lighting savings are arbitrary!

Simulation software packages have other weaknesses. If modelers maintain

awareness of these weaknesses, they can minimize some of the nastier
consequences. First, most software packages do a relatively poor job of
analyzing electrical demand in a building. This is because most utilities
assign demand charges based on the maximum power draw during a 15
minute period. O n the other hand, most software packages view peak
demand as the concurrent average hourly maximum for the various end-
uses.' Hourly simulation is somewhat more accurate in predicting the peak
demand of a class of buildings (e.g., offices) because of the diversity of
schedules and operation among the buildings.

In addition, most programs do a poor job of modeling morning warm-up.

Where resistance heat is used, these programs tend to spread the actual
spike over one o r more hours. So, in most cases the simulation
underestimates the demand.

This leads to a corollary weakness--using the model to size equipment. In

general, energy simulation programs are not designed to size HVAC
equipment. Equipment usually is designed to meet near peak loads. But,
as stated above, the program addresses peak average 1oadss2 This is not t o
say that the computer cannot be a useful tool for sizing HVAC equipment.
The point is that there are programs specifically designed to analyze HVAC
loads and t o size equipment of adequate capacity to handle these loads.
Having said this, we suggest though that the simulation package can
sometimes be useful in calling attention to gross over o r undersizing of
equipment. If the simulation output equipment selection is sharply at odds
with the designer's selection, the modeler should discuss this with the
designer. Either the model or the design might have a serious error that

It is important to understanci the concept of average hourly

maximum power as applied to modeling. For a specific end-use, such as
lighting, this is the installed wattage multiplied by the average percentage
usage during the schedule hour of maximum usage. As an example, this
hour might be 4 PM on weekdays, which is the average hourly power of
all annual weekday 4 PMs.

* Some energy simulation programs can be used to size HVAC

equipment providing the modeler has accurate design information,
understands the actual mechanical system zoning, enters schedules that
reflect peak load conditions, and runs a design day analysis. However,
since design information is often incomplete during design-phase energy
analysis, we advise modelers to leave equipment sizing to the HVAC
designer.
cooperation may discover.

Another weakness lies in absolute predictions of building energy

consumption. A Bonneville-sponsored study found that various
model/modeler groupings estimated energy consumption for small offices to
be between 66% and 127% of the utility billings (when the modelers had no
feedback from utility bills). (See reference 3.) The range for large offices
was 37% to 120% of the utility billings. Perhaps this degree of accuracy is
acceptable to some users of the information. However, we have experienced
several upset building owners who are shocked to see their utility bills
coming in 50% higher than the model estimate. We recommend that the
modeler clearly state that the simulation estimates are estimates only--
generated with the best information available at the time.

Most commonly used simulation software packages do not do a good job of

simulating the effect of thermal mass on building performance. Similarly,
most packages do not address light-dimming controls in day-lit buildings
well, if at all. Many modified bin method programs do not address solar
gains very well. Some programs have very narrow limitations for the number
of building zones that can be input. The modeler's task is to consider the
specific building to be simulated and determine what building features are
likely to be significant drivers of energy behavior. In addition, she must
consider which ECMs are likely to be of particular interest. Having
determined these things, she then must match the software package t o these
features and ECMs. If the modeler is not competent to use the selected
level of software, she should not attempt to model the building.

Controls for HVAC and lighting can be a fertile ground for ECMs.
However, the modeler must maintain great humility when simulating these
ECMs. Most simulation packages are quite crude in their handling of
controls. Controls that yield savings by shaving fractions of operational
hours don't fit easily into the hourly structure of even the powerhouse
programs. Other controls save energy with relatively complex, subtle, and
infinitely variable logic. Simulation packages on the other hand are
necessarily limited in the number of controls variations they can consider.
Therefore, the modeler typically finds herself sculpting either the program
or the ECM or both to fit each other. At best, such sculpting decreases the
accuracy of the ECM analysis. At worst, it gives totally invalid and
misleading results. In addition, most simulation packages assume ideal
controls operation. Since as-built conditions are typically somewhat less
than ideal, it is often difficult o r impossible to simulate actual operation.

Energy management systems (EMSs) are often considered as ECMs. We

lament this tendency. An ECM should describe specific control sequences,
not an overall EMS. We suggest that the incremental control changes
themselves should be considered as the relevant ECMs, rather than the
entire control system. For example, since most energy codes now require an
automatic means of setting back space temperature setpoints during
unoccupied periods, this function of an EMS should not be considered as an
ECM. On the other hand, complex control of multiple chillers serving
primary cooling load and an ice storage system, with integrated waterside
economizers, exceeds the capability of conventional control systems.3 Such
control--and the hardware, software, maintenance, and training that make it
possible--should be considered as the ECM, not the entire control system.

A final weakness of energy simulation programs is that, like any computer

program, they have very limited capabilities to compensate for bad
assumptions or sloppy input.

We have highlighted many of the pimples in simulation programs. Having

done this, we must point out that often there is no better alternative.
Though the electrical demand estimates of a model may be faulty, there is
usually no better tool for making these estimates. Though the thermal mass
algorithms are imperfect, they certainly are better equipped to deal with this
performance parameter than are hand calculations. We suggest that the
modeler follow this procedure in dealing with these weaknesses:

1) Consider the specific building to be simulated and determine what

building features are likely to be significant drivers of energy behavior. In
addition, consider what ECMs are likely to be of particular interest.

2) Consider whether any of these features or ECMs can better be analyzed

using other methods than conventional building energy simulation programs.
Alternatives might include hand (or spreadsheet) calculations, other
computer programs (such as daylighting o r ice storage programs),
manufacturer's analyses (be cautious here), and so forth. Sometimes a
hybrid approach can be useful. For example, a daylighting analysis program
could be used to generate lighting output that could be used as direct input
for the energy simulation program. The energy simulation program then can
be used to investigate the interactive HVAC effects of the daylighting
controls.

3) Having determined these things, if energy simulation software is still

indicated, select the software package that most clearly addresses these
features and ECMs. Also, select a modeler who is competent with the
selected software.

4) If possible, check the results of any analysis with another type of

analysis. For example, coarse hand caIculations might serve as a check on
the reasonableness of the energy simulation results. We recommend some
version of this with any modeling work. Later in these guidelines we will
address the subject of reasonableness checks in much greater detail.

One can argue that, increasingly, EMSs are becoming common

practice for large buildings. Insofar as this is true, it may be
inappropriate to consider one as an ECM. However, even if they are
common practice, using them to their full potential is certainly not
common practice. Perhaps conservation programs should concentrate on
this aspect of EMSs. This is a decision for conservation program
managers.
1.2 QUALITY CONTROL

Part 2 of this report addresses things modelers can do to'increase the quality
of their work. In this section we address what program managers should do
to ensure the reliability of the modeling tool. But before we can discuss
reliability, we must consider what we mean by model accuracy and reliability.

The modeler always works with incomplete information. Usually she works
with a standard weather file rather than with building-specific weather data.
Usually she must guess at lighting and equipment usage schedules as well as
equipment power densities. Eventual HVAC system operation is always
unknown at the beginning of a modeling project. Also, no energy analysis
software can exactly simulate the complexity of a real operating building.

With all of this missing or inaccurate information, and with imperfect

software, the modeler cannot hope to precisely estimate building energy
consumption. What can she expect? She, and the program managers who
depend on her, can expect to get a reasonably accurate picture of building
performance and ECM savings under standard weather and ideal operating
conditions.

Quantifying "reasonable accuracy" is difficult, and varies depending on the

complexity of the building and its ECMs. We believe it is normally possible-
-with some billing data, competent modeling, and a good choice of software-
-to estimate annual building energy consumption within about lo%, and
seasonal consumption within about 25% of the actual measured
consumption. Without the feedback of billing data, model estimates can
easily vary by 50% or more from the eventual billed data. We cannot
comment on the corresponding reliability of ECM savings estimates since we
know of no research that has been done comparing estimates to actual
measured ECM savings.4

Computer simulation of building energy performance uses an incredibly

complex series of modeler inputs, software algorithms, and computer
operations. Error is inescapable. But we can take steps to avoid or catch
the most serious errors--those errors that can lead to wrong program
decisions. Unreliable energy savings estimates (i.e. estimates that differ
significantly from actual measured building performance) can come from
several sources. Modeling error is only one of these sources. Others
include:

Different operating conditions

0 longer hours, more people, higher internal gains
0 different infiltration loads
o different weather or indoor temperatures
o different operating profiles

The Energy Edge project has gathered extensive building end-use

data. But it has not attempted to directly measure individual ECM
savings. The latter is a complex monitoring task that involves a variety of
on-off, before-after, and other experimental approaches.
 o different controls

Funded ECMs not installed

Improper ECM operation

o hardware, software, or calibration problem
o improper installation
o inadequate design
o poor controls

Modeling error generally comes from some level of inexperience o r

carelessness. Program managers can take steps to guard against these.

Inexperience is a relative term. Different software programs and different

buildings require different levels of modeler experience. Bonneville has
written qualifications to be required for the lead modeler in Energy Smart
projects (reference 13). These should be enforced both in letter and intent.
A word on intent: the qualifications state that, "The specific qualifications
may be satisfied by several people possessing different skills and experiences
which, when taken together, meet the intent of the 'lead' requirements."
The intent of this statement is to not require superwoman modelers when
all the required expertise is available and used cooperativeIy within one firm.
A firm does not satisfy this intent if it has design personnel who have no
contact with the modeling projects.

Carelessness is also a continuum. We are a11 careless to a greater or lesser

extent. Every modeler makes mistakes, and even the best modelers
sometimes make major errors. The single best antidote for this is to have
a cornpeten t modeler (other than the project modeler herself) review every
modeI. There is no substitute for subjecting a model input and output to
a second pair of sharp eyes. Perhaps not all projects warrant the expense of
such a reviews5 But we suggest that if a project is not worth the expense
of a review, then it may not be worth the expense of modeling in the first
place. Note that effective review of a model is difficult o r impossible if the
modeler has not prepared comprehensive documentation as described in
part 2.

Finally, we have prepared the modeling guidelines in part 2 in order to

improve the quality and consistency of modeling assumptions. Though these
guidelines are not appropriate for all situations, they provide a framework
that should be especially helpful to less experienced modelers. We believe
that the active support of program managers for use of these guidelines will
help make modeling a more reliable tool for conservation programs.

Review of a model normally takes from two to eight hours,

depending on the simulation software, building complexity, and quality of
the model and documentation.
1.3 RECOMMENDATIONS FOR FUTURE WORK

There is much we don't know about the reliability of computer simulation

for ECM savings estimates. Modeling has been, and will likely continue to
be, a major tool for design assistance and program evaluation. Therefore,
we would do well to support future research into such basic questions as:

Do better-tuned models give more accurate ECM savings estimates?

How reliable are the ECM savings estimates at various phases of

modeling?

Does arithmetic adjustment of model estimates using bilIing data give

reliable ECM savings estimates?

Recent work with short-term energy measurements indicates that

conventional theory may be defective in dealing with envelope thermal
transmission (references 2 and 14). Is this work valid?

What are realistic default assumptions for infiltration in different

building types?

Can simple arithmetic adjustments be used to account for the

interactive HVAC effects of lighting ECMs? If so, can we dispense
with modeling whenever only lighting ECMs are considered?

What are the relative cost-to-benefit ratios of modeling and

commissioning?

Are certain classes of ECMs no longer cost-effective when the cost of

modeling and commissioning are included? Which classes?

In addition, we note two questions that are specific to Bonneville Power

Administration programs:

How reliable are the ECM savings estimates of the Prescriptive

Can the ELCAP End-Use Index (EUI) tables be expanded by either

direct end-use measurement. o r tuned modeling to address other
climatic areas and newer construction? (See references 5 and 15.)

We also recommend consideration of a possible follow-up step to aid

consistency among modelers. This step is the creation of a library of
annotated DOE2 input files for various commercial building types. Each
input file would serve as a sort of check-list of relevant inputs with
recommended defaults appropriate to that building type. The file would be

The Prescriptive Path was created for the Bonneville Power

Administration Energy Smart project. Based on a series of prototype
building computer simulations, it defined energy savings for a number of
simple ECMs in small buildings.
liberally sprinkled with comments that explain the significance of the inputs.
The file library could reside in magnetic form on the Electric Ideas
Clearinghouse bulletin board.
2. T H E PHILOSOPHY OF MODELING

2.1 T H E PHILOSOPHY OF MODELING ASSUMPTIONS

"Assumptions" is not necessarily a dirty word. At every stage of modeling,

the modeler must make assumptions. Even when the modeler tries to avoid
making assumptions by letting a simulation program revert to default values,
she is still making assumptions. In that case, the assumption is that the
software authors were correct in their assumptions. Often the software
default assumptions are poorly documented or inappropriate for the specific
building at hand.

There are several issues here:

1) Assumptions are inevitable.

2) By avoiding the responsibility of making assumptions, the modeler may

inadvertently incorporate poorer, hidden assumptions.

3) Assumptions range in quality from good to bad. Or, from another

perspective, they range from believable to far-fetched.

4) The modeler's task is to identify when inadequate information forces

some level of assumption. Having done this, she must use the best tools
available to ensure that her assumptions are at the high end of the various
scales of quality. These tools will include deductions from O&M audits,
deductions from monitored data, data from other similar buildings,
engineering rules-of-thumb, engineering judgement, and so forth. Somewhat
obviously, the sharpest tools should be used in preference to the dullest
ones. We would consider rules-of-thumb and engineering judgement to lie
towards the duller end of the spectrum. However, for design-phase
modeling these may be the only tools availablee7

5 ) Assumptions have maximum credibility when they are supported by

several of the above tools. We recommend that the modeler check the
reasonableness of any significant assumption with a second method whenever
possible. In a later section of this report, we will present guidelines for
reasonableness for many of the more significant areas of assumption.

6) Assumptions may and should change as the level of knowledge about

the building (new construction) increases. Information about HVAC
operation, end-use schedules, etc. gains greater definition as the project
progresses. Much more information is available for an as-built model than
for the design model. Similarly, more is available late in the design phase
than early in the design concept phase. Since the modeling work may span

When we refer to design-phase modeling throughout this

document, we include the retrofit equivalent--ECM conceptual design
modeling. Both of these are distinct from the modeling that occurs later
in a project. These later models generally benefit from more accurate
knowledge about the building and its operation.
several weeks, months, or even years, the modeler must use her judgement
and maintain integrity when deciding what assumptions warrant adjustment.

7) Since the simulation results may say more about the assumptions than
about the building, the modeler must clearly document all assumptions. She
must clearly state the assumption. She must clearly state the source of the
assumption--including whether it is a program default. She must include all
relevant calculations, communications, measured data, etc. Complete
documentation allows other analysts to develop a confidence-level in the
simulation results. It may also serve as a basis for a second pair of eyes to
discover a potentially significant modeling error.

2.2 T H E PHILOSOPHY O F SIMULATION ERROR

There is a popular bumper sticker that says, in essence, unfortunate things

happen. This observation applies to modeling. Errors happen. Every
modeler makes mistakes. Every energy simulation program has bugs.
Building information is often wrong. Error can arise from the mass of
model input, from lack of understanding about the program or the building,
from building changes, and from mis-tweaking (more about that in part 2).
In sum, every model contains errors. This is true no matter how many times
the model has been examined, massaged, and corrected. Every model
contains errors.

Does this mean computer simulation is useless, or worse? No. Computer

simulation is still the best tool available for many energy analysis tasks.
However, it does mean that we should maintain humility and skepticism
about simulation results. And it also means that we should carefully
consider quality control procedures.

We cannot hope to catch all modeling errors. But we can certainly strive to
eliminate the most significant ones. While it may not be important that an
R-30 roof was input as R-28, it will be critical if a hydronic heating system
was input without a boiler or other heating source. There are relatively
straightforward quality control procedures that would catch the latter error.
The most helpful procedures entail some level of investigation of model
input and output reports and reasonableness comparisons of end-use output
with typical values for these end-uses. We discuss these procedures in some
detail in part 2 of this report.

One final dose of error philosophy: The yet-to-be-written book, "The Inner
Game of Modeling", makes the point that since error is a fact of life, it is
most constructive to just notice error and learn from it. Embarrassment,
shame, cover-up are all counter-productive. This report is in fact partly an
exercise in noting and learning from error. The Energy Edge project has
been a generous teacher.
3. TECHNICAL OVERVIEW O F ENERGY EDGE MODELING

3.1 MODELING METHODOLOGIES

A wide variety of energy simulation software packages (and modelers) have

been used in the Energy Edge project. These programs fall into two main
groups--modified bin method and hourly analysis programs. Several of the
Sponsors used bin method models for their initial screening of buildings.
All of the Sponsors used hourly programs for the design model and
subsequent models of buildings selected for Energy Edge participation. Not
all of the models for a given building were necessarily run by the same
modeler.

Selection of modeling programs was generally based on the following

criteria:

1) SCREENING: Initial screening of multiple applicants for Energy Edge

funding was done with quick, simple bin method programs. Though it was
suspected that these programs are generally less accurate than the more
complex hourly analysis programs, it was felt that they nevertheless provided
a reasonable basis for comparison among buildings. In addition, it was felt
that an hourly analysis simulation that is quickly and carelessly prepared
would likely be more wildly inaccurate than a similarly careless bin
simulation.

2) AS-DESIGNED MODEL: The as-designed model is the model that

incorporates the building design information at whatever its level of
completion. This level of completion might vary from concept design to
complete construction documents. Once a project was selected for
participation in Energy Edge (and sometimes before it was selected), an as-
designed model with baseline was run. This model uses an hourly analysis
program. Several different programs have been used. These include
DOE2.1 (B, C, and D versions), TRACE, ADM2, and ESP. DOE2.1 was
used most frequently.

3) AS-BUILT MODEL: After the building is constructed and has

operated for a period of at least 6 months to a year, an as-built model is
generated. The intent of this model is to capture the most up-to-date
knowledge about the building and its operating systems. Data for the model
are gathered from construction inspections, as-built building drawings,
operations and maintenance (O&M) audits by the Sponsor, and occasionally
from commissioning activities.

With only one exception, all Energy Edge as-built models used DOE2. The
one exception involves a high-rise building with a complex ice storage system
that cannot be adequately simulated with the present version of DOE2.
(The "E"version, due to be released in the autumn of 1991, will address this
type of system.) The as-built model for this project used the TRACE
program, with some custom-written algorithms to deal with the complexities
of the ice storage.

4) TUNED MODEL: Bonneville, in a quest for maximum analytical

reliability, has carried the Energy Edge modeling one step beyond the as-
built model. Extensive operating end-use data are collected for each Energy
Edge building. The model tuning adjusts the as-built model through a series
of iterations to calibrate it to the monitored data. Though we have
developed three different tuning methodologies (references 1 and 2), all
three share the common approach of adjusting model input using monitored
data until the simulation output matches monitored data within set
tolerances.

The primary types of monitored input are site weather, end-use schedules
and power densities for loads such as lighting and equipment, and operating
schedules for HVAC equipment, The simulation output of greatest interest
is HVAC end-use energy consumption. However, other end-uses must be
tuned before HVAC tuning can be successful.

Model tuning relics on the hypothesis that if a model is calibrated such that
its end-use energy consumption estimates approximate measured end-use
consumption, then the model predictions of ECM and total building energy
savings are more accurate than for an untuned model. We must point o u t
that though this hypothesis seems intuitively correct, it has not been
rigorously analyzed and tested.

A secondary purpose of the Energy Edge tuning work is to evaluate the

program benefits of spending more time and money o n the as-built and
tuned models. Work to date indicates frequent and sometimes significant
discrepancies among the energy consumption estimates of the four modeling
stages (screening, as-designed, as-built, and tuned models).

3.2 COMPARISON O F MODEL SAVINGS ESTIMATES

Various studies have documented the variability among the predictions of

different models and modelers. (References 3,4.) This document does not
attempt to rehash this territory. The guidelines developed in this report
focus on the input uncertainties identified in these assessments.

The Corson study (ref. 3) pursued two main research paths.8 The first had
11 different modelers use five different software packages to model four
different building types. The modeling was done in three cycles--each cycle
giving the modelers more information about the building. The first cycle
gives the modelers only basic building characteristics information. The
second cycle gives utility billing data. The third cycle gives end-use
monitored data. The tentative conclusions of this portion of the study arc:

1) The billing data feedback generally resulted in simulations that more

nearly match each other and more closely approach measured data. Cycle 1
annual energy as modeled by DOE2 ranges from 56% to 148% of the

Note that this is a massive study. We d o not claim to d o it justice

with our interpretation of the results as presented here. O n e of the main
conclusions of the study is that much further work is required ro add
meaning and certainty to the study results.
respective billing data. (These are larger differences than with some of the
bin method software packages studied.) Cycle 2 percentages of billing data
for the DOE2 models range from 93% to 106%.

2) The end-use data feedback also resulted in an improvement in

simulation results (for most programs tested) relative to the simulations
done with billing data feedback. However, these cycle 3 results are not
nearly as marked as the cycle 2 results. Also, the DOE2 annual kwh
differences increased for three of the four buildings modeled.

3) There is great variability in results among the different software

packages and among modelers.

The second main research path involved running multiple building

simulations for two of the building types to test model sensitivity to 25
different input parameters.9 The tentative conclusions of this portion are:

1) The tested parameter that is the strongest driver of energy consumption

in the two buildings is the HVAC system selected. There is great variability
among the software packages as to the significance of HVAC system
selection.

2) Parameters of secondary importance include:

wall insulation in the small building (R11 increase)
supply fan operating mode during unoccupied hours
internal load schedules (2 hour increase in lighting, equipment,
and occupancy)
building location (mild versus severe climate).

3) Tested parameters of minimal sensitivity generally include:

building orientation (rotated 90 degrees)
floor insulation (R-value increased 30%)
roof insulation (R-value increased by R-19)
south wall glazing area (small retail--convert 30% of opaque wall
area t o glass; large office--reduce 30%)
glazing R-value (increase 30%)
glazing shading coefficient (decrease 30%)
outside air CFM (increase 50%)
interior lighting power density (decrease 30%)
interior lighting heat to space (decrease 30%)
hot water usage (increase 100%)

Experience from our work does not always agree with the results of
this study. We have found that some of the parameters shown in this
study t o be of minimal significance as energy drivers have sometimes been
of major significance with specific buildings. Also, starting values for the
parameters may have a strong role in determining importance. For
instance, a change in wall insulation R-value from 0 to 11 could be
critical, whereas a change from 11 to 19 could be insignificant. Thus we
warn against offhandedly ignoring the effect of any of these parameters.
Experience and informed judgement are always valuable tools.
 economizer operation (turn off)
supply air volume (increase 30%, holding outside air CFM
constant)
cooling COP (increase 30%)
space setpoint temperature (increase winter occupied setpoint by
5 F degrees)
cooling equipment part load performance (increase specified
capacity 50%)
number of occupants (increase 50%)
vacation period, unoccupied (add July)
building mass (change from medium to heaviest category)

In addition to the conclusions from the two main research paths, the study
called in to question the common modeling rule-of-thumb that though
estimates of base building consumption may vary widely among different
models and modelers, the ECM energy savings estimates will show little
variation. The Corson study saw results that are totally opposed to this rule-
of-thumb. The study results indicate that the ECM savings estimates vary
more widely than do the widely-varying base building estimates. In our
Energy Edge work, we have seen some results that also support this
conclusion. This is of global significance! If valid, it means that we
modelers can no longer find solace in the belief that a sloppy base model
can nevertheless yield accurate ECM savings data. And if we modelers can't
find solace in this belief, then the program administrators who routinely base
financial decisions on our simulations must feel some unease also.

3.3 ENERGY EDGE COMPARISONS

Lawrence Berkeley Laboratory (LBL) is responsible for impact analysis of

the Energy Edge program. Part of their work has involved comparing the
design-phase computer models to the tuned models. This report section
owes much to an unpublished LBL memorandum that summarizes the early
results of this analysis.

As the Energy Edge program has progressed, LBL has found substantial
differences in the building systems and operating conditions between early
design intentions and actual occupied buildings. Part of LBL's evaluation
efforts has included tracking these changes. Although this task is hampered
by lack of data, such as clear reference to original modeling assumptions,
LBL has identified and compared key building characteristics (ref. 16).

The main finding of the analysis is that the early design predictions and the
tuned model estimates show tremendous variation in the estimated energy
savings of the ECMs. We see, for example, in an 8500 ft2 office building
that the total energy use of the tuned model is about the same as the design-
phase baseline model, but the total savings by the ECMs were similar to the
design-phase model. However, the savings of individual ECMs differ by as
much as a factor of five between the two models. (See Figure 1.)

There appear t o be several reasons for these differences. At least two (and
to some extent, all) of the ECMs owe their differences to the dissimilarity
in baseline assumptions in the two models.
Figure 1

ECM Energy Savings Comparison

Small Office, 8500 sq.ft.

..............................................

..............................................

Vestibule Roof lnsul Lighting

Wall lnsul Low-E Glass Non-funded
Predicted Tuned eewq2llblcomp.wql
dubal ecm-save

End-Use Energy Consumption Comparison

...................
........................................

....................
....................

Base Edge Base Edge Monitored

Predicted Tuned
eewq2Ublcomp.wql
dubd EUI
In a 4100 ft2 fast food restaurant, we see significant differences between the
ECM savings estimates for four of the five ECMs. (See figure 2.)
In this building the difference in the savings estimates for the lighting ECM
is primarily due to differences in baseline assumptions. The difference for
the hot water heat pump ("ht pump DHW") is due to a mistaken assumption
in the design-phase model regarding hot water usage. The monitored data
allowed us to improve this assumption in the tuned model.

In a 3025 ft2 medical office, we see relatively small differences between the
ECM savings estimates for three of the four ECMs. The fourth ECM shows
a tremendous difference between the two estimates. (See figure 3.)
In this case the difference appears to be due to the fact that the funded
lighting contrdls (occupancy sensors and daylighting controls) are not
working. The tuned model simulates this as-operating condition, whereas
the design-phase model assumes proper operation.

Each difference between the performance of the ECMs in the design-phase

versus tuned models can be assigned to one or more of the following four
categories:

1) Improper or different ECM operation (hardware, software, o r

calibration problem, incorrect installation, poor design, poor control,
different controls, etc.)

2) ECM not installed (or new ECM installed)

3) Different assumptions about operating conditions of ECM o r building

(operating profiles, weather, infiltration, occupancy, etc.)

4) Modeling technique differed (error, different software, different

baseline assumptions, etc.)

From the analysis of the three buildings described above, it appears that
categories 1 and 4 are the main causes of differences in ECM savings
estimates between the two modeling phases. However, we also see examples
of the other two categories.

3.4 LESSONS LEARNED FROM MODEL CALIBRATION

The Energy Edge project has required us to examine a small number of

building models in exhaustive detail. For most buildings we have had the
benefit of well analyzed end-use monitored data, and sometimes
supplementary measurements of ECM or system behavior.

For most buildings we are charged with calibrating the model end-use
monthly, seasonal, and annual energy consumption to the monitored data.
The extent of this activity is unique in the annals of energy modeling, and
has enabled us to catch a glimpse of certain inalienable modeling truths.
We have already alluded to some of these in the appropriate sections. But
we shall summarize all of these truths in this section.
Figure 2

ECM Energy Savings Comparison

Fast Food Restaurant, 4100 sq.ft.

..............................................

..............................................

Ht Pump DHW Ext tight Non-funded

Heat Rec Int tight
Predicted Tuned eewqUblcomp.wq1
rncd ecm-save

End-Use Energy Consumption Comparison

Fast Food Restaurant, 4100 sq.fk.

..............................................................................

..............................................................................
...................
...................

Base Edge Base Edge Monitored

Predicted Tuned
r
Int Light Heating Cooling eewqablcornp.wq1
rncd EUI
Fans/Aux DHW Plugs
.Figure 3
ECM Energy Savings Comparison
Medical Office, 3025 sq.ft.

..................... ..................... ..................... .........................

EconfHi COP Lighting

Predicted Tuned eewq2Vblcomp.wql
sisk ecm-save

End-Use Energy Consumption Comparison

Medical Office, 3025 sq.ft.

.........................................................................................................................................

........................................ ..............................................................................
..............................................................................

...................

...................

I 1 I

Base Edge Base Edge Monitored

Predicted Tuned
Int Light Heating Cooling eewqEilblcomp.wq1
dsk EUI
FanslAux r A Out light Plugs
 1) One cannot overemphasize the importance of the baseline definition.
We have seen that differences in baseline assumptions were responsible for
most of the greatest differences in ECM savings estimates between the three
design-phase and tuned models discussed in the previous section. Codes
simply do not address all the aspects of model input. Therefore modeler
judgement is unavoidable.

Program managers must realize that there is often some degree of

unavoidable arbitrariness in the definition of the baseline model. Different
competent and ethical modelers may make different assumptions about
various baseline parameters. This becomes particularly significant when
these parameters are affected by an ECM. The implication of this is that
the ECM savings estimates must then also have a degree of arbitrariness.

In part 2 of this report we discuss baseline derivation in some detail. Both

modelers and program managers should read that section.

2) Many of the most significant energy drivers can be quantified only by

educated guess--even with extensive monitored data. This is especially true
during design-phase modeling and for new construction. The most obvious
example of this is infiltration, though ventilation air volume and schedule,
internal equipment loads and schedules, and HVAC system control are also
significant. Infiltration and ventilation air assumptions can overshadow most
other inputs for buildings that are subject to a significant heating load.
These assumptions can greatly affect ECM savings estimates, especially when
directly addressed by an ECM.

Often, parametric analysis of the extremes of reasonable values can give the
modeler insight into the importance of these uncertain energy drivers.
Parametric analysis is also helpful in establishing the error bounds related
to specific inputs.

In part 2 we provide some guidelines for these assumptions. However, both

modelers and program managers should realize that the science of
infiltration is crude, and that savings reliability can often be diminished by
this. When a modeIer analyzes an ECM that relies heavily on infiltration
assumptions, she should clearly document the assumptions and the degree
of uncertainty about them.

3) Small buildings are often disproportionately affected by occupant

intervention. We found that it is sometimes impossible to calibrate the
models for small buildings. Occupant manipulation of a single thermostat
or a single window in a one or two zone building can completely mask the
known input parameters. We know of nothing that can be done in the
design-phase models to address this issue. (Though, again, parametric
analysis of the impact of thermostat setting variability may be useful in
establishing the error bounds.) The lesson here is not a solution. The
lesson is one of awareness. End-use and ECM savings estimates from
design-phase models of small buildings may differ significantly from the
actual end-use consumption and ECM savings due to such occupant
intervention.
4) Choice of tuning methodology can strongly influence ECM savings
estimates. We recently completed a model tuning study of a small credit
union in Idaho. In analyzing this building, we used two different
methodologies. One of these methodologies drew o n work by the National
Renewable Energy Laboratory (NREL, formerly Solar Energy Research
Institute, SERI) involving short-term energy monitoring (STEM). This
approach directly measured the building load coefficient and, indirectly, the
envelope overall U-value. The startling conclusion of these tests was that
the measured envelope U-value is about 40% of the value as calculated by
ASHRAE methods! Yes, this means that STEM believes the envelope
behaves more efficiently than as predicted by conventional theory. Assuming
that the methodology is valid, one possible explanation for this is that the
envelope operates as a heat exchanger--heating infiltration air as it filters
through.

Much work remains to validate this methodology, but these results should
certainly give modelers pause. We are not at this time recommending that
modelers depart from conventional methods for U-value calculation.
However, one should note that, depending on which theory is embraced, the
savings estimates for any ECM related to envelope performance may differ
widely.

5 ) Errors happen, happen. The modeler should never assume her model
to be correct. Part 2 deals at length with error checking and model
debugging. Suffice it to say here that models are never correct. They are
only less incorrect as the modeler takes greater care.

6) ECMs often do not operate as intended. Another Energy Edge project

lesson is that ECMs (and, for that matter, buildings) often do not work as
intended by the original design. Dynamic systems (lighting controls,
economizers, energy management systems, other HVAC controls) are the
most frequent examples of this.

Several conclusions can be drawn from this lesson. The first is that design-
phase models that assume proper ECM operation will often not be
indicative of actual ECM savings or building energy consumption. The
second is that, if program managers are serious about procuring ECM
savings, then the programs must include commissioning of the ECMs and
related systems. They must also support intelligent operations and
maintenance procedures for the life of the measures. The third is that
HVAC tuning can be complicated by unintended operation.

7) Simplified analysis can bear useful results. Sometimes a model can be

used for analysis of only a part of a building or for certain ECMs only. This
can greatly simplify the modeling task when appropriate. An Energy Edge
project example of this was an existing large officelretail building with a
complex glass facade and massive construction. We wished to approximate
the potential savings due to air-side economizers and night flushing. As a
first cut, we modeled only interior zones. This was based on the assumption
that since the installed capacity of interior HVAC units grossly exceeded
that of perimeter units, and since the building appeared to be satisfying
heating and cooling loads, then the interior zones represented the vast
majority of heating and cooling energy consumption.

From this crude analysis we found that the ECM savings were not at all
close to the cost-effective criteria. Therefore further analysis was not
warranted, and we had avoided significant modeling effort and expense. The
lesson is that model complexity is not an end in itself. The modeler should
always be alert to the potential for simplified modeling.
PART 2 TECHNICAL GUIDELINES

1. GENERAL MODELING STRATEGIES

1.1 T H E ROLE O F THE ANALYST

Conserving energy requires living on the edge. Excessive amenity levels are
avoided, but so is deprivation. HVAC systems are sized with adequate
capacity but little margin for error. New, relatively complicated and
untested technologies are introduced where the tried and true once reigned.
Designers, manufacturers, installers and building operators and
administrators are forced up against a learning curve.

Energy savings analysis increases investment in conservation by offering

building owners the information needed to weigh first cost against reduction
in annual operating costs. Savings calculations also serve as the basis for
utility funding in demand side management programs. Nevertheless, it is
important that energy conservation not be confused with energy analysis.

Analysis is only a small link in the chain of activities that leads to a

successful ECM. Once estimated savings have been used to select an ECM,
it is effective manufacturing, design, installation and operation that
determine actual savings. The analyst's role is not only to predict typical
energy savings but also to promote actual savings through all phases of the
project.10

Critical parameters, assumed in the analysis, need to be written down by the

analyst to serve as performance specifications, along with assumed operating
setpoints and schedules. The analyst needs then to check design documents,
bids and equipment submittals to assure that the performance specifications
are being met. The design documents, whether they are done by a
manufacturer or a consultant, should indicate that adequate attention has
been assigned to fitting the general technology to the specific project.

Ideally, an ECM description should include, from an early stage, a written

consensus between the designer, customer and analyst of how the success of
the ECM is to be evaluated after installation. For HVAC and controls,
check-out trend logs should be planned in advance wherever possible. The
analyst can also promote clear labelling of equipment, training of the
building operators, and good user manuals. To last, an ECM must be
reliable, convenient and understandable.

lo We use the terms "analyst" and "modelern somewhat

interchangeably in this report. Insofar as we intend a difference, it lies in
usage of "modelern in a more restricted sense of "one who models". The
analyst, on the other hand, is the modeler who also provides design
assistance, encourages commissioning, does hand calculations, etc.
1.2 BALANCING DETAIL WORK WITH TIME FOR CRITICAL
THOUGHT

Computer simulations require so many detailed inputs there may be little

time to sit back and look at the overall picture and still turn a profit. Be
sure to simplify the simulation to an extent that there's time to think about
the results from a distance. Analysis requires judgement, which requires
time for calm reflection.

At first it is tempting to model each schedule, each system, each strange

wall, each unconditioned space and each wall layer with intricate detail. But
the inputs and outputs can easily become unmanageable. Checking inputs
and outputs for reasonableness and careless errors takes time. A huge
number of detailed inputs full of errors is of no advantage to anyone.

The ultimate goal in simulation is to focus any desire for detail o n inputs
that make a big difference. To some degree, the critical inputs vary from
one building or ECM to the next. However, the following guidelines may
be generally useful in reducing the amount of time spent on detail work.
Most of the points that are summarized here are discussed in greater detail
later in this report.

Interior Walls. Do not model interior walls between air spaces

that are roughly the same temperature unless they are needed to
properly simulate thermal mass or light-dimming controls in day-
lit buildings. However, the modeler must take care that she allow
some path for heat transfer from interior zones with significant
internal gains.

Utility Rooms. In general, don't model mechanical (or electrical)

rooms or mechanicaI room walls. The heat loss from the
equipment usually keeps the rooms warm, and that energy is
already accounted for in the HVAC model as equipment
inefficiency.

Permanently Shaded Glazing. Treat windows that are

permanently shaded as north-facing without shade, o r as unshaded
with a low shading coefficient.

Schedules. Since schedules are highly uncertain in any case, don't

create two schedules that are only off by 10% unless that
difference is the specific goal of a control ECM.

Roof and Ground Losses. With tall buildings, pay little attention
to the roof, doorway and ground heat loss characteristics unless
they are directly involved in an ECM.

Hand Calculations. Calculate consumption by domestic hot

water, exterior lights, and equipment located in unconditioned
space by hand (i.e. with a spreadsheet). That way you can change
those calculations without rerunning the program.
 Sensitivity studies. We highly recommend the modeling practice
of running parametric simulations of parameters in question to
test their significance in a specific building. If changing the values
within the range of uncertainty has little effect on the output,
then the modeler can arbitrarily select a mid-range value and
move on.

Lighting Inputs. If lighting is fairly evenly distributed, calculate

a watts per square foot input to use throughout, even for retrofit
baseline simulations.

Spreadsheets. Some modelers find it helpful to develop the

envelope and internal gain inputs on a computer spreadsheet.
(See appendix VI.) Some modelers even go so far as to use a
spreadsheet to help generate their DOE2 BDL files.

Pre-simulation Review. If possible, check the level of detail in

zoning and schedules with any reviewers before you get started.

Preserving Carefully Done Work. If you perform a careful

detailed calculation to create an input, file the calculation
carefully so that you don't have to redo it a month later when
checking your work.

Focus on ECMs. Even with the baseline inputs, focus on the

inputs that will be directly affected by the ECMs to be modelled.

Caution: The net heating and cooling loads on a zone are sacred. The
modeler should not move walls, lights, equipment, etc between zones.
Multiple spaces can be joined into a single zone, but the respective lights,
walls, floor area, etc should be manipulated together, and not capriciously
assigned to different zones.

1.3 NEGOTIATIONS BETWEEN THE MODELER AND T H E

COMPUTER SIMULATION

The good modeler knows how to negotiate profitably with her software.

Part of successful negotiation is arriving at the table with an opinion about

what the results should be. The modeler should have access to information
about typical:

Btu per square foot per year total consumption by building type,

end-use breakdown fractions,

seasonal variation of end-use consumptions, and

a reasonable range of inside space temperatures.

The modeler should also have some sense of what the energy savings for
each ECM will be, either from back of the envelope calculations or previous
modeling experience.

These prejudgments form the modeler's side of the argument. If the

simulation produces results that fall dramatically outside the range of
anticipated results, the modeler's most reasonable first reaction is to try to
bring the simulation into line--always remaining open to the possibility that
the simulation is correct. The order of attack is important, and should be
roughly as follows:

1) Careless Errors in the Inputs. Look for careless errors in the inputs.
Examine both the actual input and the simulation program input echo
reports.

2) Simulation Output. Examine parts of the simulation output that might

lead to some clarification of the difference between simulation results
and the expected values.

3) Understanding of the Simulation Algorithms. Reread the appropriate

sections of the simulation users' manual to check whether you've
understood the simulation properly, and correct any errors resulting
from such misunderstandings.

4) Understanding of the Building or the Design. Think about the physical

process being modelled and whether the model - as described in the
user's manual - has captured the major elements. If not, consider a
change in the inputs or a hand calculated correction to the outputs.

5) Increased Attention to Detail in the Inputs. Consider redoing more

carefully any inputs that were developed in a rush if it seems the
difference might create a more reasonable outcome.

6) Fiddling with Inputs with a High Degree of Uncertainty. Tweak

uncertain inputs within a reasonable range of values to move the
simulation results toward the answers you expected. (This is a complex
and dangerous topic. "Responsible tweaking" requires modeler
experience, judgement, and integrity.)

If the results are still mysterious, the modeler may try a parametric run or
two to make sure she understands how the simulation works on algorithms
related to the source of disagreement. As an example, she might input
infiltration at two different values during unoccupied hours--say, 0.2 and
0.5 air changesfhour. Or she might try two runs--one with 5% minimum
outdoor air ventilation and one with 20%.

The overall strategy is thus first to increase accuracy, then tweak, and finally
to consider that either there is a bug in the simulation or an error in the
preformulated answers. If the simulation has been around for over 10 years,
most of the bugs have probably been worked out. To distinguish between
a bug and a sober lesson in reality, it may be helpful to talk with someone
who is familiar with the software and talk with people knowledgeable about
the technology in question. For unexpected ECM savings, try hand
calculations to see whether the same results can be approached manually.
Do not tweak all of the uncertain inputs before checking for careless errors,
Otherwise you may find yourself in the humbling and inefficient activity of
tweaking back to center after careless errors have been found.

If the modeler doesn't start each simulation with an idea of the reasonable
range of outcomes, she will be at a disadvantage in routing out careless
errors and misunderstandings about how the simulation works. O n the
other hand, a modeler who resorts to unrealistic inputs to generate what
appear to be the "right" outputs may be working from erroneous
preconceptions. It is important to know when to stop arguing and to start
listening.

Computer analysis is helpful not only in fine-tuning ECM savings

calculations within the range of typical values but also in identifying overly
optimistic ECM expectations.

Simulations may also help dislodge outdated notions of energy use indexes
and end-use breakdowns. As design fashions shift, so d o building
consumption patterns.
2. SOURCES O F MODEL DISCREPANCIES

There is a growing body of knowledge that identifies recurring modelers'

biases in constructing building models. End-use monitoring through the
ELCAP and Energy Edge programs has lead to surprising new information
about equipment power densities and lighting and equipment use during
unoccupied hours. Review of design phase modeling for Energy Edge
buildings has also revealed some typical errors resulting from modeler
oversight or misunderstanding of how simulations use input data.

Since these findings may be particularly helpful to modelers, they are

summarized here. Each of these topics is also covered in greater detail later
in this report.

2.1 EQUIPMENT POWER DENSITY

We identify two significant biases that relate to equipment power density:

1) Underestimates from using rules of thumb. Equipment loads (also

referred to as plug-loads and receptacles), tend to be vastly
underestimated. The time-worn rule-of-thumb for office spaces has
been 0.5 w/ft2, not including large main-frame computers. Recent
studies indicate that a power density of 1.0 to 2.0 w/ft2 is more
appropriate.

Overestimates from using nameplate data. When modelers have access

to audit data, the opposite tendency occurs. Use of nameplate capacity
as the connected load in a computer model overestimates consumption.
Many pieces of equipment, like photocopying machines and printers,
draw power in varying amounts through their operation cycle. The
peak, or nameplate consumption, is only experienced for a fraction of
the time the machines are on. As a general rule of thumb, we advise
that nameplate ratings be multiplied by 113 to get the ap ropriate
power density input for typical electronic office equipment. I!'
2.2 UNOCCUPIED EQUIPMENT AND LIGHTING SCHEDULES

Modelers often assume that the lighting and equipment loads go t o zero
during unoccupied periods. Monitoring data have shown that in fact 10%
to 30% of the lighting and at least 30% of equipment loads are on during

" There is much variation among different types of electronic

equipment in the ratio of average to rated power. Personal computers
tend to be constant load machines with average power at 25% to 40% of
rated power. Copiers and printers have ratios that can vary widely, and
that are primarily a function of the duty cycle of the machines. Average
power for such machines is generally between 15% and 30% of rated
power. Because of the preponderance of personal computers in the
receptacles load of the typical office, the 113 rule of thumb is a reasonable
approximation. (Reference 21, Harris et al, and reference 22, Piette.)
unoccupied hours in typical office buildings.12 Similar values apply to
other commercial building types.

2.3 WINDOW SHADING

In the course of reviewing design-phase simulations of Energy Edge

buildings, we found that interior and exterior shading are among the most
commonly overlooked model inputs. Often this occurs either as a simple
oversight o r because the modeler has no information on which t o base
assumptions. Existence and use of curtains are the norm rather than the
exception since glare and overheating tend to occur in direct sunlight. The
modeler should assume, unless there is firm evidence to the contrary, that
the building has interior, occupant-operated shading. Extensive shading by
adjacent buildings o r trees, as well as shading of the building by itself all
need to be taken into account by the modeler.

Computer simulations vary widely in their ability to model internal and

external shading. Compromises often need to be made. For example,
curtains o r shading by adjacent buildings can be modeled by input of an
artificial shading coefficient. The modeler is encouraged to model shading
accurately within the limits of the program, and in no case to ignore shading
just because it is uncertain or difficult to model.

2.4 WINDOW AND WALL U-VALUES

Modelers commonly ignore the effects of window frames and metal wall
studs o n U-values. For multiple paned windows with metal frames that have
n o thermal breaks, an overall window U-value more than the U-value of the
double glazing should be used. Metal wall studs should also be taken into
account when calculating the average wall U-value. Heat loss through walls
and windows is otherwise significantly underestimated.

2.5 HEAT LOSS T O GROUND

Some computer models overestimate heat loss to ground from floors in

contact with the soil. Heat loss to ground from floors occurs primarily at
the perimeter. Calculations based on the entire floor area are likely to
overestimate heat loss. If a simulation calculates heat loss from floor to
ground based on a floor area input, the modeler is encouraged to input the
floor perimeter in linear feet or to input an artificially low U-value. The
artificially low U-value is more convenient if the floor area is used to
calculate infiltration or equipment loads as well as conduction losses. (See
reference 11, p. 111.118.)

l2 The unoccupied fractions of peak lighting and equipment power

appear to be somewhat dependent on building size. Small offices tend to
have lower unoccupied fractions than do large offices.
2.6 OPERATION ASSUMPTIONS

Modelers typically assume that both the baseline and the as-designed
building operate according to design intent. Commissioning and auditing
experience constantly screams to us that this is just not the case! Though
we are not suggesting that the design-phase modeler simulate broken
buildings, we are suggesting that the modeler be aware of this issue and be
humble about her simulation estimates.

Commissioning work in the Energy Edge project and other northwest energy
conservation programs is providing much anecdotal evidence that certain
ECMs frequently do not work according to the designers' intent. Controls-
related ECMs are probably the most common example. For instance, we
have never seen economizer controls on packaged equipment to operate
properly as-installed. Energy management systems typically are only partially
utilized--in direct contrast to the designer's fantasies. Such anecdotal
evidence suggests that computer simulation based solely on design intent is
quite misleading as to the true cost and energy savings of these ECMs.

We recommend that the modeler, in her role as a provider of design

assistance, make herself and the building owner aware of these issues. She
should especially emphasize that model predictions presume proper
installation and operation of the relevant building systems. It is critical that
the building owner (and, for that matter, the energy provider) understand
that it is not sufficient to just analyze and fund such ECMs. Ongoing
attention to operation and maintenance is required.
3. FOCUSING O N THE INPUTS THAT HAVE A SIGNIFICANT
IMPACT

It is important in computer modeling of buildings to focus on the inputs

that are most critical to making an accurate estimate of energy savings from
the ECMs under consideration. Any input directly affected by the ECMs
should be given careful scrutiny in both the baseline and the post-ECM runs.
In addition, glazing U-values and shading coefficient, infiltration, ventilation,
fan schedules, lighting and equipment watts per square foot, and thermostat
settings are always important. The risk of significant error for other inputs
depends to some degree on the size of the building and whether o r not it is
new construction.

3.1 SMA.LL BUILDINGS

In small buildings, critical inputs include roof U-values, heat loss to ground,
and heat transfer through unconditioned spaces such as attics, storage areas,
and garages.

3.2 LARGE BUILDINGS

In large buildings, zoning, economizer controls, air distribution system type,

and multiple-zone HVAC system controls become more critical than in
small buildings, because many of the HVAC zones have no exterior walls
and therefore are cooled even in the winter.

3.3 RETROFIT PROJECTS

For retrofit projects, the greatest difficulty is determining how the existing
equipment is controlled. Manual adjustments to mixed air setpoints and
unoccupied period equipment operation can have a critical impact on
building consumption, but are not easy to ascertain. Air flow rates and fan
and pump consumption may also be difficult to determine if the original
system was dramatically oversized and the system is being operated at flow
rates much lower than were intended.

When applying ECMs to the simulation of an existing building, it is critical

to input non-default values for important HVAC parameters. Supply CFMs,
and chiller and heat pump capacities should be input by the modeler for the
post-ECM run (i.e. carried from the baseline to the post-ECM run) even if
they were aIlowed to default for the baseline, because otherwise the addition
of any ECM affecting peak heating and cooling loads will cause the entire
HVAC system to be resized.

Occupant discomfort in an existing building may be related to a reduction

in energy consumption. Burned out light bulbs that haven't been replaced,
inadequate ventilation, or undersized cooling equipment may save energy
while reducing the quality of the space being served. For such cases, the
baseline simulation may tend to overestimate baseline consumption if it
models normal operating conditions. If the comfort problem is corrected as
part of the ECM installations, modeled ECM savings may be incorrect. For
that reason, it is important in retrofit projects to discuss with the owner any
needs to improve amenity levels beyond the existing condition and to
consider the implication of such needs on ECM savings.
4. SIMULATION INPUT

Proper selection of simulation inputs requires understanding of building

technologies as well as computer models. In this section, recommendations
are made for developing simulation inputs. When an understanding of the
physical process being modeled is critical to developing the appropriate
model inputs, that process is described.

4.1 ZONING

This section paraphrases cxtcnsively from an article in the BLAST News

(reference 9), and blends in our own opinions. We therefore take full
responsibility for any faults in the material, but take only partial credit for
its successes.

4.1.1 The Basics

The goal of any simulation is to take something that is extremely complex

(a building) and to model it as simply as possible yet as accurately as
necessary. One of the most critical steps is the zoning of the building. The
more complex the building, the more important this step becomes.
Buildings with a small number of zones are much easier to manage than
buildings with an abundance of zones.

Aggregation of loads into zones, systems, and plants can have a significant
impact o n energy consumption, particularly for large buildings and buildings
served by multiple-zone heating and cooling systerns.13

Zoning in simulation models is based on, but is not identical to, HVAC
zoning. An HVAC zone is defined by an individual thermostat and that part
of the air distribution system that responds to that thermostat. A designer
will often give individual thermostatic control to several different groups of
occupants in an area that can be expected to have a relatively homogenous
heat balance. A model zone, on the other hand, represents simply a mass
of air o n which a heat balance is performed. There can be one o r many
HVAC zones in a model zone. There will rarely be more than one model
zone per actual HVAC zone.

When describing a building in a model, the first step is to zone the building.
When placing zone boundaries, it is very important to remember that:

A zone represents a mass of air on which a heat balance is performed.

Surfaces, scheduled loads, and controls provide mechanisms for energy

l3 We frequently refer to "multiple-zone systems" in this report. This

nomenclature is not to be confused with the more common term, "multi-
zone system". A multiple-zone system as used here is an HVAC system
that serves more than one zone and responds on some level to multiple
thermostats. A multi-zone is one example of a multiple-zone system;
VAV, dual duct, VVT, and so forth are other examples.
 flow into and out of a zone.

This means, for the model, that the entire zone has one mass of air at a
single temperature. Any surfaces which are assigned to a given zone
exchange energy with the zone air mass and with the other surfaces of the
zone. Any scheduled Ioads and controls which are assigned to a given zone
exchange energy with the zone air mass and with the zone surfaces.

Before discussing ways to zone a building, there are some common myths
about zones which must be dispelled:

Myth 1:
A zone must r e p r e s e n t an enclosed volume.
This is not necessa ry... If a zone has a hole in it, it will not let in
masses of outside air. Surfaces, controls, and scheduled loads are the
only means by which energy can enter or leave a space. Many zones
may be completely described with a single exterior wall and a piece of
internal mass.

Myth 2:
A zone must r e p r e s e n t a continuous volume.
This is not necessary. It is possible to have a single zone consist of
several spaces which are far removed from each other. For example,
all of the bathrooms in an entire building might be combined into a
single zone.

Myth 3:
A zone must a l l be on one f l o o r .
This is not necessary. Very often, it is useful to have a single zone
include rooms on several different floors.

Remember that a model zone is a somewhat abstract thing--it is not limited

by the geometry of specific rooms or of the HVAC distribution system.
However, since buildings are divided into rooms, a zone will typically be a
collection of rooms.

Start With One Zone

One approach to zoning a building is to start with the entire building as one
zone and then subdivide that zone as needed. Any building may be modeled
as one zone, if desired. This would be a very simple model, but it can be
done. If a single-zone model is sufficient for the needs of a project, then
there is no need to go any further in zoning the building.

There are many different criteria which may be used to determine additional
zone boundaries. Five basic criteria are usage, controls type, solar gains,
perimeter or interior location, and fan system type. These five
characteristics are sufficient to define almost all of the necessary zone
boundaries, yet there may often be special conditions which require
additional zones to be created.

1) Zoning by Usage. Rooms which differ greatly in usage typically need

 to be described in separate zones. All rooms which are included in a
zone should have similar internal loads.

Example A
A kitchen with high internal loads should not be grouped into the
same zone as a storage room with low internal loads. If both of these
rooms were modeled in a single zone, the kitchen loads would be
distributed evenly throughout both rooms; it would be as though there
was no wall between the kitchen and the storage room. The modeler
may, however, decide that this distinction is trivial to the major energy
patterns of the building. In this case, the zones could be combined.

Example B
Ten offices with similar lighting levels and controls, occupancy rates,
and equipment loads may be described with a single zone. These
offices may even be on different floors, and they do not necessarily
have t o be adjacent to each other. Since all of the rooms will
essentially behave the same, nothing is lost by combining them into a
single zone.

2) Zoning by Controls Type. Rooms with different temperature control

strategies o r thermostat schedules (or thermostat setpoints) typically
need to be described in separate zones. A zone can only have one
control profile which is active for any given hour, so all rooms which
are included in a zone will be controlled the same way.

Example A
Fifteen classrooms are heated/cooled from 8am to 5pm and are set
back from 5pm to $am. All of these rooms may be included in a single
zone.

Example B
A hospital office area is occupied from 7am to 7pm and is set back
from 7pm to 7am. The hospital patient rooms are never set back. The
offices and the patient rooms should be placed in at least two different
zones to model the setback periods properly.

3) Zoning by Solar Gains. Rooms which have greatly differing solar gains
should not be included in the same zone, because the effects of the
solar gains will be diluted throughout the entire zone.

Perimeter zones with windows should be assigned at least one zone for
each direction of the compass. Otherwise multiple zone system
controls are difficult to model accurately, and solar overheating of one
zone is likely to be misrepresented in the simulation as free heat in
another area of the building.

For further simplification (and somewhat less accuracy) the perimeter

zones can combine north and east sides, and south and west sides.
Zones with unglazed exterior surfaces or extensive heat loss through
interior walls can also be lumped in a single zone if they are otherwise
similar.
 Example A
A north-facing zone should not be combined with a south-facing zone.

Example B
A core of interior offices should not be combined with an atrium
unless a single thermostat controls the HVAC for both.

4) Zoning by Perimeter or Interior Location. For large buildings, careful

attention should be paid to separation of interior and perimeter spaces
when breaking the building down into zones. Ignore the location of
interior walls in this case. Interior spaces with no exterior walls, roofs,
or floors are cooled year round. Perimeter areas stretching from
exterior walls with windows inward to about 12 to 15 feet from the wall
are often heated. If these perimeter spaces and interior spaces are
lumped into common zones, the simulation will dramatically
underestimate winter heating and non-economizer cooling
consumption.

5) Zoning by Fan System Type. Rooms which are served by different

types of fan systems should not be included in the same zone, since a
single zone may be served by only one type of fan system.

Example A
A room which is served by a fan coil system should not be combined
with a room which is sewed by a variable air volume (VAV) system.

Special Considerations:

There will often be special conditions in a building which require additional

zoning in order to model the building conditions accurately. The modeler
must always analyze the thermal aspects of the building to determine
whether the zoning will be sufficient. Here are a few things to watch for.

Shading
When zoning is based on solar gains, remember to consider shading
effects.

Large Open Spaces

Models generally assume that all zones are well-mixed. If there is
reason to believe that different parts of a large space may experience
different conditions, then the space may be divided into more than one
zone. Remember, though, to provide some means for heat transfer
between these zones. This may be accomplished with mixing or
interzone surfaces.

Unconditioned Spaces
When describing unconditioned spaces, such as attics and crawlspaces,
consider the boundaries carefully. For example, if the attic is separated
into more than one section, so that the sections will not always be at
the same temperature, then it may be necessary to use more than one
zone to describe the attic.
It is important when combining spaces into a single zone that all of the
elements of each space are included. Do not move, for example, the lighting
o r a window from one zone into another without moving also the respective
floor area, equipment and any associated exterior surfaces.

4.1.2 Assignment of Zones to Systems

In computer simulations, once the net heating and cooling loads have been
calculated for a zone (usually on an hourly basis), the loads are combined
with any loads due to ventilation or change in thermostat setpoint and
passed to the simulation of the air distribution system, which we refer to
simply as the "system".

Zoning of Single Zone Systems:

Zone assignment is not particularly critical for single zolre systems, except
that the following two conditions should be met.

1) If some of the building is heated or cooled by a different system type,

different fuel, or at a significantly different efficiency level than another
part of the building, that distinction needs to be made by assigning the
respective zones to separate systems.

2) If multiple zones are assigned in a computer simulation to a single-

zone air distribution system, some simulations will assume that each
zone is served by a separate air distribution system, while other
simulations will add reheat and serve all of the zones with a single
system. Since the latter strategy is far less efficient, the modeler needs
t o make sure that the strategy taken reflects the actual design o r
building.

In existing buildings, single zone systems as installed are sometimes serving

incompatible spaces, such as interior and perimeter areas, without much
success. Discomfort caused by improper zoning is difficult to model. The
simulation is likely to overestimate consumption if proper zoning is
modeled.

Conversely, extensive use of baseboards to offset inadequate air distribution

system zoning in existing buildings can introduce an inefficiency that isn't
reflected in a simulation of a single zone system serving the same area. In
such cases, greater simulation accuracy may be achieved by modeling the
single zone system as a multiple-zone system with reheat o r baseboard heat.

Zoning of Multiple-Zone Systems:

For multiple-zone systems, it can be important to assign zones to air

distribution systems accurately. In general, it takes more energy for an air
distribution system to serve zones with diverse simultaneous (i.e. at any
given time) heating and cooling needs than it takes to serve zones with
similar loading patterns. Two extreme examples will illustrate the case.

Fans Schedules. If a building contains 4 zones that are occupied 12

 hours a day and one zone that is occupied 24 hours a day, the energy
consumption for heating and fans is going to be higher if the 24 hour
zone is served by the same air distribution system as the other zones.

Cold (and hot) duct central supply temperatures. If a building contains

some zones that require 52 F supply air for cooling year round and
other zones that are in heating during cold weather, energy
consumption for heating (and possibly cooling) is likely to be higher
if all zones are on the same air distribution system.

If a computer room or a 24 hour part of the building is served by its own air
supply system, adding those zones to a multiple-zone system serving the rest
of the building will result in a significant overestimate of energy
consumption.

4.1.3 Assignment of Systems to Plants

Where one or more air distribution systems is served by one o r several

boilers and chillers, the heating and cooling loads on the respective air
distribution systems are passed to a separate "plant" section of the program.
Unitary systems, which include the supply fan and all heating and cooling
equipment within a single enclosure, are modeled without a "plant"
simulation.

Plant assignments (the assignment of one or more air distribution systems

to a specific group of boilers and chillers) can be important if different fuels
o r efficiency leveb characterize different boilers and chillers in the building.
Plant assignments are also important because they affect where on the part
load curve the boiler or chiller is operating.

Building energy simulation software generalIy offer limited capability for

assigning different air distribution systems within the same simulation t o
different boilers and chillers. Inaccuracies that might result from this
limitation should be considered qualitatively by the modeler in presenting
the simulation results.

Sometimes software limitations may make it advisable for the modeler to

simulate a particular building in two or three separate parts (i.e.
independent models). This, of course, adds to the complexity of the
modeler's work, and requires careful accounting to not lose track of any of
the parts. Software limitations that could make this necessary might include:

inability to assign different air distribution systems to different

plant equipment,

insufficient allowance for required number of zones,

inability t o accept multiple occupancy or operating schedules for

different building sections.
4.2 ENVELOPE LOADS

4.2.1 Infiltration

Infiltration can be one of the most significant energy drivers in a building

simulation. This is also the simulation parameter about which the modeler
is likely to have the poorest information. For these reasons, modelers often
view infiltration with mixed feelings. On the one hand, it is frustrating to
have to guess at such an influential parameter. On the other hand, because
the truth is safely hidden, the modeler often feels free to adjust ("tweak")
this parameter within the wide bounds of reasonableness in order to match
measured data.

The ASHRAE Standard 90.1 recommends procedures for model input of

infiltration (reference 8):

"Infiltration shall impact only perimeter zones. When the HVAC

system is ON, no infiltration shall be assumed to occur. When the
HVAC system is OFF, the infiltration rate for buildings with or
without operable windows shall be assumed to be 0.038 C F M / ~of~ ~
gross exterior wall."

An exception is provided for hotels and motels. For these buildings, the
above infiltration rate is to be assumed for all hours.

Though we have little data to support this contention, we believe that the
above guidance is somewhat simplistic. The assumption of zero infiltration
during hours of HVAC system operation rests on the assumption that
system operation will result in building pressurization. However, this
presumes a well-designed, well-balanced, and properly operated air
distribution system. It also presumes the absence of other infiltration-
related effects such as tall building stack-effect, a high frequency of occupant
or customer entry and egress, normally-open loading docks, and so forth.
The presence of any of these effects should prompt the modeler to
reconsider her infiltration input.

Similarly, the mandated value for infiltration during hours of no HVAC

operation is restrictive and simplistic. Though the value may be generally
appropriate, the standard gives no guidance as to when this value applies.
Is it wind-dependent or independent? Is it temperature dependent? We
recommend that 0.038 C F M / is~ a~ reasonable
~ beginning assumption, but
that the modeler consider the characteristics of the building being modeled
to determine its ultimate suitability. We further recommend, if the software
package being used has the capability, that the modeler split the volume
between wind-dependent and wind-independent infiltration. We note,
however, that DOE2 does a relatively poor job of simulating wind-dependent
infiltration. DOE2 modelers often rightly choose to model infiltration as
wholIy wind-independent. Turning our attention to baseline building input,
we recommend that this model use the same infiltration input as the design
model unless a specified ECM directly affects infiltration.
4.2.2 Window Unit U-values and Shading Coefficients

Modelers commonly assume that the winter U-value of a single-glazed

window is 1.1, and that the U-value of a double-glazed window is 0.49.
However, these values do not consider the effect of the window frame. An
aluminum frame with no insulating "thermal break" can degrade these U-
values to 1.23 and 0.78, respectively (reference 6). Given the high U-values
of glazing relative to walls and roofs, this bias can have a disproportionately
large effect on the building model. The opaque frame can also have a
significant effect on the overall window unit shading coefficient.

Typically, the design-phase modeler will not have accurate information about
the window units to be installed. We recommend that, unless the design
specifically includes an ECM relating to glazing and frame type, the modeler
assume a double-glazed window with a non-thermal-breakaluminum frame.
If the modeler does have knowledge of the specific window types to be
installed, then we recommend that she use the manufacturer's rated window
-
unit U-values (and shading coefficients) as a first preference. If these are
not available, or are suspect, then the modeler should either refer to the
ASHRAE Fundamentals Handbook (1989), Chapter 27, Table 13, or use the
Lawrence Berkeley Laboratory computer program WINDOW 3.1. The
modeler should also adjust shading coefficients to account for opaque
portions.

DOE2 modelers should note that the input for glazing heat transfer is
GLASS-CONDUCTANCE, not U-VALUE. The former parameter adjusts
the U-value for outside film coefficient and wind speed. The DOE2
reference manual provides a formula for making this conversion.

4.2.3 Wall U-values

One cannot assume the U-value of a wall (or roof or floor section) to be the
same as the U-value of the insulation it contains. In some cases the U-value
of the wall-section may be less than that of the insulation due to other
layers, air-spaces, etc. And in some cases the U-value may be substantially
higher than that of the insulation due to wall construction. Metal stud
construction is a prime example of the latter. (See Reference 8, Appendix
E, for a thorough technical treatment of this issue.) In general, though, stud
spacing, thickness, and type all influence the wall U-value.

The State of Oregon Energy Code Compliance Manual (reference 7)

presents a useful set of tables for dealing with this issue. We recommend
that the modeler uses these tables where appropriate. These tables can be
found in appendix I.

Sometimes insulation is installed in such a manner that it's theoretical U-

value is not realized. This can occur when batt insulation is compressed in
a wall space or installed with major gaps. The as-built modeler should be
aware of such deficiencies and adjust nominal U-values to reflect the as-
installed conditions.
Architectural features of a building sometimes make it unusually difficult to
calculate an appropriate U-value. When the modeler encounters such
features, she should first evaluate whether they are likely t o have a
significant effect on building energy behavior. If the features affect only a
small portion of the building, then the modeler may decide that quick-and-
dirty calculations may be adequate.

If the features may have a significant effect, the modeler should take more
care in these calculations. Sometimes, this may involve adding zones to the
model--for instance, to model an unheated area above the top floor ceiling,
o r an attic. Sometimes, it could involve modeling the feature with a high
and a low in a range of possible U-values. This would bracket the effect of
assumptions about this feature on the model.

Recent work at the National Renewable Energy Laboratory suggests that

effective U-values may actually be 60% lower than those calculated using the
standard ASHRAE methods. One project measured a small building's
overall U-value on the basis of heating consumption and outside air
temperatures.14 Until the SERI methodology and results have been
closely examined and duplicated, it would be premature to revise the
recommendations given above. It will be interesting however, to follow the
outcome of further investigation.

4.2.4 Shading and Solar Gain

When a building has significant areas of glazing, solar gain through that
glazing can be a major driver of building energy performance. Accurate
solar gain calculations must take into account the building latitude, the time
of year, the time of day, the orientation of the building surface, the physical
characteristics of the surface, and shading. Hourly computer simulations are
generally equipped to accurately take care of all of these factors except
shading, without much inconvenience to the user.

Exterior shading is usually uncontrolled by the occupants, and exerts a

seasonal effect on building energy behavior. Ground-based shading such as
trees may block the low winter sun more than the high summer sun. O r
they may loose their leaves in winter--leading to the opposite effect.
Ignoring this may lead to gross miscalculation of heating energy, cooling
energy, or both.

If a building is to be located in an urban area amidst multiple high-rise

buildings, then those buildings will offer exterior shading. Most software
programs have some means of entering such shading. For new construction,
if the design drawings include landscaping plans, then the modeler should
input major shading features from these plans. If the building involves
architecturaI features such as fins or overhangs that will shade the windows,
these features must be input.

l4 This project was done cooperatively with Bonneville. It applied the

STEM techniques to an Energy Edge small office. The results of this study
are reported in reference 17.
Another commonly overlooked type of exterior shading is the building itself.
Often one wall of a building may shade other walls. Again, this can have a
significant effect on the simulated performance.

Occupant-controlled interior shading can be a more difficult, but equally

important, energy driver. Usually the type and usage of interior shading
devices is not known until well after the design-phase model is required.
However, it is reasonable for the modeler to assume that some form of
interior shading will be instaIled if there is a significant area of untinted
glazing--especially on the south and west sides of a building. Lacking
guidance from the owner and architect, we typically assume shading by
venetian blinds. As for scheduling of usage, we typically assume that west-
side blinds will be shut from 3pm to 8am the following morning, and that
south-side blinds will be shut from l l a m to 3pm. Sometimes we adjust
these schedules seasonally.

ASHRAE Standard 90.1 (reference 8) requires that glazing be assumed to

be internally shaded by medium-weight draperies closed one-half time. It
further requires that the shading be calculated as effective over one-half the
glazed area in each zone. Though we see this requirement as more arbitrary
than our normal assumptions, we recommend that the modeler use
whichever she feels most comfortable with.

It's important to remember that shade will tend to increase heating

consumption in the winter and reduce cooling consumption in the summer.
Similarly, solar gain that occurs in the early morning is likely to reduce
heating consumption, while solar gain in the afternoon will more often
increase cooling loads. A shading coefficient, which applies equally year
round, is therefore not going to properly represent operation of curtains,
deciduous trees, or external structures that result in varying degrees of
shading.

Window setback and overhangs are generally easy to input into simulations.
Movable curtains can be modelled in DOE2, but cannot be modelled in
many programs. Shading by adjacent buildings and other parts of the same
building tends to be time-consuming, even when possible. It's advisable for
that reason to model windows that are in almost permanent shade as either
north-facing or with a very low shading coefficient.

Keep in mind that for most simulations, each zone floats in space, with no
designated geometrical shape or relationship to the other zones. The
simulation has no idea if one part of the building shades another unless that
information is input explicitly.

4.2.5 Daylighting

Installation of additional glazing and controls to automatically reduce

electric lighting during sunlit periods is referred to as daylighting. The
HVAC effects of glazing are easy to model in simulations; the reduction of
electric lighting in response to varying levels of natural light is much more
difficult to simulate. Even lighting design programs that model only
daylighting make some serious simplifying assumptions, such as zero direct
beam (versus diffuse) solar radiation.

In most building energy simulations, the only way to model the lighting
effects of daylighting is to input a reduction in the peak lighting load, or a
change in the lighting schedule.

Unfortunately, this method tends to underestimate the energy savings from

daylighting controls. Natural light is usually brightest during warm weather,
so daylighting reduces cooling energy requirements while having little effect
on heating. Simulations that can only model daylighting controls by a
change in the average lighting levels year round wiIl tend to overestimate
both heating and cooling.

For this reason, the modeler is advised to model a daylighting ECM with
daylighting algorithms if they are available in the simulation being used for
overall analysis.

DOE2.1 (versions C and later) does offer a daylighting program. 4 1

sunlight entering the space is assumed to be spIit in half, with half being
evenly distributed over the ceiling and the upper half of the walls, and half
being evenly distributed over the floor and the lower half of the wall area.
The modeler inputs the minimum foot-candle requirement, and the location
it is required. Specific architectural daylighting elements, such as light
shelves, cannot be directly simulated by the program, so their attenuation of
light entering the space may be evaluated by the modeler and included in the
shading coefficient.15

4.2.6 Thermal Mass

The well known equation for heat loss (or gain),

Q = U-value * Area * (Ti, - To,,)

fails to account for solar gain or thermal lag. While such an equation is
useful for sizing equipment, using worst case (design) conditions, it is less
accurate for calculations of annual energy consumption. The strength of
computer programs as tools to estimate annual energy consumption lies
largely in their ability to account for solar gain and thermal lag, two factors
which are not easy to model in hand calculations.

When the outside temperature and/or inside temperature is varying with

time, the above equation is no longer correct for any given moment, because
the heat loss or gain by the space is affected not only by current
temperatures, but also by the history of previous temperatures. Given
dynamic, (non-steady state) outside or inside temperatures, heat is stored in
the walls and roofs, creating a thermal lag. The rate at which a change of

l5 DOE2 function commands can be used to input the effects of

various architectural elements. In buildings where daylighting control is a
significant ECM, an architectural scale model can be created to generate
data for the function commands.
outside temperature is registered inside the building is a function of the
insulation value, and even more importantly, the weight of the wall o r roof
through which the heat is passing.

Thermal mass affects the timing of cooling loads as well as energy storage
behavior. Thus inputs describing mass can be important in estimating
cooling load shapes and coincident peak demand in buildings of heavy
construction.

Heavy buildings are likely to show higher savings from night flushing than
light buildings. Heat absorbed by heavy walls during hot days is released to
the atmosphere and to the interior of the building at night. Cool night air
can be used to cool the building at night, displacing use of mechanical
cooling that would otherwise be needed the following day.

The amount of energy required for morning warm-up o r cool-down is

greater in heavy buildings than in light buildings (unless the building is
permitted to warm up or cool down after occupancy). Heat transfer
decreases as the difference between the inside temperature and the outside
temperature decreases. In a light building, the inside temperature
approaches the outside temperature (during setback, setup, or night shut off)
faster than in a heavy building with the same configuration and envelope U-
values.

Accurate building weight simulation improves the accuracy of night setback

and night flushing savings estimates, as well as the effects of solar gain o n
HVAC loads.

DOE2 provides two methods for dealing with transient heat gains in a
space--precalculated weighting factors and custom weighting factors. The
program uses the precalculated factors as a default. So, once again, if the
modeler declines to select a method, she has, by default selected one.
Quoting from the DOE2 Reference Manual, p.TII.143 (reference 1I),

"To aid the user in deciding which of the above methods to use for
HVAC calculations, the following can be stated. The Precalculated
Weighting Factor method requires the least computer time and
produces the least accurate results. The Custom Weightinp Factor
method is more accurate than the Precalculated Weighting Factor
method, but requires more user-input effort and slightly more
computer time. In the following cases use of the Custom Wei~hting
Factors is suggested:

Buildings with thermostat set-back and/or set-up

All passive solar buildings
Masonry buildings
Heavy construction buildings
Any building in which it is necessary to define the
distribution of the solar radiation within the building
Buildings located in sunny locations with large amounts of
solar energy entering the spaces."
We have found that the use of custom weighting factors instead of
precalculated weighting factors has made a significant difference in heating
and cooling end-uses in buildings that meet some of the above descriptions.
Other modelers have also noted a significant difference. The modeler
should follow the DOE2 reference manual recommendations concerning
when to use custom weighting factors.

In our work, we have found that switching from precalculated t o custom

weighting factors generally makes a more significant difference than
changing the values of the keywords in the custom weighting input. Though
we have not researched the question exhaustively, it appears that the most
critical custom weighting input is the area and construction of interior walls
(including floors and ceilings). As long as the modeler takes some care in
her input of actual interior wall characteristics (area, construction, U-value,
and location), her input for the remaining custom weighting keywords
(FURNITURE-TYPE, FURN-FRACTION, AND FURN-WEIGHT) will
have relatively little effect. She may find it convenient to enter an adiabatic
wall rather than standard interior walls to adequately simulate the mass
without affecting the inter-zonal heat transfer. She must also use layer-type
input for exterior walls as opposed to the "quick" U-value input.

To use custom weighting factors, the modeler must input a FLOOR-

WEIGHT of zero under the appropriate zone SPACE-CONDITIONS
command. Reasonable input for the FURNITURE-TYPE keyword is
HEAVY for a predominance of file cabinets, bookshelves, etc., and LIGHT
for a predominance of chairs, softwood furniture, etc. Reasonable input for
the FURN-FRACTION keyword is 0.2 to 0.3 for a typical office layout.
Reasonable input for the FURN-WEIGHT keyword is 2 to 5 (lb/ft2) for the
typical office. Extensive file storage, heavy equipment pallets, library stacks,
and so forth would prompt a higher input value.

DOE2 also provides a keyword for description of the portion of the solar
radiation coming through the glazings in the space that is absorbed by the
particular interior wall, floor, or ceiling under which the keyword is input.
Generally, the DOE2 defaults for this keyword, SOLAR-FRACTION, are
adequate. These defaults are: 60% of the incoming solar radiation is
absorbed by the floor; the remaining 40% is distributed to the other named
surfaces in the space, according to their surface areas. The modeler should
override these defaults only when they are a clear misrepresentation of the
actual geometry of the space.

4.2.7 Unconditioned Spaces

Heat transferred through surfaces (wall, floor or ceiling) between

conditioned and unconditioned spaces can be modeled in either of two ways.

Method 1: The unconditioned space is input as a zone in the program, and

the surface between the two zones is input as an interior surface.

Method 2: The surface between the two zones is input as an exterior

surface, with provisions made to see that solar gain and the additional
thermal lag of the adjacent space are roughly accounted for.

The advantage of the second method is speed: it doesn't require

descriptions of the exterior surfaces or conditions of the unconditioned
space.

The disadvantage of the second method is that it is difficult to model the

thermal lag and solar gain characteristics of an unconditioned space as a
simple surface. If, for example, an office is adjacent to an unheated
warehouse, the effects of outside temperature changes will be transmitted
through the warehouse to the office at a slower pace than they would be
through a light exterior wall. There probably is some sort of exterior wall
which the user could input which would approximate the same insulation
value, thermal lag and solar gain characteristics as an attached warehouse,
but it is not obvious what that surface would be like. The best that one can
do when forced to use such a procedure is to make the interior surface
which is modeled as an exterior surface heavy enough to roughly cover the
effects of the thermal mass of the unconditioned space.

If there is a liberal flow of outside air through the unconditioned space, (as
in most commercial garages), it is advisable to model the interior surface
as an exterior surface, because the temperature of the conditioned space will
be similar to the outside temperature. The surfaces should be described as
north walls, o r horizontal surfaces facing down (with a 180 degree tilt in
DOE2), to minimize solar gain.

Do not model mechanical rooms, or the walls of such rooms. They are
heated for free by the waste heat from HVAC equipment.

In small buildings, heat loss and heat gain between conditioned and
unconditioned spaces can have a significant impact on the overall building
energy consumption. In those cases, it may be worth modeling
unconditioned spaces as zones. For large buildings, where the overall effect
of the unconditioned space is small, the added detail work involved in
modeling unconditioned spaces as separate zones is usually not worth the
effort.

4.2.8 Interior Walls

It is often unnecessary to model an interior wall if the thermostat schedules

for the zones o n either side of the wall have identical setpoints and hourly
profiles. Heat transfer does not occur unless there is a temperature
difference.

However, there are two important exceptions to this recommendation. First,

the modeler must take care that an interior zone with high internal heat
gains during periods of HVAC non-operation has some means of
transferring this heat to adjoining perimeter zones. Otherwise, these interior
zones may see unoccupied period temperatures float absurdly high. This in
turn can lead to improper morning pick-up loads and generally
overestimated occupied period cooling loads.
Second, when thermal mass is important to a building's energy behavior, the
modeler should elect to calculate custom weighting factors. She can
simulate the mass of an interior zone by entering an adiabatic wall with
surface area equal to that of the actual interior walls.

If there is a significant temperature difference across an interior wall for a

significant period of time, the wall should be described in the model. There
are typically two ways to describe an interior wall in a building simulation.
Either the modeler is asked to identify the zone on the other side of the wall
o r to give the average temperature of the air on the other side of the wall.

Be sure not to input an interior wall under both adjacent zones. That would
result in duplicating the heat transfer.

In a large building, very often the heat transfer between the conditioned
spaces and an adjoining massive unconditioned space (such as a parking
garage) will be insignificant relative to the other heat balances in the
building. If the modeler decides not to model such a massive unconditioned
space as a separate zone(s), the simplest approximation is to input under the
adjacent zone an interior surface, and give the temperature on the other side
of the wall as the average annual temperature.

Summarizing our recommendations, model interior walls only if:

(1) the thermostat schedules and setpoints for the zones on either side
of the wall are not identical, or

(2) the zone has significant internal gains during periods when the
HVAC system is not operating, or

(3) an adiabatic wall is needed to properly simulate thermal mass,

4.2.9 Above-Ceiling Spaces

Most commercial buildings have spaces several feet high located between the
ceiling of each conditioned space and the floor (or roof) above it. Such a
space is unique in several respects. It gains heat from recessed fluorescent
fixtures, loses and gains heat through the exterior plenum walls, and loses
and gains heat from heating and cooling supply air ducts. In some buildings,
HVAC return air travels unducted through these spaces. In these cases, the
spaces are termed ceiling plenums. The temperature of the ceiling plenums
is not thermostatically controlled.

Heat losses and gains which occur in a ceiling air plenum affect the amount
of energy consumed for HVAC because they affect how much heat is lost or
gained by the return air between the conditioned space and the supply fan.
The above-ceiling space temperature also affects heat loss o r gain through
the ceilings of the space below and any floor above, and the amount of heat
loss o r gain from supply air ducts.

In most cases the walls of the above-ceiling space can be included in the
zone directly below it, The walls, and heat from lights would be included in
the inputs for that zone. The greatest inaccuracy of such an approach occurs
for true return air plenums, and only when economizer cooling is operating
or a high minimum outdoor air percentage is in effect. During these
conditions, the heat of lights and heat gain through the plenum walls would
still be modeled as entering the space rather than being dumped outside
before entering the supply fan and coils. Energy consumption for cooling
tends to be low during periods of economizer cooling, so this inaccuracy is
not large. However, the inaccuracy may be significant in the case of a high
minimum outdoor air ercentage since this condition occurs during all hours
of HVAC operation. 11
4.2.10 Heat Loss to Ground

Heat loss through surfaces in contact with the ground is only a very small
part of the overall heat loss for a multiple story commercial building. For
such a building, accuracy is not very important, though care should be taken
t o avoid gross overestimates. Tn single-story commercial buildings, heat loss
to ground can be a significant portion of the building's total heat load.

The temperature of the ground is more stable than temperature of the air,
particularly as depth increases. But the temperature of the soil is
significantly affected by the presence of the building. One cannot simply say
that the ground temperature is usually around 50 O F so the underground wall
or floor heat loss can be modeled as,

Q = U x A x (Ti, - 50 degrees)

because in fact, during the heating season, the soil under a heated building
and around underground walls is warmer than the normal ground
temperature. Use of this type of calculation, though, is common in the
more simple software packages, and tends to overestimate heat loss to
ground.

According to the ASHRAE Handbook of Fundamentals (1989, page 25.6),

heat loss through underground floors and walls is best understood as flowing
from the building interior, through the ground to the outside air. The
ground acts as thermal mass and insulation. The longer the path through
the soil from the building to the outside air, the lower the heat loss. Heat
loss therefore decreases with depth and distance from the building's
perimeter.

If a simulation uses the full floor area to calculate heat loss from a floor in
contact with the soil, it is best to either input the linear footage of the
floor's perimeter where the simulation asks for area, or to input an
artificially low U-value. The low U-value is the better approach if the floor

l6 If the DOE2 modeler feels that for a specific building it is

important to account for the heat exchange between the return air stream
and the plenum walls, she can model the plenum as ZONE-
TYPE=PLENUM and, in the SYSTEM command for the adjacent
conditioned zone, model the RETURN-AIR-PATH=PLENUM-ZONES.
is likely to be important to the building thermal mass behavior or if the
floor area input is also used for calculating other information such as
infiltration or zone equipment loads. The DOE2 modeler should use the
actual floor construction for calculation of the custom weighting factor
parameters. The artificial U-value can be calculated as,
U-EFF = U,,,,,, x (length of perimeterJfloor area)

4.2.1 1 Weather

It is our experience from Energy Edge and other modeling projects that
minor variations in weather input data usually have a relatively minor effect
on simulation results. Edge put a lot of time and money into monitoring
actual weather at the building sites. However, in our modeling we have
found that it rarely made a significant difference in annual end-use energy
consumption whether we used the site-gathered weather file or the QpicaI
Meteorological Year (TMY) weather file for the closest available site.
Qpically we would find about a 5% or less annual difference between
simulations using the two types of weather files. Monthly differences were,
of course, more noticeable. But for design-phase modeling, monthly energy
consumption is usually not of prime importance.17

However, we have found an exception to the statement that minor local

variations in weather usually has a minor effect on simulation results. In
one Energy Edge simulation we found that an average local 5 F degree
increase in dry bulb temperature from the TMY data resulted in a 70% to
80% increase in HVAC energy consumption. We believe that this degree
of effect can occur when the thermal balance temperature of a building is
close to the average annual dry bulb temperature. The effect may also be
tied to the volume of outdoor air introduced through infiltration and
ventilation.

We recommend that the design-phase modeler use the best readily available
weather data for the location closest to the project site. But we don't
generally recommend that the modeler (or the utility) spend additional time
to gather and process extensive site-specific weather data. A compromise
position that may have merit is to monitor or otherwise acquire local
outdoor dry bulb temperature only, and to integrate this series of values into
the TMY weather file.

l7 The Corson report (reference 3) also generally backs up our

experience. In the sensitivity study phase of that work, Corson found
that, depending on the software used, switching the weather file between
Eugene, Oregon (a site with relatively moderate summers and winters),
and Richland, Washington (a site with relatively extreme summers and
winters), usually resulted in less than a 10% difference in the estimates
for annual energy use. (However, most of this difference is in the heating
end-use.)
4.3 INTERNAL LOADS

The modeler must estimate end-uses such as equipment (also referred t o as

"plug-loads" and "receptaclesn) and lighting both to account for the power
consumed by the end-use and for the cooling (or in the case of refrigeration,
heating) load imposed on the space by the end-use. In general, there are
three types of input parameters associated with such end-uses. These are
peak power consumed, a profile (schedule) of fraction of peak power used
in any given hour, and a fraction of hourly energy consumed that results in
a load to the space in which the equipment is located. Following is a
generic discussion"of these three parameters. Section 4.3.1 begins a more
detailed discussion of these parameters as they apply to plug-loads,
computers, lighting, refrigeration, and cooking.

1) Peak Values

Rated, o r peak, consumption rarely occurs on most pieces of equipment.

The modeler should only give the rated equipment capacities if the
simulation offers a load factor (averagefpeaklopd fraction), or calculates the
equipment load input hourly using an operating schedule profile that
permits fractional amounts. Note that even the peak consumption in a
weekly equipment schedule is generally not equal to the rated equipment
capacity, because the weekly peak used in computer simulations represents
only the maximum hourly average that occurs in a typical week. For copiers
and printers, the equipment peaks may occur for only a half a minute at a
time and may far exceed the average during any hour.

2) Schedules

We recommend that the modeler use Table 13-3, Building Schedule

Percentage Multipliers, in ASHRAE Standard 90.1-1989 (reference 8) as the
first source for building schedules. This is reproduced in appendix I11 for
the modeler's convenience. However, we further recommend that these
multipliers be modified in certain circumstances--as discussed in sections
4.3.1 through 4.3.3. The modeler must use good judgement in application
of recommended multipliers in the table. For instance, the warehouse
schedules are not appropriate in general to refrigerated warehouses.
Similarly, the retail schedules are not appropriate to refrigeration in g r o c e j
stores.

If believable information concerning schedules is available from the owner

or design team, the modeler should consider this information. For instance,
the owner of a grocery store may intend 24 hour operation, but the Table
13-3 schedule states 12 hour operation. In this case it is appropriate to
input the owner's intended schedules.

For new construction, the modeler should be aware that, even if the input
makes reasonable assumptions about load schedules, the simulation results
may be far from actual building energy consumption when the building is
only partially occupied. Partial occupancy is the normal state for buildings
that are not owner-occupied during their first months or years. So, unless
the modeler specifically attempts to model this start-up period, she should
not expect the simulation estimate to match billing data during this period.

3) Percent of Heat to Space

The objects located in a building may affect the energy consumption of the
building in two ways: they may consume energy directly, and they may affect
the amount of energy consumed for HVAC. People and hot food are
examples of objects which affect HVAC consumption without consuming
energy directly; equipment generally affects both. This section addresses
internal gains by such equipment, including office equipment, computers,
refrigeration, and task (not ceiling) lighting.

If a piece of equipment is completely enclosed by the building and has no

exhaust stack or vent, the equipment heats the building at the same rate it
consumes energy to run, even if the equipment (such as a typewriter) has a
primary purpose other than heating the building. Energy is consemed, and
any machine ultimately dissipates mechanical energy in the form of heat as
a result of friction.

If equipment has an exhaust hood or vent, some of the energy consumed

directly by the equipment escapes up the stack, and the amount of heat
contributed to the space is less than the amount of energy consumed directly
by the equipment.

The modeIer needs to find out how the simulation addresses internal gains.
Some programs assume that the heat to space equals the energy consumed
by the equipment, other programs ask for the percent heat to space. Some
programs ask for two inputs: one to cover electric consumption by
equipment, the other to give the amount of heat to space. Most programs
offer an input for equipment that has no effect on the HVAC. If a program
assumes the heat to space for internal equipment equals the electricity or
other fuel consumed, the portion of the equipment load that is vented
directly by exhaust hoods can be entered under the program's option for
modeling external equipment. Such artificial inputs are difficult to track and
should therefore be used only for important loads.

Chapter 26 of the ASHRAE Handbook of Fundamentals, 1989, is a good

source of information concerning the amount of energy consumed by various
sorts of equipment, and their effect on HVAC requirements.

4.3.1 General Plug Loads

Peak, or Connected, Load:

As the connected load for lighting drops, the connected load for office
equipment and computer is rising dramatically. Old rules of thumb, such as
.5 watts per square foot for plug in loads are no longer even approximately
correct. Information about current use patterns are available from several
end-use studies.

These studies indicate that, given no actual building data, an operating

power density of 1.0 to 1.5 w/ft2 for omce equipment and personal
computers is appropriate.

One study (ref. 4) suggests that the actual plug usage in six large office
buildings was over five times hi her than predicted. Modelers' estimates
K
ranged from 0.64 to 1.73 Kwh/ft -yr, but sub-metered data indicate 7.2 to
10.7 ~ w h / f t ~ - ~Most
r. of this discrepancy apparently was due to the
assumption for peak power density.

Schedules:

According to recent end-use monitoring studies, power consumption by

office equipment during occupied hours far exceeds what has generally been
assumed by energy analysts. Results of the ELCAP study suggest that, as
a general rule of thumb, equipment consumption during unoccupied hours
should be set at no less than 30% of the peak load.

Monitoring data from the Energy Edge Program, and ELCAP data
(reference 5) indicate that unoccupied equipment usage in some building
types lies in the range of 40 to 70% of occupied usage. ELCAP data gives
the folIowing ratios of unoccupied to occupied (peak) weekday equipment
usage: l8

BUILDING TYPE UNOCCUPIED/ PEAK W / F T ~

PEAK RATIO

Office 0.40 1.50

Retail 0.30 0.76

Grocery 0.67 1.04

Restaurant 0.36 2.13

Warehouse 0.50 0.35

We recommend that these values for unoccupied equipment fraction be used

during design phase modeling unless there is specific design information to
support a different assumption. The modeler may also wish to use the
ELCAP values for peak power density when she lacks more detailed design
information.

l8 The ELCAP "other loads" end-use does not include refrigeration

or food preparation equipment. Also note that ELCAP data represent an
average among a11 monitored buildings in a given building type. Thus the
data for groceries are an average of stores that may operate for anywhere
from 12 to 24 hours.
4.3.2 Computer Rooms

In office building simulations, personal computers must be distinguished

from large computers. Large computers are usually on 24 hours a day,
served by their own HVAC systems. On the other hand, personal computers
are often shut off during unoccupied hours, and are served by general
HVAC building systems that cycle supply fans at night.

If internal gains from mainframe computers are averaged into the general
equipment watts per square foot, the building heating loads are likely to be
underestimated, and the 24 hour operation of the computer HVAC will not
be well represented unless the entire building is operating on a 24 hour
schedule.

The modeler is advised to distinguish between large computers and small

computers when developing equipment Load inputs, and to create a separate
zone and air distribution system in the simulation for any significant
computer rooms.

As with any attention to accuracy, this precaution can be ignored if the load
in question is too small to warrant detailed treatment. Also, in new
construction, a computer room that has no interactions with the rest of the
building can be completely left out of the model if it is unaffected by any
ECMs.

The effects of large computers relative to general office equipment and

personal computers a n be assessed by examining the following ELCAP end-
use data.

Annual Equipment
Consumption, Power Density,
End-Use k wh ~ f t ~ - ~ r w/ft2

Large Offices ( >30,000ft2):

General Office Equipment 0.84 0.66
Personal Computers 1.07 0.64
Large Computers 1.98 0.23

Small Offices ( c 30,000ft2):

Genera1 Office Equipment 0.93 0.73
Personal Computers 0.46 0.27
Large Computers 1.58 0.18

In most other building types examined, consumption by large computers was

negligible.
4.3.3 Lighting

Schedules:

ELCAP monitoring studies show that lighting loads, like equipment loads,
are higher during unoccupied hours than has generally been assumed by
modelers. The ELCAP results (reference 5) indicate that the following
ratios of unoccupied to occupied (peak) weekday lighting usage should be
used unless building-specific data are available:

BUILDING TYPE UNOCCUPIED/ PEAK W / F T ~

PEAK RATIO

Office 0.29 1.56

Retail 0.10 1.44

Grocery 0.58 2.00

Restaurant 0.14 1.67

Warehouse 0.06 0.68

The modeler may also wish to use the ELCAP values for peak power density
when she lacks more detailed design information.

Heat to Space:

Building simulations track both the direct electric lighting consumption, and
the indirect effect of lighting on HVAC requirements. The direct electric
consumption is input either as watts per square foot, or watts per zone.

For surface mounted or suspended fixtures, the electricity which is used by

the lights ends up as heat in the conditioned space, increasing the cooling
load or decreasing the heating load accordingly. Recessed fixtures may loose
heat to an unconditioned attic in small buildings, or to an unconditioned air
plenum in large buildings. If the attic is well vented, or if the HVAC system
in the large building uses an air plenum return, part of the heat generated
by lighting may escape from the building without contributing t o HVAC
loads. For that reason, most computer programs ask for the fraction of the
heat from the lights that goes into the space. Other programs ask whether
or not the lighting is "vented".

The most common method of cooling fluorescent fixtures is to pass the

return air through them to an air plenum above the ceiling tiles. In some
cases the air is passed through slots on the side of the fixture; in other cases
the air is passed though the interior of the fixture to a ducted return. The
percent of heat from the lights which goes to the space is a strong function
of the fixture design, the air flow rate through each fixture, and whether or
not there is ducted return. With a VAV system, the air flow rate varies.
Some fixtures experience a greater air flow than others due to their location
relative to the fans.
For these reasons, the percent of heat from the lights which goes to the
space is a rough estimate for vented light fixtures. It is generally assumed
that somewhere between 35% (for some ducted return air fixtures) and 80%
(for plenum return) of the heat goes to the space from fixtures with vented
return. A 1988 study by S.J. Treado and J.W. Bean gives extensive estimates
of space heat gain for various configurations of fluorescent lighting fixtures
(reference 17). For incandescent light fixtures and fixtures vented to supply
air, the fraction of heat to the space is roughly one.

4.3.4 Refrigeration

Refrigeration is the largest energy end-use in grocery stores and refrigerated

warehouses. It can also be a significant end-use in other building types.
According to ELCAP data, central refrigeration equipment in grocery stores
consumes, on an average, 43 k ~ h / f t ~ - ~ r .

Most software packages either do not sirnulate commercial refrigera tion, o r

simulate it poorly. DOE2.1 (versions C and later) is an exception, offering
the capability to simulate many characteristics of commercial refrigeration
systems and to investigate many possible refrigeration ECMs.

Modeling Refrigeration with DOE2:

We used DOE2 to analyze a large grocery in the Energy Edge program.

The DOE2 estimate of the refrigeration end-use consumption was within 5%
of the estimate of the refrigeration system manufacturer, who used software
designed for and tested against their equipment. We know of no other
comprehensive energy simulation software that can claim this flexibility or
accuracy.

For this reason, we recommend that modelers use the DOE2 program for
rnodeIing any building that contains a significant amount of commercial
refrigeration equipment. The input for this end-use can be quite complex.
It typically requires a large amount of supporting hand calculations. The
modeler should carefully calculate case loads, auxiliary loads (anti-
condensate heaters, evaporator fans, and case lights), and zone loads.
Controls must also be carefully considered. Carelessness in any of these
inputs can invalidate not only the refrigeration simulation, but also the
simulation results for the entire building since refrigeration systems usually
interact strongly with the building HVAC.

We offer a few specific tips concerning DOE2 refrigeration input:

Anti-condensate heaters - Split the rated heater kW between the

refrigerated case and the space in which the case is located. Based on
manufacturer recommendation, we have directed 85% of the kW to the
case and 15% to the space. This accounts for the full heater wattage,
but attributes a more reasonable amount to a refrigeration load to the
case.

Case lights - Split between case and space. Based on manufacturer

recommendation, we have directed 80% of the kW to the case and 20%
to the space.

Condenser type - The default is water. If you don't want this, be sure
to enter air.

Heat recovery - Consider how heat recovery interacts with floating

head pressure control. Both control strategies are common practice in
large refrigeration systems. In some systems, the condensing
temperature may be elevated to permit heat recovery. In others, the
savings of floating head pressure control take precedence. DOE2
simulates only the former case. In either case, if there is no heating
load in the space, then the head pressure usually is allowed to float
down to about 10 degrees F above the ambient temperature (minimum,
about 60°F).

System definition - The modeler should try to obtain the

manufacturer's R/S sheet (refrigeration schedule) to aid her in defining
the system input. Information is also often available from case
manufacturers about electricity consumption and heat gain per linear
foot of case. Hand calculations can also be performed using the
compressor capacities, assumed load factors, and standard COP'S as a
function of case temperature.

Whenever possible, the types and linear footage of refrigeration display

cases, and capacity and location of compressors should be obtained
from audit data o r design drawings for use in the hand calculations
necessary to develop refrigeration inputs.

Sometimes though it is not possible to obtain actual design or as-built

information. Architectural Energy Corporation has prepared a draft
handbook for analysis of ECMs (reference 185. In this handbook they
have provided some very useful tables for analysis of commercial
refrigeration. Included are typical case design loads per lineal foot,
energy consumption by auxiliaries per lineal foot of case, and default
compressor EERs for different temperature cases.

Since DOE2 allows only three refrigeration systems per zone, it is

often necessary to combine the actual systems to meet the three system
limitation. Case (or walk-in) location and evaporator temperature
should be the main considerations for system combination.

Pull-down loads - Sometimes grocery walk-in coolers are stocked with

warm product. The refrigeration load related to bringing the
temperature of this product down to the desired temperature is called
thc "pull-down load". This load can represent a major portion of the
total refrigeration load in some extreme situations. Therefore, it is
important that the modeler try to determine whether, and to what
degree, such a situation exists. She may do this through audits of
existing buildings or through interview of the operators of new
buildings.

DOE2 has no capability for direct simulation of pull-down loads. If

 the modeler has determined that a specific building will have a
significant pull-down load, she should use hand calculations to
determine the load. She should then add the load to the appropriate
REFG-ZONE-LOAD input an offsetting SOURCE-BTU/HR in
the loads section.

Modeling Refrigeration with Other Software Packages:

The anaIyst should use DOE2 whenever simulating any building in which
commercial refrigeration can be expected to be a major energy end-use or
where refrigeration ECMs are considered. This would include, at a
minimum, all grocery stores and refrigerated warehouses. But sometimes the
analyst encounters buildings such as restaurants or hotels in which
refrigeration is a relatively minor end-use. In these cases it is permissible
to use another simulation program.

When using a simulation that doesn't model food refrigeration equipment

as such, the modeler has to manipulate other inputs to approximate direct
consumption by refrigeration equipment and its effects on HVAC. HVAC
effects can be the most difficult to represent.

Refrigeration equipment generates heat at the condenser, but it produces a

cooling effect next to the refrigeration case.

If the condenser is not attached to the body of the refrigerator (and the
condenser is located outside the zone where the refrigeration case is
located), the space heat contribution at the condenser will be greater than
the value of the electricity consumed by the equipment, and the space
cooling contribution at the body of the refrigerator will also be greater than
the value of the electricity consumed by the equipment. If the compressor
is located in an unconditioned area, heat from the compressor can be
ignored except when considering heat recovery options.

Self-contained refrigeration equipment is relatively simple to model because

the net effect on the space will be a heating effect equal to the electricity
consumed by the unit. Where most of the compressors and condensers are
located outside the conditioned space and heat recovery is performed
mechanically rather than via air flow through the compressor room, the net
effect of the refrigeration cases on the spaces they are in is a negative
internal gain.

When the analyst must resort to hand calculations, the direct energy
consumption by the compressor, and the effect on the HVAC energy
balances can be roughly calcuIated as follows:

1) Electricity consumed by the compressor, hourly =

(rated capacity in kW / COP) x load factor.

2) Heat removed from the space around the refrigeration case, hourly =

(rated capacity x load factor) - E,,

3) Heat rejected at the condenser, hourly =

rated capacity + electricity consumed by the compressor =

rated capacity x ( 1 + 1/COP ) x load factor.

where E , is the electricity consumed each hour by strip heaters and display
lights in the refrigeration cases served by the compressor; and the load factor
is the average actual load on the compressor divided by its rated capacity.

These load factors are useful for general reference, but may not be accurate
if refrigeration ECMs are being considered, or if an existing building has
already undergone extensive refrigeration remodel. The loads on
compressors are decreased by decreasing the store ambient temperature and
humidity, reducing the amount of energy consumed for case lights and
defrost, the existence of case doors or strip curtains, or the orientation of
the case (tub vs vertical shelves). Roating head pressure controls and
parallel piping of multiple compressors also dramatically lower compressor
loading.

If any of these features has been retrofitted into an existing building, the use
of the standard compressor load factors may overestimate consumption. For
that reason, on retrofit projects, it is helpful to discuss with the store
operator a rough history of the refrigeration systems, and use low load
factors if it seems appropriate.

According to a study done by Seton, Johnson and Ode11 for the Oregon
Department of Energy, a load factor of 0.8 is typical of compressors under
normal operation. A load factor of 0.6 is typical of compressors working on
a refrigeration case which was originally designed to be open, but has been
retrofitted with a cover to save energy.

The same study gives the following COPS as typical:

1.0, for ice cream cases (-20 to -15 degrees F)

1.2, for low temperature cases (-5 to 0 degrees F)
1.75, for medium temperature cases (30 to 45 degrees F)

However, if manufacturer's COP ratings are available for a specific

installation, these should be used.

Inputs to the simulation will depend on the type of inputs permitted by the
simulation. If the program accepts negative internal gains, the heat removed
from the space by refrigeration can be easily described once the above hand
calculation has been performed. Otherwise, a negative internal gain is most
easily modeled by creating a fake internal wall next to an air space of fixed
temperature, with a UA delta-T equal to the hourly heat loss to the
refrigeration cases.

Inputs for the direct electric consumption of food refrigeration equipment

also depend on the types of inputs offered by the program. Many
simulations assume that any internal equipment loads create an internal
HVAC heat load equal to the direct equipment consumption. Description
of equipment that actualIy causes a net cooling effect on the space needs t o
be performed carefully in such models. Once the cooling effect has been
described, as recommended in the above paragraph, the direct electric
consumption for refrigeration can be calculated by hand and added to the
simulation results, or can be input to the simulation as equipment outside
the building conditioned space.

If a simulation does not offer refrigeration models, ECM savings can be

sometimes approximated by hand, and the resulting impacts on equipment
consumption HVAC can be input to the simulation. As an example, savings
from reduction in case lighting or electric strip heaters can be calculated as:

Savings = (change in direct consumption by the lighting or electric

strip heaters) x (I + 11COP).

4.3.5 Cooking

Cooking equipment consumes a significant percentage of the total building

consumption in restaurants. According to ELCAP data (reference lo), food
preparation equipment in restaurants has an average power density of 9.6
watts per square foot (continuously operated, plus 1.89 watts per square foot
of intermittently operated) and consumes on the average 9.16 kwhiyr per
square foot. These figures cover only electric equipment. Gas cooking was
also in use in many of the buildings examined.

Note that the electric consumption by cooking equipment is generally less

than the rated capacity, because the electric consumption is regulated by the
cooking requirements at any given time. Note also that the cooling load
imposed by the cooking equipment is less than the electricity or other fuel
consumed directly if it is vented to an exhaust hood.

Page 26.10 of the 1989 ASHRAE Handbook of Fundamentals gives

recommended values for heat gain to space from various types of cooking
equipment, with and without an exhaust hood.

In dining areas, a significant amount of latent and sensible heat are given off
by the food served. ASHRAE recommends assigning 50 Btu/hr of internal
load (25% latent and 75% sensible) for each meal served (Fundamentals,
1989, page 26.9).
5. HVAC SYSTEM SIMULATION

5.1 SYSTEM SELECTION FOR SIMULATIONS

All energy simulation programs offer the modeler some latitude in HVAC
system selection. The modeler will be offered a menu including systems
such as variable air volume (VAV) with reheat, fan-powered VAV, single
zone packaged rooftop, multi-zone, dual-duct, heating and ventilating only,
two-pipe hydronic with fan coils, and so forth.

The familiar names of system types offered by the programs can lull the
modeler into thinking that there is a good match between the actual or
designed system and the simulated system. From a system simulation
perspective, an HVAC system is a combination of equipment type (along
with their operating characteristics, part load performance, etc.) and system
control. Most programs will default to specific equipment types and control
sequences once a system type has been selected. Sometimes these defaults
are representative of the actual design, but often they are not. And in some
programs, these defaults cannot be overridden.

The actual operating differences between one air distribution system type
and another often depend more on differences between the user-specified
control options than characteristics intrinsic to the distribution system types.
Such control options include economizer cooling, central supply air
temperature reset, fan duty cycling, and variable air volume. User-specified
control options are restricted for some air distribution system types, so the
user should take this into account when selecting a system.

Therefore, the modeler must ensure that the important characteristics of the
design or actual building are being captured mathematically by the program.
If they are not, the modeler should consider using a system type that may
not be the same as the actual designed system, but that more closely
represents its operation than the program default for that system. If the
program does not offer good options, the modeler should consider use of a
different program. Remember that the Corson study showed that choice of
system type was the single most important driver of energy consumption.

The modeler should carefully question the default simulation of system type.
What control options are available for each system type? Is reheat being
assigned automatically if multiple zones are assigned to a "singlett zone
system, or does the model assign a separate single zone system assigned to
each zone? Does the default dual duct system have the same number of
supply fans as the actual system? Is the default dual duct system supplying
a constant total flow to each zone or a variable total flow? Are the supply
fans continuously on, on during occupied hours, or cycling on at all times?
Is the reheat source electric or gas? Does the system type allow outside air
ventilation? And so forth, ad infinitum.

5.2 HOW AIR DISTRIBUTION FITS INTO THE OVERALL

SIMULATION

In a computerized building energy analysis, the amount of heating and

cooling required by a space is typically calculated by a section of the
program called "loadst1.The space loads account for envelope heat transfer,
solar gains, and internal gains.

Once the loads section of the program is completed, the results are sent to
a section of the program that models the HVAC air distribution systems.
In this program section the ventilation air load, the coil heating and cooling
loads, the zone reheat and mixing loads, the supply air flow rate, and fan
energy consumption are calculated. (Reheat and mixing loads refer to loads
which may occur as a result of the need to serve several zones at once,
either by reheating mechanically cooled air or by mixing heated and cooled
air. They are loads which would not occur if single zone systems served the
same space loads.) If the HVAC equipment is packaged (i.e. not served by
a central plant hot and chilled water system), then HVAC energy
consumption will also be calculated in the systems section of the program.

However, if the HVAC equipment is not packaged, then the "system loads"
are sent on to a section of the program that models the boilers, chillers,
pumps, cooling towers, and so forth. That section of the program includes
central plant equipment efficiencies and part load curves, and calculates the
total HVAC energy consumption.

5.3 CAPACITY, LOAD FACTORS AND PART LOAD EFFICIENCIES

5.3.1 User inputs

When entering capacity, power, or efficiency for equipment it is important

to know what the simulation is going to do with it. Is a fan kW going to be
applied to a part load curve, cycled on or off, or taken at face value for the
entire time the fan schedule says "on"? Is a lighting or equipment kW going
to be multiplied by a load factor or an hourly fraction in a schedule
somewhere? Is a chiller model going to take a straight user-input average
efficiency, or is the loading going to be compared to the capacity and used
to find the part-load efficiency?

Constant volume fans and pumps are usually sized to run at about 80% of
the nameplate hp. In existing buildings the fans and pumps may be running
well below capacity. For a constant volume supply fan or pump, the
modeler should input a kW roughly equal to what the fan or pump is
actually consuming since the simulation will use that input as the average
(constant) consumption.

Boiler and chiller capacity inputs should reflect the nameplate rating, not
the actual average consumption since the simulation will in most cases apply
a part load multiplier to the rated power. This is particularly important for
retrofit projects where the chillers are significantly oversized.
5.3.2 Default Capacities

If the modeler has access to the actual designed HVAC equipment

capacities, it is always better to use these capacities than to let the program
size the equipment.

In modeling existing buildings, use of default values for sizing HVAC

equipment may lead to inaccuracy in modeling existing buildings if the
system is oversized or undersized for the current application. However,
sometimes the modeler may have inadequate knowledge about the existing
building equipment, and thus must resort to auto-sizing of equipment (i.e.
program defaults). If she does this with the baseline (existing) building
model, it is important that she not continue the use of autosizing when
analyzing ECMs. Once the program has selected equipment size for the
baseline buiIding, these sizes should be input as the equipment sizes for
succeeding ECM simulations (unless an ECM specifically addresses
equipment size). Use of defaults for ECM runs in retrofit projects will
result in simulation of a comprehensive HVAC redesign that is not
occurring in reality.

5.3.3 Sizing Factors

In existing buildings, the capacity of equipment generally has been selected

on the basis of design outside temperatures dictated by code. Additional
capacity for morning warm-up and severe weather is accounted for by
heating and cooling sizing (or safety) factors. Most computer simulations
allow the modeler to input the sizing factors and the design conditions.

These sizing factor inputs can be important when the modeler allows the
program to auto-size equipment. If the modeler inputs equipment
capacities, however, the sizing factors are irrelevant. In our experience,
designers use heating equipment sizing factors of 1.1 to 1.5, depending on
the type of equipment and control, severity of climate, and inclination of the
client for legal adventures. Generally cooling equipment sizing factors are
1.0 to 1.2, since warm-up is not a consideration.

The DOE2 modeler has access to another input that affects auto-sizing of
HVAC equipment. This is the SIZING-OPTION keyword in the systems
module ZONE command. DOE2 defaults to a FROM-LOADS value for
this keyword. This default uses the LOADS single space temperature input
for both heating and cooling modes, resulting in an overestimate of peak
loads. It also assumes unconditioned spaces and plenums to be at a constant
temperature. If the modeler is depending on the program to size her
equipment, she should generally override the default with the ADJUST-
LOADS value.

5.3.4 Part Load Curves

Most energy simulation software uses part load curves to determine HVAC
equipment input power at those times when the equipment experiences a
load less than its rated capacity. Most equipment operates in a part load
condition most of the time. Chillers, fans, boilers, packaged air conditioning
equipment, refrigeration equipment, and so forth are all affected by part
load operation. In most cases, the equipment efficiency is different at the
various degrees of part load than at full load. Sometimes the equipment is
more efficient at part load than at full load; but in most instances, it is less
efficient.

Energy simulation can be strongly affected by part load curves. The energy
consumption of a piece of equipment at a given part load is a function of
three numbers: the full load or rated capacity of the equipment, the part
load (i.e. the percent of full load), and the efficiency multiplier for that
specific part load. We have already discussed rated capacities. Sometimes
these are input by the modeler based on design or audit information, and
sometimes these are automatically calculated by the program to satisfl the
calculated peak loads plus a specified or default safety factor. The accuracy
of the part load function depends on the accuracy of the rated capacity
input.

Most simulation programs do a competent job of calculating part load

fraction. This is simply the equipment load (directly or indirectly related to
heating or cooling load) at a given time divided by the full load capacity.
However, not all programs competently simulate part load performance.
Some programs hide their assumptions about part load performance. In
others, including DOE2, the modeler can directly input part load
performance.

Part load curves vary significantly depending on the type of equipment and
control. The curves can be linear, quadratic, cubic, and so forth. If an ECM
directly affects a part load curve, the modeler should ensure that the affected
equipment is described properly in the simulation. She should carefully
examine the output reports and use hand calculations as an adjunct to check
the simulation and to analyze part load efficiencies that can't be easily
simulated. Condenser water reset, chilled water reset, and variable-pitch
axial vane fans will create part load consumption patterns that are likely to
be lower than those calculated in simulation unless artificial inputs are used.

5.4 CONTROLS

Controls are one of the most important factors in determining energy

consumption. Theoretical savings for controls ECMs are easily
overestimated. Accurate predictions of energy savings can only be achieved
if controls ECMs are defined by a sequence of operations common to the
model, the installed system, and the use of the system once it has been
installed.

5.4.1 Sequence of Operations

The action of a set of controls is best defined by a sequence of operations.

(See the example in appendix IV.) Although general terms like optimum
start, intelligent recovery, and cold deck reset are often thrown about as
though their definitions were clear-cut, that is not the case. The energy
savings from any one of these strategies depends on the details of how it is
achieved. A modeler cannot accurately model any of these functions unless
a sequence of operations has been established, either by the analyst or by
another party in the project.

Unfortunately, it is often very difficult to get manufacturers to document

how their controls work. Sometimes the information is viewed as
proprietary. Other times the manufacturer's representatives don't
themselves understand the control logic being used. Designers and building
owners are also sometimes vague about what the functions of a proposed
control system will be.

Information about specific control components (e.g. thermostats, sensors,

actuators) is best obtained by talking with the manufacturer's design
engineers at their central office. Information about the sequence of
operations, though, is best obtained, often with difficulty, from either the
contract drawings or from the controls contractor.

5.4.2 Role of the Analyst

Communications between the analyst and the other parties in any project
involving controls are important to achieving consistency between estimated
and actual ECM savings. We note here several of the activities that are
critical to successful controls ECMs. Some of these activities are typically
the responsibility of one of the project contractors. But if the analyst sees
that a critical task is not being done, she or someone on the energy team
should step in.

1) Promote Development of a Clear Sequence of Operations.

In any communications between the analyst and the designers,
manufacturers, owners or building operators, it is important to seek a
common understanding and documentation of the sequence of operations
underlying the estimated savings. In these communications, the analyst will
need to find out from the designers and manufacturers what is being
proposed, and may also need to inform the other parties involved in the
project which aspects of the controls are most critical to savings.

Clarity can sometimes be achieved during the design and bidding phase
through use of performance specifications, written either by the designer or,
if necessary, the analyst.19 These specifications should include sequences
of operations to be performed by the new controls. The sequence of
operations should not be left up to the contractor developing the controls

l9 A performance specification differs from a full design specification

in that the performance specification gives enough design information to
thoroughly define the design intent. This should include a general
description of equipment required, functions to be performed, efficiency
requirements, general controls sequence of operation, start-up scope, and
general requirements for documentation and training. The full
specification goes well beyond this, giving complete details of equipment
to be purchased, codes to be followed, start-up procedures, submittal
requirements, and so forth. See appendix V for several examples of
performance specifications.
software, to be uncovered by the owner for the first time after installation.

2) Promote Documentation and Building Operator Training.

Even a simple night setback thermostat can be difficult to use properly
without adequate docurnenta tion. Control documentation should include a
description of the intent of the controls under all modes of operation, a
description of the control algorithms being used, and a clear explanation of
how to change the control settings and schedules. The performance and full
specifications should require adequate documentation, and training of the
occupant or building operator.

The customer needs to understand from the start that the controls cannot
operate effectively unless the setpoints and schedules fed to it are good.
User-friendliness of controls also is extremely important. Unfriendly
controls are ignored, bypassed, or otherwise improperly used.

3) Post Installation Evaluation.

For expensive control ECMs projects, the analyst should meet with the
owner and the designer during the design phase and discuss how the building
owner will be able to tell whether or not the controls system is working after
it has been installed. It also may be worth adding a few monitoring points
if necessary to assure that the performance of the new controls can be
evaluated after it is in (e.g., trend log capability for flow, duct static pressure,
and fan consumption to evaluate VSD control on large motors).

The Building Operator's Effect

The computer simulation is not equipped to model the building operator's

mind. This is a serious shortcoming. Building controls include an automatic
component and a manual component. The two are linked by the mind of
the building operator.

Where sophisticated controls are installed, the actual building operation will
usually be limited by the ability of the building operator to understand the
controls and the building. For these cases, a computer simulation will often
underestimate consumption.

On the other hand, the baseline run for a large building with few automated
controls will tend to overestimate the HVAC consumption since the building
operator often will be much smarter than the automatic control logic at his
disposal. H e or she will be constantly manipulating the controls according
to a higher order of understanding than any of the formal logic they can
offer. Most building operators will perform manual mixed air and supply air
temperature resets seasonally if possible, and will turn off the boilers and
chillers in a pattern that minimizes reheat, even where there is some
compromise in comfort. This is particularly true of existing buildings with
older, simpler control systems.

This tendency to underestimate the efficiency levels achieved using simple

automatic controls and overestimate the efficiency .levels achieved with
complex automatic controls can result in excessive estimates of savings for
controls ECMs.
To minimize inaccuracy for retrofit projects, the analyst should visit the
building at night to see what equipment is on (lights, fans, pumps, etc.). She
should also interview the building operator.

Malfunction:

Computer simulations are poorly equipped to model malfunctions.

Inaccurate sensors, controls wired in backwards, and malfunctioning dampers
and valves all can have a dramatic effect on the accuracy of baselines for
retrofit projects, and actual ECM savings in both new and existing buildings.

5.4.4 Specific Control Operations

Warm-up Controls:

The amount of time it takes for the space to come up to the occupied
setpoint depends in the simulation, as it does in real life, on the thermal
mass of the building, the capacity of the heating system, and the difference
between the temperature of the space prior to occupancy, and the occupied
setpoint.

Most programs begin to simulate warm-up at the hour that the thermostat
setpoint changes from the unoccupied to the occupied value. If the space
does not reach the occupied setpoint in the first hour of occupancy, the
simulation will generally begin to keep track of "hours of loads not met".
Loads not met during other parts of the day are cause for concern, but one
or two hours a day of loads not met during winter morning warm-up is
usually not a concern.

If a building warm-up begins either via manual or automatic controls before

occupancy, the "occupied" setpoint should not be assigned to occupied hours
only. The computer simulation that waits until the change from setback
thermostat setpoints to occupied heating space temperature setpoints will
tend to underestimate heating consumption.

Unless a simuIation is explicitly modeling optimum start, either (1) input

a change in thermostat setpoint an hour or two prior to occupancy or (2)
input a thermostat setting schedule that ramps up the heating setpoints
prior to occupancy. A ramp of 3 F per hour is typical for heat pumps, 5 F
for other heating equipment. The ramp brings on the fans and the heat
earlier on cold days than warm days, as does optimum start, but without the
fine tuning afforded by optimum start self-correction capabilities.

DOE2 offers an optimum start option, but many programs do not.

Heat Pump Setback:

Heat pump back-up heat (typically electric resistance) is usually staged to

come on when the actual space temperature drops more than about 4
degrees below setpoint (or a lesser temperature difference over a protracted
period of time, if integral controls are used). Use of a simple setback
thermostat would thus bring on the back-up heat on every winter morning
as soon as the thermostat setpoint changed back to its occupied setpoint.

Some of the "intelligent recovery heat pump thermostats" create a

thermostat setpoint ramp with a slope of 3 degrees per hour so the heat
pump may have a chance to meet the warm-up load during mild weather
without use of back-up heat. (Other intelligent recovery thermostats alter
the slope of the ramp somewhat on the basis of historical data, or apply a
more complicated algorithm. In general, it is difficult to obtain information
on the more complex methods.)

Computer simulations can be programmed with a thermostat setpoint ramp,

if the simulation offers a thermostat setpoint schedule. Some of the simpler
programs limit the heat and cooling setpoints to one set during unoccupied
hours and one during occupied. These programs generally do a poor job of
simulating heat pump setback.

Perimeter Heat:

Perimeter heat is a wild card in computer simulations since the relationship

between air distribution system zoning and perimeter heat is often complex
and difficult to quantify.

Air distribution system perimeter zones are usually 12 to 15 feet deep. In a

computer simulation, the heat gain from lights and equipment for that full
12 foot depth is subtracted from winter heat losses through the windows and
exterior walls when the HVAC load is calculated. By contrast, perimeter
heating systems - hot water, steam, and electric baseboards - are designed to
meet the window losses right at the inside window surface. A properly
operated perimeter heating system will in this way tend to increase comfort
while reducing efficiency (assuming the baseboard heat source is less
efficient than the air system source). Simulations add the envelope heat
losses to the lighting hcat gains for the perimeter zones. Thus simulations
tend to underestimate heating consumption by canceling window heat losses
with internal gain from lights.

Perimeter heating systems in large existing buildings are often controlled as

a function of outside air temperatures. In many cases, the air distribution
system may actually be cooling air that is overheated by the perimeter heat.
When that occurs, the computer simulation will dramatically underestimate
heating consumption. During mild weather the net heat load on a 12 foot
perimeter may be zero (heat loss through the glass being compensated by
heat gain from the lights), while the heat loss through the glass is still
significant.

In most software packages, baseboards are modelled more or less like reheat.
Since nothing can be done about the inaccuracies mentioned above, it is
only necessary to differentiate baseboards from reheat if they are served by
a different fuel, if they have outdoor air reset, or if they are controlled to
provide heat when the fan system is not on.
Economizer controls:

Computer simulations of economizers generally ask for the high limit shut-
off temperature and whether the control is based on dry bulb o r enthalpy
sensors. Humidity sensors a,renotoriously unreliable, and humidity generally
is not an important economizer performance parameter in the northwest.
Therefore, there is reason to promote use and simulation of dry bulb
economizer controls. The modeler should note that simple non-differential
dry bulb and enthalpy controls open the outdoor air damper when the
outdoor air temperature or enthalpy is less than the control setpoint. To
simulate such control, the modeler must enter that setpoint. Otherwise the
program may default to differential control, comparing outdoor air to return
air. This usually results in an underestimate of cooling energy consumption.

In some large buildings, the building operator closes the outside air dampers
when the chiller is on during mild weather, because the chiller does not
operate well at low loads. Thus the analyst should address chiller part load
operation in conjunction with economizer controls as part of the audit of
existing buildings and as part of the design phase of new buildings. If a
chiller has a minimum allowable loading level, that can be reflected in the
simulation by designation of a low "high limit shut-off temperature" for the
economizer.

Energy Monitoring and Controls Systems (EMCSs):

A building owner may be interested in an EMCS for a number of reasons,

including less maintenance than pneumatics, flexibility, responsiveness, alarm
o r avoidance of environmental problems, and reliability.

Within the simulation, by contrast, an EMCS is simply a set of independent

control functions, such as cold deck reset and optimum start, that are
modeled the same way for an EMCSs as for discrete controls.

As with any control savings estimate, the estimate of savings for an EMCS
must start with a list of the intended control functions (e.g. optimum start,
condenser water reset, cycling of the fan during unoccupied hours). T o
make an accurate analysis of savings, the analyst will then want to know the
specific sequence of operations to be employed. (See appendix V.)

After the installation, an EMCS may be more likely to save the estimated
amount of energy than discrete controls because use of monitoring points
and trend logs permit the owner, building operator, or building
commissioner to evaluate the performance of the new control system relative
to the intended sequence of operations.

Simulation of ECMs that involve EMCSs requires careful consideration of

the baseline condition. With existing buildings, the baseline is simply the
existing set of controls sequences as operating. With new buildings, the
baseline must include code-required controls sequences such as unoccupied
setpoint setback and reset and/or VAV control in multiple-zone systems.
5.5 SIMULATION OF MULTIPLE ZONE SYSTEMS

5.5.1 General System Characteristics

Most commercial buildings simultaneously require perimeter zone heating

and interior zone cooling during cold weather. Many commercial buildings
are served by multiple-zone HVAC air distribution systems that serve both
interior zones and perimeter zones. A multiple-zone system has to provide
simultaneously for the various conditions in all of the zones it serves. To
do so at a reasonable first cost, multiple-zone systems most often cool
supply air (either mechanically or with outside air) centrally and then heat
up some portion of that cooled air at those zones that require heating.
Depending on the design and health of the governing control system, this
can be a highly efficient or highly inefficient process.

Constant volume reheat systems and constant volume dual duct or multi-
zone systems without hot and cold deck reset have similar potential for
inefficiency since they all mix heated and cooled air when neither heating
nor cooling is needed. Dual duct and multi-zone systems have additional
inefficiencies introduced by duct leakage and zone damper leakage.

Today, multiple-zone systems have become more efficient by using hot and
cold deck reset and variable air flow rates for both reheat and dual duct type
systems. Nevertheless, the modeler still needs to produce accurate
simulations of reheat and mixing controls because,

1) baseline runs for existing buildings may be dominated by reheating or

mixing, and

2) mixing and reheat continue to play a lesser, but still significant, role in
new multiple-zone systems.

When a zone load varies, it can be met by varying the zone supply CFM or
the zone supply air temperature. Zone supply CFM can be varied in a
straightforward manner within certain limits if the system is VAV. If a
system is constant volume, variation of the zone supply temperature is the
only way to meet the varying zone loads.

Since supply air is cooled centrally it must be available at the coldest

temperature needed by any zone to meet the zonal cooling load. During
times when no zone requires 55 degree air, the central supply air
temperature can (and usually should) be raised to the highest temperature
that will still cool the zone requiring the most cooling (the "critical zone").
This automatic control is called central supply air temperature reset, and is
a common strategy for energy efficiency in multiple-zone buildings.20

20 Cold air reset may or may not save energy in a VAV system. The
increased cold air temperature at cooling part loads requires greater zonal
air flow than would be required without reset. This in turn increases fan
consumption, decreasing the savings of reset control. The net savings can
be determined using some of the more powerful simulation programs.
Since zone loads can vary independently from one another, every multiple-
zone system has some means of raising the zone supply air temperature for
one zone while lowering it for another. Central supply air reset is not
sufficient to provide this service. For all but the "critical zones", the central
cold supply air to each zone has to be heated even if the system has reset.
Each non-critical zone therefore has a "reheat" heating coil, or a set of
mixing dampers which mix the central supply cold air with central supply hot
air. In addition, depending on the system type, each zone may have a
damper for throttling the central supply air (VAV) and sometimes a fan for
introducing a secondary air stream consisting of warm ceiling plenum air to
the zone.

Simulations generally offer three hot and cold deck temperature controls:
constant, reset based on outside air temperature, and reset based on the
needs of the coldest and warmest zones. The modeler should be careful in
making this selection to pick the option that most closely matches the
existing building or proposed design. If reset based on outside air
temperatures is selected, the modeler usually has the option of determining
the exact reset schedule.

To the extent that heating and cooling loads in the critical zones are a linear
function of the outside air temperature for a certain range of outside air
temperatures, outside air reset can work well. To the extent that solar gain
is not directly linked to outside temperature, or that occupancy and internal
gain schedules vary during occupied hours, outside air reset may not
satisfactorily track zonal cooling loads. When modeling outside air reset
controls in a computer simulation, the modeler often must input the
proportionality relationship between the supply air temperature and the
outside temperature.

It is important to note that reset of the cold duct supply air temperatures
should ideally affect both control of the central cooling coil valve and
control of the outside airlreturn air mixing dampers. In other words, the
mixed air setpoint and the supply air setpoint should be reset together.
Otherwise the advantages of cold deck reset will be seen on reducing the
cooling consumption but not for reducing the heat consumption. O n dual
duct systems that have a separate set of outside air mixing dampers for the
hot duct, the same issue applies.

5.5.2 Multi-Zone and Dual Duct Systems

Multi-zone and dual duct systems generally have very similar models in
energy simulations. The only actual difference between the systems as-
installed is the location of the zone mixing dampers. In a multi-zone system,
the zone mixing dampers are located together near the supply fan. In a dual
duct system, the zone mixing dampers are located near each zone at the end
of the ductwork. However, a given program may have significantly different
defaults for equipment or operation depending on whether the multi-zone
or the dual duct option is selected. Therefore, the modeler must be alert (as
always) to the defaults in the program.

Some multi-zone and dual duct systems have been retrofitted to VAV
operation either by conversion to a VAV reheat system o r by changing the
zone mixing dampers so that they may alter the total (sum of hot and cold)
air flow rates to the zones. If a program doesn't offer a VAV multi-zone or
dual duct option, the most accurate simulation to use is probably a VAV
reheat simulation.

5.5.3 Variable Air Volume Systems

Variable air volume systems vary the heating and cooling delivered to each
zone by varying the amount of air supplied to the zone. By reducing the
amount of reheat (or mixing, as in multi-zone and dual duct systems), VAV
systems reduce the energy required for heating and cooling. By reducing the
total amount of air, VAV systems reduce the energy required for fan
operation.

VAV systems have central mixing dampers for outside air and recirculated
air, a central cooling coil, and, often, central supply air temperature reset
based on the critical zone. Some VAV systems have both supply and return
fans. The VAV volume damper for a zone reduces the cold central supply
air CFM rate to the zone below the design (maximum) flow rate when the
cooling load is low or heating is required.

From the point of view of energy efficiency it would often be desirable to

temporarily eliminate air supply to a zone. For several reasons, there is a
limit to how much the air flow rate to a zone can be reduced.

1) Ventilation requirements. The amount of outside air delivered to a

zone at any given time depends on both the position of the outside air
mixing dampers and the amount of central supply air delivered to the zone.
During cold weather, if the outside air mixing dampers are set at a
minimum, the amount of central supply air required for adequate zone
ventilation can be relatively high. In any case, during occupied hours, a
minimum amount of central supply air is required during occupied hours to
provide ventilation.

2) Air distribution. A certain air flow rate is required to operate zone

heating coils and VAV terminal units, and to adequately distribute the
conditioned air through a space.

3) Fan limitations. Fans are unable to operate at less than a certain

fraction of their design air flow rate. This limit is a function of the fan type
and the type of air flow control.

The minimum zone supply CFMs required for air distribution can be met
without use of centrally cooled air by providing a zone fan to recirculate air
directly from the conditioned space over the reheat coil and back into the
space. Such VAV systems, called "fan powered terminal unit VAV" reduce
the amount of reheat by reducing the minimum zone supply CFM.

If the system being modelled is a fan-powered VAV, the minimum zone flow
rate is set by ventilation requirements during occupied hours. During
unoccupied hours, it is zero.
Air flow to each zone is never dropped to zero during occupancy in VAV
systems. In actual systems, the zone dampers and fan controls have
minimum settings which take all of the above criteria into account. In
computer simulations, the user is required to input the minimum air flow
rates for VAV systems, as a percentage of the maximum. Such information
can sometimes be found in specifications or on the mechanical drawings, but
would be difficult to measure on-site because there may be many zones, and
the minimum air flow rate for each only occurs under certain zone
conditions.

The modeler should input a minimum supply air CFM of no less than
~ ~ a design supply air CFM of no less than 0.8 cFNl/ft2
0.3 C F M / ~and
unless she has audit or design information that states otherwise.

The efficiency of VAV systems in saving fan energy depends on the type of
VAV control. Reduction of zone supply air volume is accomplished by a
VAV air damper, which closes or opens to alter the air flow to the zone.
As the zone dampers close to varying degrees, the pressure drop through the
duct system increases, which increases the pressure drop across the central
supply fan, which in turn reduces the central supply CFM and the
horsepower consumed by the fan by one of the following method^.^'

1) Riding the fan curve (a simple increase in pressure drop through the fan,
resulting from closing volume dampers in the distribution system). Note
that this method often results in high air leakage losses through ducts and
dampers as a result of the high pressure in the duct system.

2) Discharge damper. Discharge dampers increase the pressure drop across

a fan without increasing the pressure in the ductwork.

3) Inlet vanes.

4) Variable speed fan drive.

Some VAV systems actually have a constant air flow through the fan but
reduce reheat by dumping unneeded supply air directly back to the return
air without passing it through the zone. If a computer simulation doesn't
offer this bypass option, the VAV simulation will underestimate fan energy
consumption for such systems.

5.5.4 Water-Loop Heat Pump Systems

Most simulation software programs do a mediocre job (at best) of modeling

21 Some designers are using an alternative control logic called

terminal regulated variable air volume (TRVAV). With TRVAV, the
EMCS sums sensed air flow at each of the terminal units and sets fan air
flow accordingly. This theoretically decreases fan energy consumption by
allowing the fan to operate in part load conditions at a lower static
pressure. Simulation of this type of fan control requires a program that
allows input of a user-defined fan curve.
water-loop heat pump systems. And few programs can directly simulate the
added complexity of a ground-water source. It is important that the modeler
seek a complete understanding of the specific software algorithms related to
the simulation as well as of the actual system to be simulated.

In many cases the defaults of the water-loop heat pump system algorithm
cannot be overridden by the modeler. This happens even with powerful
programs such as DOE2. DOE2 does not allow different operating
schedules for the individual zone heat pumps, and does not allow
economizer operation. When the modeler encounters limitations such as
these, she should ask herself whether an artificial input will more accurately
simulate the actual system performance.

The simplest artificial input for a ground-water source loop heat pump
system is to input individual packaged single zone units with hydronic
heating and cooling coils for each zone. Then input a central plant with
chiller, boiler, and tower COPs calculated to simulate the heat pump COPS.
at the normal water loop temperature. Circulation pump power can either
be manually calculated or calculated with program pump algorithms. This
approach can also be used for water-loop systems that do not have a ground-
water source. However, the accuracy will be poorer for this application
since the actual heat pump COPs will vary with the wider range of loop
water temperatures.

If the modeler chooses to use default program algorithms to simulate

ground-water source systems, she should be careful in her accounting of
tower and boiler power. We recommend that she input a very high COP (or
low energy-input-ratio) for this equipment. She may also wish to input a
non-electric fuel for the boiler to aid with accounting. If the program does
not allow user input for the maximum and minimum loop temperatures, she
should either use a different program or resort to the artificial input
described above.

5.6 VENTILATION

Ventilation and infiltration are powerful drivers of energy consumption in

northwest buildings. All simulation programs allow separate inputs for
ventilation and infiltration. However, programs vary widely in the ways they
combine ventilation and infiltration loads to calculate the total outside air
flow into the building.

Many programs assume zero infiltration whenever the fan is on, based on
the theory that the outside air being introduced to the building through the
ventilation system will pressurize the building, creating a net outflow of
conditioned air.

In reality, homogenous pressurization of a building is difficult. Wind

patterns, air balance, thermal buoyancy of warm air in a tall building on a
cold day, volume dampers, dirty and clean air filters, and control of return
fans off of building pressure sensors all affect the building pressurization.
In small buildings, entry and egress, and occupant operation of doors and
windows add to these factors. Even with the fans on, there is likely to be
infiltration is some parts of a building.

While it may be optimistic to simulate no infiltration when the fans are on,
DOE2 handles this differently. It adds the infiltration load to the ventilation
load when the fans are running. Thus if the modeler thinks infiltration is
likely to diminish when the supply fans are running (a likely supposition),
she must adjust the infiltration and ventilation schedules to reflect reduction
of infiltration during fan operation. This can greatly increase modeling
complexity, particularly if any ECMs affect fan schedules.

Some simuIation programs compare the infiltration and the ventilation, and
take the larger of the two. One program cuts the infiltration to zero when
fans are running only if the fan static pressure is above a certain threshold.

Since ventilation and infiltration are often large loads, it is important for the
modeler t o find out how the simulation does handle infiltration when the
fans are and are not running. The modeler should also convert ventilation
and infiltration into a common unit of measure so she can compare their
magnitudes. Otherwise, it is difficult to gauge the significance of fan
operation and outside air damper positions.

If the simulation does not automatically reduce infiltration when the fans are
on, and the modeler feels that the building is successfully pressurized, she
should decrease the ventilation rate input to compensate for the greater
infiltration rate.

Even when the program does permit input of an occupied infiltration

schedule, the modeler may wish to avoid the complexity of coordinating the
infiltration and ventilation schedules. She can do this with input of a
constant infiltration value equal to unoccupied infiltration, and input of a
ventilation volume equal to the actual ventilation volume plus the change
in infiltration due to fan operation. She. must also input a ventilation
schedule.

For existing buildings, the analyst should consider the condition of the
outside air dampers before deciding on the minimum outside air setting for
the simulation inputs. Dampers that haven't been properly maintained tend
to have a high leakage rate when they are at the minimum setting.

If infiltration is input in air changes per hour, the modeler should be careful
not to overestimate infiltration loads in zones with high ceilings and no
windows. If minimum outside air is input as a percent of the supply CFM,
she should input accurate supply CFM, especially in simulations of existing
buildings.

Some simulations permit the modeler to input an hourly schedule of

minimum outside air settings. This is particularly helpful for modeling
controls that set the minimum outside air at zero during unoccupied hours
t o prevent unnecessary heating for ventilation when ventilation is not
needed. It is important when using an hourly schedule for minimum outside
air settings to remember that the leakage rate through closed dampers is not
zero. For existing dampers, visual inspection of the position of the blades
in the "closed" position is warranted. For new construction, a 5 t o 10%
leakage rate can be assumed, o r the modeler can ask the designer what
leakage rate is going to be specified.

Indoor air quality questions are being pursued through lawsuits and
academic studies. Standard ventilation rates are increasing, so this
component of the energy simulation is likely to become even more critical.
Use of low leakage outside air dampers for 100% recirculation during warm-
up, minimum supply air volumes in VAV systems, and reduction in fan
operating hours may be limited in some cases by concerns for indoor air
quality.

5.7 FAN SCHEDULES AND SUPPLY CFM

Fan schedules and supply CFM are important in determining heating and
cooling consumption because they are closely linked to outside air
ventilation HVAC loads, to the amount of reheat or mixing of hot and cold
air that takes place in multiple-zone systems, to central plant equipment
operation, and, of course, to fan motor energy consumption.

In computer simulations, the fan operation during unoccupied periods is

typically designated as either off, cycling on only as necessary to maintain
heating and cooling setpoints, or on continuously. The amount of time the
fan is off has a dramatic effect on energy consumption.

If the minimum ventilation air is input as a percent of the total supply CFM,
the supply CFM becomes important in determining the outside air loads on
the HVAC system.

.In a dual duct system, or a system with central cooling and zone heating,
high supply CFMs and extended fan schedules increase the amount of air
that is centrally cooled and then reheated. In large buildings, it is common
during mild weather to turn on fans during unoccupied hours when the
boilers and chillers are still off. Simulations may for that reason tend to
overestimate heating and cooling during unoccupied periods unless the
boiler and chiller schedules can be, and are, input accurately.

In new construction, the supply CFM is generally determined by the design

cooling loads. However, designers will usually specify a certain minimum
supply CFM to ensure adequate air distribution and ventilation.

For retrofit work, the existing supply CFM may be more than what is
required to satisfy current loads. If that is the case, allowing the simulation
to assign supply CFMs may underestimate heating, and to a lesser extent,
cooling loads. The original design CFM should be read off the mechanical
drawings or, preferably, the latest air balance report, unless the building
operator o r owner indicates that there has been a significant undocumented
change in the air flow rates.

For both new and existing buildings, the modeler should check that the
input for design supply CFM/~~'looks reasonable. If audit o r design
information is available, these should be used. If not, very rough rules of
thumb can be outlined as follows:

For constant volume systems:

perimeter areas, South, East and West: 0.5 to 3.0 C F M / ~ ~ ~

perimeter areas, North: 0.5 to 2.0 CFM/ftL

interior areas: 0.75 to 2.0 C F M / ~ ~ ~

For Variable Air Volume Systems:

all areas: 0.5 to 1.5 C F M / ~ ~ ~

5.8 FANHEAT

The supply fan itself can impose a significant cooling load. The temperature
rise of the air across the fan is typically 1 to 3 degrees F. The temperature
rise across the space, when it is being cooled, is roughly (78°F - 55°F) =
23°F. The fan heat therefore can be roughly 10% of the cooling load.

If the motor is located in the air stream, the temperature rise across the fan
is higher than for motors located outside the ductwork. Heat gain from
typical electric motors is given on page 26.8 of the 1989 ASHRAE
Handbook of Fundamentals.

Most simulations will automatically calculate the temperature differential

(delta-T) across the fan on the basis of air flow rates, fan energy
consumption, fan and motor efficiency, and fan motor placement (assumed
or input), so this is not an input the modeler normally needs to worry about.
It is however helpful to note its importance in case the modeler decides for
any reason to input the fan delta-t rather than using the default value. The
modeler also should remain alert to programs that do not automatically
calculate the fan delta-T. Should she encounter such a program, she can add
a zone load equivalent to the effect of the fan delta-T.
6. OTHER ENERGY CONSUMPTION

6.1 EXTERIOR AND "HIDDEN" ENERGY USERS

This is a loose category that includes both energy loads external to the
building and easily-missed loads within the building that do not contribute
to HVAC heating or cooling loads. External loads might include car washes,
exterior lighting, swimming pools, gas pumps, water pumps, sidewalk heating
systems, and so forth. Non-HVAC driving interior loads might include
elevators, laundries, process equipment, and so forth. If the modeling intent
is to estimate the actual utility bills that the owner wiIl be paying, then it is
imperative that the modeler try to calculate or simulate these loads.
However, if the modeling intent is to estimate either HVAC system
performance or the incremental savings of ECMs that influence the HVAC
systems, then it is less important to simulate the hidden energy users.

There is at least one internal load that is easily missed but that can strongly
drive building energy use and HVAC performance--a large mainframe
computer system. Such systems can have an annual energy use index (EUI)
on the order of 2.0 k ~ h / f t ~for r office buildings (reference 10). To
- ~large
put this in perspective, this is the largest contributor to the equipment end-
use for both large and small offices, as determined by the ELCAP study.
The modeler should specifically ask the owner and design team whether they
plan such a computer system for the building in question. Since these
systems are typically served by a stand-alone AC system, the modeler may
decide that the purpose of the design-phase modeling can be met without
considering the computer center and its supporting equipment.
7. THE BASELINE BUILDING

7.1 GENERAL DISCUSSION

We cannot over-stress the importance of the baseline building model. The

energy savings calculation always has two parts--the building to be evaluated
(with one or more ECMs) and the baseline building. Error in the calculated
savings can arise equally from error in the as-designed building or from error
in the baseline. But we have found that modelers most often are sloppy in
their generation of the baseline model.

The Energy Edge project attempted to comprehensively define baseline for

the range of building types encountered in the project. However, one of the
lessons learned from the project was that it is impossible to cover all of the
building aspects in a baseline definition. Nevertheless, it is possible to state
a hierarchy of sources for definition. Energy Edge selected the following
hierarchy (starting with the highest source):

1. Model Conservation Standards Equivalent Code (February, 1985)

2. Common Practice Matrix. PECI convened a group of architects
and engineers to define baseline glazing area and HVAC system
type for a broad range of commercial building types.
3. Modeler judgement when neither source 1 nor 2 directly
addresses an issue.

In addition to these sources, Energy Edge provided modelers with a list of

"parameters held constant". Quoting from the Energy Edge Technical
Requirements:

"The following building parameters shall be held constant between the

MCS Design and the .7 MCS Design building.

1. Weather - hourly ambient wet and dry bulb temperature, wind

speed and direction, and solar weather data. The data should be
representative of the construction site.

2. Occupancy and Function - Number of people, schedule(s) of

occupancy, and sensible and latent heat gains.

3. Usage schedules of installed interior and exterior lighting. This

means the schedule for which lighting is made available either
through daylighting or electrical lighting.

4. Zoning of Building Area - At a minimum, each different

occupancy located in the building (e.g. office space, cafeteria, etc.)
shall be modeled with a separate set of zones. Perimeter zones
shall be between 8 and 20 feet deep for the MCS Design.

5. Internal Equipment (e.g,, elevators, computers, office machines,

refrigeration, etc.) - Operating schedules and internal heat sources
to conditioned spaces.
 6. Ventilation Air - Fan operating schedules and minimum amounts
(in CFM) of outside air.

7, Heating and cooling system schedules - Months of the year, days

of the week, and hours of the day when heating and/or cooling
are available from the primary and secondary system equipment
(except where thermal storage and/or night flushing and/or EMS
controls are shown to save energy).

8. Thermostats - Space conditioning set points and their related

schedules of operation (except where variable space temperature
techniques can be shown to save energy).

9. Floor area and wall area.

10. Energy Sources - Energy source for each end-use (no fuel
switching). Heat recovery is not precluded.

11. External shading from existing buildings and vegetation located at

the construction site."

Energy Edge abandoned the Common Practice Matrix relatively early in the
program. Instead it treated both glazing percentage and HVAC system as
parameters held constant (PHC)--unless either is directly affected by an
ECM. (As an example, skylights might be considered an ECM when tied to
a daylighting lighting control system. Similarly, a water-loop heat pump
system might replace the baseline air-to-air heat pumps.)

The MCS generally held its position in the baseline source hierarchy
throughout Energy Edge. Similarly, the PHC held its position. However,
interpretations were required of both in numerous instances. The most
common requirement for interpretation of the PHC related to exceptions
when a parameter was affected by an ECM. For instance, lighting, HVAC,
or equipment schedules could be changed between the baseline and as-
designed buildings if an ECM specifically provides automatic controls to
make that possible. Item numbers 3,5,6,7,8, and 11 each has the potential
for ECM exceptions.

The MCS is Swiss cheese when it comes to defining baseline. We have

already mentioned that it doesn't address a host of parameters. In addition,
when it does address a parameter, it sometimes does so with a laxity that
defies common-sense. An example of this are the MCS requirements for
wall U-value. The maximum acceptable U-value for most commercial
building walls is 0.25. Though glazing typically elevates the overall U-value
of a wall area, the MCS requirement often leads to a calculated insulation
value o n the order of R-5to R-7. This is not common installation practice!
The modeler should input for the baseline the minimum insulation value
that both satisfies MCS and represents common construction practice.

Carrying this one step further, we recommend that the modeler be given the
option of using the building that the program appIicant submits as hislher
non-program proposed design for the baseline, providing that that proposed
design at least satisfies MCS.

7.2 BASELINE ISSUES

Generation of a baseline model can raise some interesting technical issues.

We have already dealt with several of the code-related issues. Now we turn
our attention to some others.

HVAC equipment downsizing is one issue that often arises when baseline
derivation is discussed. This relates to the concept that a more efficient
building (lower lighting loads, lower solar gains, lower heat loss, etc.) should
be designed with HVAC equipment of a smaller capacity than the less
efficient baseline building. Our Energy Edge experience (primarily with
small to medium size offices so far) indicates that equipment downsizing
usually had little effect on estimated ECM savings or ECM ranking.

Also, it is our observation that designers tend to ignore or distrust the

potential downsizing benefits of ECMs and therefore size HVAC for
efficient buildings much as they would for the hypothetical baseline.
Designers are especially cautious about potential future increases in
equipment and lighting power density. Finally, incorporating downsizing in
the modeling process adds considerable complexity. Therefore, to simplify
our modeling, we generally ignore downsizing as an issue. In design-phase
modeling, the modeler should use the designer's selection for the HVAC
equipment capacity for as-designed building. She should use the same
equipment for the baseline building unless the output indicates that
increased capacity is necessary to meet loads. In the latter case she should
let the software program select the capacity required, being alert to possibIe
unrealistic sizing by the program.

We have previously alluded to the issue of lighting controls. What is the

appropriate baseline for such ECMs? If a daylighting ECM has been
modeled by enabling a program switch for daylighting, then the baseline
model simply turns off that switch. In the case of occupancy sensor controls,
the modeler should resort to ASHRAE Standard 90.1 requirements. The
modeler should simulate the effect of occupancy sensor lighting control by
multiplying the relevant baseline lighting power densities by 0.7. Note,
though, that this factor should be used only for modeling the ECM, not for
increasing the allowed baseline lighting power density.

Infiltration is discussed in some detail in section 4.2.1 of this report. The

modeler .should consider infiltration as a parameter to be held constant
between the as-designed and baseline buildings unless an ECM is designed
to directly decrease infiltration in a predictable manner.

Another baseline-related issue that seems to confuse some modelers is

whether the baseline and as-designed model should have different
assumptions concerning proper operation.

One could argue that since, statistically, most building occupants cannot
program programmable thermostats, then the modeler should simulate such
buildings without night-time setbacks. This arguer might continue with the
proposal that an energy management system (EMS) be considered an ECM.
We counter that it is unreasonable to assume improper operation for a
baseline technology and proper operation for a n ECM that is at least as
complex. Since design-phase modeling generally assumes proper building
operation for the as-designed model, it is appropriate to assume the same
for the baseline model. For new construction, the modeler should assume
proper operation of all building systems in both the baseline and as-
designed models. For existing buildings, the modeler should assume actual
operation for the baseline model.
8. ERROR CHECKING AND DEBUGGING

8.1 GENERAL ERROR CHECKING APPROACH

At the same time the modeler is keeping track of how the software works
and assessing its accuracy, she also needs to keep a vigilant eye on her own
contributions to the simulation.

Computer building energy simulations require a large number of inputs,

some of which are less than exciting. An accurate simulation requires that
the modeler provide an accurate detailed description of the building to be
simulated, and inputs that information into the computer without careless
errors. Patience with detail is an important part of good simulation work.

For accuracy:

Check calculation of inputs for errors prior to keyboard input.

Check to make sure the units of measure (e.g. kW, hp, feet, Btu)
agree with instructions in the simulation manual.

Check inputs for typographical errors before running the model.

Check output to see whether end use consumption and hourly

and monthly profiles are reasonable. (Include comparison of
actual and simulated consumption where possible.)

Do not "tweak uncertain inputs before the previous steps have

been taken.

Early detection of errors reduces the amount of time spent on reruns; and
careless input errors are sometimes more easily detected by examining the
input than by working backwards from the output. Results in the output
which appear abnormal may be due to any number of causes, including
software glitches or influences on actual consumption which have not been
explained to the modeler by the occupants.

8.2 INPUT CHECK

Most simulation software packages provide some form of report that echoes
the modeler's input. Some programs will change the modeler's input under
certain circumstances (if the input is viewed within the simulation as falling
outside the realm of reason). But the echo reports typically document the
adjusted input. Careful inspection of the input echo reports is thus usefuI
for two purposes--to see what the program thinks the modeler input, and to
see how the program changed what the modeler input.

What should the modeler specifically look for? We recommend the

following checks, as a minimum:

building orientation
zone definitions
 wall, roof, floor, and window definitions (orientation, area, zonal
affiliation, construction, U-value, glazing shading coefficient)
lighting and equipment power densities
operating schedules (for lighting, equipment, fans, thermos tat
setpoints, plant equipment, occupancy, etc.)
HVAC definition (CFM, input power, zones served, minimum
outside air percentage, system type, heating and cooling
capacities, fan schedule)
plant equipment definition (equipment type, capacities, rated
efficiency, part load efficiencies)
utility rate schedules if economic analysis used

8.3 OUTPUT CHECK

It is not at all unusual that a model will appear to be correct even though
it is correct for the wrong reasons. Offsetting errors do not make a
successful model. Though the model may accurately estimate annual energy
use, if it does so for the wrong reasons, then the ECM savings analysis may
be wildly inaccurate. The modeler must ensure that her model adequately
simulates actual building operation even before she checks whether the end-
use EUI results are "reasonable".

As a minimum, we suggest that the modeler check for the following in the
output reports (not all software packages include all appropriate reports):

Loads not met. Do the HVAC systems and plant as input satisfy
the building heating and cooling loads? If they do not, it could
signify improper loads, systems equipment, or controls input. It
also could signify an inadequate design. O r it might signify
several other things (including reasonable operation, as in
morning warm-up). In any case, a significant value of loads not
met is a serious warning signal. In DOE2, we find a "% of hours
any system zone outside of throttling range" (in the BEPS report)
greater than about 5 7 % to be of concern. The SS-F reports
provide further information for trouble-shooting.

Fan schedule. This is the source of many common modeling

errors. Many software programs have different fan schedule
defaults for different HVAC systems. Fan schedule error
contributes to model inaccuracy primarily through incorrect fan
power and ventilation air simulation.

The modeler can guard against fan schedule error by investigating

whatever relevant report the software provides. In DOE2, both
the SS-C and the hourly reports can be useful for this
investigation. In other programs there may be similar reports, or
the modeler may have to deduce fan operation from total monthly
fan consumption.

Note that this is a critical input, and one that is often in error.
We are not recommending that it be checked through an input
echo report! The input echo may not be a true indication of how
 the schedule is actually used by the software. We strongly
recommend that this input be checked through the program
output--either directly or with manual calculations.

Ventilation air. Like fan schedule, this input is a frequent source

of model error. Ventilation air is a direct contributor to heating
and cooling load through the equation,
Btu/hr=1.08 x CFM x (To-T,). 22

Error can arise from either the ventilation CFM o r from the
ventilation schedule. Software programs often tie the ventilation
schedule to the fan schedule. We discussed this effect and how
to troubleshoot it in the previous item. But some software
programs allow input of a distinct ventilation schedule. When
this is the case, the modeler should check whatever output report
gives an indication of the ventilation simulation. In DOE2, the
most direct indicator of simulated ventilation schedule is the
systems hourly report #39, "OUTSIDE/TOT CFM".

There is another reason for debugging ventilation air simulation--

economizers. The effect of an economizer is the variation of the
ventilation air volume as a function of several temperature or
enthalpy conditions. We have often found that when we think we
have input a certain economizer logic, the program simulates
something else. We recommend that the modeler take advantage
of whatever output report is available to verify the economizer
simulation. Again, in DOE2, the systems hourly report #39
along with the appropriate outside air temperature (or enthalpy)
are the most direct reports for verifying this simulation.

Simultaneous heating and cooling. This is a less frequent source

of error than the previous two items. And sometimes
simultaneous heating and cooling is representative of how the
building will actually operate.

Nevertheless, the modeler should check the output reports to

ensure that the software program simulated the HVAC system
operation in a way consistent with her (and the designer's)
intention. If the HVAC system or control system modeled does
not allow simultaneous heating and cooling, then the output
should show none. Conversely, if the systems modeled are based
on simultaneous heating and cooling (e-g. variable air volume--
VAV--with reheat), then the output should show some. Note
that not all programs provide output that will enable the modeler
t o determine whether simultaneous heating and cooling occurs.

Errors due to conflicting inputs. Some programs allow conflicting

22 where Btu/hr is the heating or cooling load imposed by ventilation

air, CFM is the volume rate of the outside air, To is the outside air
temperature, and T, is the return air temperature.
inputs. Conflicts may be resolved internally without notifying the
modeler. Inspection of hourly reports for selected seasonal
conditions (hot, cold, and swing) is an excellent way of assuring
that the program works as intended.

HVAC system and plant COPs or EERs. Once again, what the
modeler inputs is not always what the program simulates. What
a devious world we model in! Simulation programs sometimes
recalculate the efficiency input using hidden defaults, auxiliary
energy consumption, and so forth. Only by investigating the
relevant output reports can the modeler ascertain that her intent
was honored.

We recommend that, at a minimum, the modeler use the output

reports to calculate seasonal heating and cooling COPs. She can
do this by dividing the total cooling (or heating) output for a
given time period by the relevant equipment input for the same
period (all units in Btu). For vapor compression air conditioning
or heat pump equipment, one would expect seasonal COPs
ranging from about 2 to 3.5. Chillers could range from 2 t o 5.
Values much higher o r much lower should prompt further
investigation.

HVAC system operation. The algorithms for simulation of

HVAC system operation are complex. Various software programs
simulate HVAC with varying degrees of success. And various
modelers of course simulate HVAC with varying degrees of
success. Simulated HVAC controls are never an exact
representation of actual building controls. Some system types--
such as variable air volume (VAV), water loop heat pumps,
ground source heat pumps, and thermal energy storage--are
especially difficult to model. Some control sequences--such as
night flush cooling, heat pump morning warm-up, and
refrigeration heat recovery--are especially difficult to model.

Because of the capacity for error, the modeler should check

output to verify that design intent was simulated correctly.
Sometimes when the simulation and the design intent are not in
agreement, this may be a red flag for design error. O r it may be
a red flag for simulation error or inadequacy. In either case, the
modeler should notice the flag and act accordingly.

When debugging the model for HVAC simulation error, it is

often necessary for the modeler to refer to the software
documentation. Energy simulation software uses various synthetic
algorithms to model HVAC systems. These algorithms
incorporate assumptions, simplifications, and sometimes errors.
It is frequently essential for the modeler to understand these
algorithms in order to understand the simulation input and
results.
8.4 EUI CHECKS

In many cases the software package used does not provide the input o r
output reports alluded to in the previous subsection. For this reason, and
for an additional safeguard of model reliability, we recommend that the
modeler compare the baseline building simulation results to independently
collected data for a similar building type. For the use of northwest energy
conservation programs, the Lawrence Berkeley Laboratory has compiled data
from three sources: ELCAP monitored end-use data, the Northwest Power
Planning Council 1991forecast (reference 191, and the SBW Consulting Inc.
prototypes study (reference 20). These data are combined in a single table
with statistically derived EUI ranges for each major end-use and for a variety
of building types (see appendix 11).

The baseline model should be used for this comparison since, as a code-
based building, it is more likely to be similar to the ELCAP new commercial
construction buildings than the more efficient as-designed building. Since
the baseIine model is derived from the as-designed model, if the baseline
end-use EUIs are similar to the statistical data, this implies that the as-
designed model shares the same validity.

Discrepancies between the simulation results and the statistical data do not
necessarily indicate an erroneous model. These discrepancies may be due
to justifiable differences between the modeled building and the average.
building represented in the data. Such justifiable differences might include
weather files, operation, schedules, equipment power densities, and so forth.
Nevertheless, statistical data are a good indication of typical end-use EUIs
for commercial buildings in the northwest.

The modeler should compare the end-use EUIs in the building baseline
simulation to the data in appendix 11. The modeler should troubleshoot the
simulation input whenever a simulated end-use EUI differs from the
corresponding EUI high or low range value by more than +35%. If the
troubleshooting convinces the modeler that the input and the simulation are
correct, she should discuss and justify any such discrepancies.
9. DOCUMENTATION

9.1 PURPOSE O F DOCUMENTATION

The as-designed model (and the corresponding baseline) may be used in

several different ways. Its primary use of course is to estimate ECM savings
for the building owner, developer, design professionals, facility personnel,
and program decision-makers. Beyond this, the as-designed and baseline
models may be used as the basis for a future as-built model, o r even a tuned
and calibrated model. They also may be subjected to expert review. Finally,
they may serve the modeler herself as a basis for other building simulation
projects.

In all cases, proper technical use can be made of the models only if the
modeler has thoroughly docurnentcd the assumptions, calculations. sources,
and so forth that she used when developing the models. Without this
documentation, future users can never know whether a given input was
based o n calculation, informed assumption, guess, desire, or error. As a
corollary, lack of documentation promotes distrust of the simulation results.

Note that the model documentation package as discussed here may overlap
with, but is not the same as, the building design assistance report. That
report addresses the primary modeling purpose only--to estimate ECM
savings for the building owner's team and program decision-makers. The
model documentation package, on the other hand, is a technical document
that addresses everything a reader would have to know to fully understand
the building model.

9.2 DOCUMENTATION REQUIREMENTS

In this subsection we present the general requirements for model

documentation for Bonneville's Energy Smart program and associated
programs. Though these requirements may seem to some to be excessive,
we have found that this extent of documentation generally saves much time
in the long run--not only for other parties, but for the modeler herself.

The modeler should document for both the as-designed and the baseline
building the following:

State company names, contact names, addresses, and phone numbers

for as many of the following as available:
0 architect
0 mechanical engineer
0 electrical engineer
0 lighting designer
0 owner
0 general contractor
0 mechanical subcontractor
o electrical subcontractor
0 modeler

State software package selected and reasons for selection.

List information sources available to and used by the modeler (e.g.
O&M audits--give dates and auditor names, building drawings,
specifications, phone conversations with designers, job conferences,
etc.).

General building description. Include building area, location, proposed

usages and occupancies, number of floors, number and allocation of
electricity meters, and so forth.

Envelope construction description. Include both actual proposed

construction details (materials, layers, R-values, stud type and spacing,
etc.) as well as how the construction is represented in the models.

W A C system description. Include actual proposed equipment

specifications as well as how the systems are modeled. Include
manufacturer's data where these are needed to support calculations.
Explain model departures from specifications. Describe assumptions
regarding fan control, minimum outside air volume and control,
setback, warmup, reset, and so forth. Describe central plant
equipment and control. State design criteria, if known.

Lighting description. Describe lighting fixture, lamp, and ballast types

and installed power density for main building areas. Describe
switching and other controls. Describe exterior lighting and controls.

Base electrical loads description. State power density assumptions

(and sources of assumptions) for office equipment and other base (not
to be confused with baseline) loads. Differentiate between different
building areas where appropriate (e.g. cafeteria, office, warehouse).
State assumptions and sources for other base loads such as elevators,
refrigeration equipment, process equipment, and cooking equipment.

Operating schedule assumptions. Include all relevant zonaI schedules

for each scheduled end-use. State sources of schedules. (It's
acceptable to state that an assumption is the result of the modeler's
best guess.)

Clearly define building zoning as modeled. Describe the rationale of

zoning if it departs from actual installed W A C zones, Include floor
plans with modeled zones clearly marked.

Include code compliance check calculations for both models. Clearly

state all relevant assumptions.

Describe fully each ECM that has been analyzed. Follow a format that
includes:
o ECM name
0 short description
o baseline description as relates to the ECM
o specifically how the ECM differs from baseline
0 cost estimates for baseline and ECM--include by item,
materials, labor, contractors' overhead and profit, design
 fees, contingency, commissioning, and annual maintenance.
State sources for all values.
0 incremental ECM cost

Provide a table showing a summary of the modeling results. This table

should include electricity usage and demand, fossil fuel usage,
electricity energy use index (EUI) in k ~ h / f t ~ -and
~ r ,total fuel EUI in
~ t u / f t ' - ~for
r the baseline model, for each ECM, and for the package
of all recommended ECMs.

Provide a table showing the annual end-use EUIs (electric, fossil, and
both) of the baseline model. Also show the ELCAP end-use EUIs for
the relevant building type and calculate percentage differences between
the two. In the text of the documentation explain any percentage
differences greater than 35%.

Provide a table summary of all recommended ECMs. This table

should include for each ECM:
0 ECM description
0 electricity demand savings
0 electricity consumption savings
0 other fuel savings
0 total incremental cost
0 estimated measure life (from Bonneville Technical
Requirements)
0 levelized cost (from requirements of specific program)
0 baseline basis for savings estimates (i.e. MCS, MCS code
equivalent, proposed building, etc.)

Describe any software package limitations that might affect these

specific models. Does the model provide direct simulation of the actual
systems and operation? If not, describe how this was handled.

Describe' any known bugs, errors, discrepancies, conundrums, etc. in

the models. Describe any unusual insights into the energy behavior
of the building that may have been gained during the course of
modeling. These might include discovery of parameters to which the
building consumption is particularly sensitive,

Provide glJ backup calculations and assumption sources. This is

critical. These would include, but not be limited to, infiltration values,
U-values, lighting and equipment power density, equipment and plant
operation curves, glazing parameters, and other calculated inputs.

Provide hard copies of the model input and output, If the models used
DOE2, the output should include reports LV-C, LV-D, SV-A, SS-D (if
there is a central plant), PV-A, PS-B, PS-C (if there is a central plant),
and BEPS. Also include ES-D if the economics module was run.
Finally, include any other reports that are necessary for the reader to
understand important results of the modeling.
 REFERENCES

Kaplan Engineering, 1989. Energy Edge Pilot Study. Available from

Bonneville Power Administration, Commercial Programs Branch.

Kaplan Engineering, Portland Energy Conservation Inc., 1992.

Energy Edge Simulation Tuning Methodologies. Available from
Bonneville Power Administration, Commercial Programs Branch.

Gale C. Corson Engineering, 1990. A Comparative Evaluation of

Commercial Building Energy Simulation Software. Available from
Bonneville Power Administration, Commercial Programs Branch.

Momentum Engineering, Ecotope, Synergic Resources Corporation,

1988. Major Projects Rule Phase /I Evaluation - First Year Report.
Available from Seattle City Light, Contract No. C87-325.

Pacific Northwest Laboratory, 1989. Description of Electric Energy

Use in Commercial Buildings in the Pacific Northwest - ELCAP.
Available from Bonneville Power Administration, Commercial
Programs Branch.

ASHRAE, 1989. Handbook of Fundamentals. American Society of

Heating, Refrigerating and Air-Conditioning Engineers, Inc.

State of Oregon, 1990. Energy Code Compliance Manual.

Available from Oregon Department of Energy.

ASHRAE, 1989. Standard 90.1- 1989. Energy Efficient Design of

New Buildings Except Low-Rise Residential Buildings. American
Society of Heating, Refrigerating and Air-Conditioning Engineers,
Inc.

M.J. Witte, BLAST Support Office, 1987. "Simplifying BLAST Input

- Describing as Few Zones as Possible." BLAST News, Vol. 5, No. 3.
University of Illinois at Urbana-Champaign.

Pacific Northwest Laboratory, 1990. Commercial Equipment Loads,

ELCAP. Available from Bonneville Power Administration,
Commercial Programs Branch.

Lawrence Berkeley Laboratory, 1980. DOE-2 Reference Manual,

Part 1. National Technical Information Service.

Lawrence Berkeley Laboratory, 1988. Technology Assessment:

Electronic Office Equipment. Available from U.S. Department of
Energy.
Bonneville Power Administration, 1991. TechnicalRequirements for
the Energy Smart Program. Available from Bonneville Power
Administration, Commercial Programs Branch.

Kaplan Engineering, Portland Energy Conservation Inc., 1992.

Energy Edge East Idaho Credit Union Simulation Tuning. Available
from Bonneville Power Administration, Commercial Programs
Branch.

Battelle Pacific Northwest Laboratories, 1989. Resolution of ELCAP

Metered End-Use Data and Regional Commercial Building
Prototypes. Draft report for Northwest Power Planning Council.

Lawrence Berkeley Laboratory, 1992. Energy Edge Impact

Evaluation, Middle Overview. Available from Bonneville Power
Administration, Commercial Programs Branch.

Treado, S.J., and J.W. Bean, December, 1988. The Interaction of

Lighting, Heating and Cooling Systems in Buildings - Interim
Report. NISTIR 88-3860, National Institute of Standards and
Technology.

Architectural Energy Corporation, February, 1992. Engineering

Methods for Estimating Impacts of Demand-Side Management
Programs - Draft Handbook.

Northwest Power Planning Council, May, 1991. Northwest

Conservation and Electric Power Plan.

SBW Consulting, Inc., November, 1990. Analysis of Commercial

Model Conservation Standards Study. Conservation Analysis
Summary. Volume 1. Available from Bonneville Power
Administration, Commercial Programs Branch.

Harris, J., J. Roturier, L.K. Norford, and A. Rabl. November, 1988.

TechnologyAssessment: Electronic Office Equipment. LBL-25558
Rev., Lawrence Berkeley Laboratory, Berkeley, California.

Piette, M.A., 1992. Personal communication.

Glazing (discussed on page 42):
We recommend that, unless the design specificalIy includes an ECM relating
to glazing and frame type, the modeler assume a double-glazed window with
a non-thermal-break aluminum frame. If the modeler does have knowledge
of the specific window types to be installed, then we recommend that she use
the manufacturer's rated window unit U-values (and shading coefficients) as
a first preference. If these are not availabie, or are suspect, then the
modeler should either refer to the ASHRAE Fundamentals Handbook
(1989), Chapter 27, Table 13, or use the Lawrence Berkeley Laboratory
computer program WINDOW 3.1. The modeler should also adjust shading
coefficients to account for opaque portions.

HVAC Equipment Capacity (page 82):

In design-phase modeling, the modeler should use the designer's selection
for the HVAC equipment capacity for as-designed building. She should use
the same equipment for the baseline building unless the output indicates
that increased capacity is necessary to meet loads. In the latter case she
should let the software program select the capacity required, being alert to
possible unrealistic sizing by the program.

Infiltration (page 41, 82):

The modeler should consider infiltration as a parameter to be held constant
between the as-designed and baseline buildings unless an ECM is designed
to directly decrease infiltration in a predictable manner.

We recommend that 0.038 C F M / ~is~a ~reasonable beginning assumption for

infiltration, but that the modeler consider the characteristics of the building
being modeled to determine its ultimate suitability. We further recommend,
if the software package being used has the capability, that the modeler split
the volume between wind-dependent and wind-independen t infiltration.

Insulation (page 81):

The modeler should input for the baseline the minimum insulation value
that both satisfies MCS and represents common construction practice.

Interior Lighting (page 56):

The following ratios of unoccupied to occupied weekday lighting usage
should be used unless building-specific data is available:

BUILDING TYPE UNOCCUPIED/ PEAK w/FT2

PEAK RATIO

Office 0.29 1.56

Retail 0.10 1.44

Grocery 0.58 2.00

Restaurant 0.14 1.67
Warehouse 0.06 0.68
The modeler may also wish to use the ELCAP values for peak power density
when she lacks more detailed design information.

Interior Walls (pape 48):

Model interior walls only if:

(1) the thermostat schedules and setpoints for the zones on either
side of the wall are not identical, or

(2) the zone has significant internal gains during periods when the
HVAC system is not operating, or

(3) an adiabatic wall is needed to properly simulate thermal mass.Do

not model an interior wall if the thermostat schedules for the zones
on either side of the wall are the same in setpoints and hourly
profiles.

Be sure not to input an interior wall under both adjacent zones. That would
result in duplicating the heat transfer.

Miscellaneous Equipment (page 53):

The modeler should only give the rated equipment capacities if the
simulation offers a load factor (averagefpeak load fraction), or calculates the
equipment load input hourly using an operating schedule profile that
permits fractional amounts.

An operating power density of 1.0 to 1.5 w/ft2 for office equipment and
personal computers is generally appropriate.

Results of the ELCAP study suggest that, as a general rule of thumb,

equipment consumption during unoccupied hours should be set at no less
than 30% of the peak load.

We recommend that the following values for unoccupied equipment fraction

be used during design phase modeIing unless there is specific design
information to support a different assumption:

BUILDING TYPE UNOCCUPIED/ PEAK w/FT2

PEAK RATIO

Office 0.40 1.50

Retail 0.30 0.76
Grocery 0.67 1.04
Restaurant 0.36 2.13
Warehouse 0.50 0.35

The modeler may also wish to use the ELCAP values for peak power density
when she lacks more detailed design information.
The modeler is advised to distinguish between large computers and small
computers when developing equipment load inputs, and to create a separate
zone and air distribution system in the simulation for any significant
computer rooms.

Occupancy Sensors (page 82):

The modeler should simulate the effect of occupancy sensor lighting control
by multiplying the relevant baseline lighting power densities by 0.7.

Operating Assumptions (page 83):

For new construction, the modeler should assume proper operation of all
building systems in both the baseline and as-designed models. For existing
buildings, the modeler should assume actual operation for the baseline
model.

Optimum Start (page 68):

Unless a simulation is explicitly modeling optimum start, either (1) input a
change in thermostat setpoint an hour or two prior to occupancy o r (2)
input a thermostat setting schedule that ramps up the heating setpoints prior
to occupancy.

Reality Checks (page 88):

The modeler should compare the end-use EUIs in the building baseline
simulation to the data in appendix 11. The modeler should troubleshoot the
simulation input whenever a simulated end-use EUI differs from the
corresponding EUI high or low range value by more than ~35%. If the
troubleshooting convinces the modeler that the input and the simulation are
correct, she should discuss and justify any such discrepancies.

Schedules (page 52):

We recommend that the modeler use the Building Schedule Percentage
Multipliers (appendix 111) as the first source for building schedules.
However, these should be modified per the following recommendations.

Supply Air ~ F ' M l f (pape

t ~ 74):
The modeler should input a minimum supply air CFM of no less than
0.3 CFM/~~'and a design supply air CFM of no less than 0.8 C F M / ~unless
~~
she has audit or design information that states otherwise.

Thermal Mass ( p a ~ 47):

e
The modeler should follow the DOE2 reference manual recommendations
concerning when to use custom weighting factors.

Tweaking (page 29):

Do not tweak any of the uncertain inputs before checking for careless errors.

Unconditioned Spaces (page 48):

If there is a liberal flow of outside air through the unconditioned space, (as
in most commercial garages), it is advisable to model the interior surface as
an exterior surface.
APPENDIX I U-VALUES
APPENDIX I1 ENERGY USE INDICES
APPENDIX I11 ASHRAE OCCUPANCY SCHEDULES
APPENDIX IV SAMPLE CONTROLS SEQUENCE OF OPERATIONS
APPENDIX V SAMPLE PERFORMANCE SPECIFICATIONS
APPENDIX VI SAMPLE CALCULATION SPREADSHEET
APPENDIX I U-VALUES
Table 3a
FRAMING R-VALUES
Wall Type of Framing Spacing Insulation R-Value Correction Factor Effective R-Value
Framing
Wood, 2 x2 16" O.C. 0.90 (Air Space) 112% 1.01
This table can be 5 75% 3.75
used to find the R-
Value for framing 7 65"/0 4.55
and framing cavity 9 57% 5.13
in a wall. The 24" O.C. 0.90 (Air Space) 109% 0.98
values from this 5 80% 4.00
table can be added
to the R- Values of 7 7 11
'0 4.97
other layers to 9 64% 5.76
calculafe the Rt of
the wall. Wood, 2 x 4 16" O.C. 0.90 (Air Space) 116% 1.05
wood framing 9 82% 7.38
values are based on 11 77% 8.47
20% and 15% 13
framing factors for
16 inches and 24 19
inches spacing. 0.90 (Air Space)
Metal framing
values are based on
Department of 13 77% 10.01
Energy, 10 CFR 19 66% 12.54
Part 435.
If the insulation is Wood, 2 x 6 16" O.C. 0.91 (Air Space)
compressed, 11
adjusted the
insulation R- Value
19
in accordance with 22
manufactures 30
recommendations. 24" O.C. 0.91 (Air Space)
Use the adjusted
insulation I?- Value 11
in this table. 19
22
30

Metal, 2 x 4 16" O.C. 0.90 (Air Space)

9
11
13
19
24" O.C. 0.90 (Air Space)
9
11
13
t9

Metal, 2 x 6 16" O.C. 0.91 (Air Space)

11
13
19
22
24" O.C. 0.91 (Air S p a c e )
11
13
19
22

Source: State of Oregon Code Compliance Manual

Table 3a (Cont.)
FRAMING R-VALUES
Roof/ Type of insulation Correction Effective
Ceiling Framing Spacing R-Va tue Factor R-Value
- - -.- - .- - -- ---
and Floor
Wood, 2 x 6 16" O.C. 11 94% 10.34
Framing 13 92% 11-96
This table can be 19 85% 16.15
used to find the R- 22 81% 17.82
Value for framing
and framing cavity
in a roof/ceiling or
floor. The values
from this table can
be added to the
R-Values of other
layers to calculate Wood, 2 x 8 16" O.C. 13
the Rt of the roof/
ceiling or floor.
19
22
Wood framing
values are based
on lo%,and 6%
framing factor for
16 inches and 24
inches spacing
without air spaces Wood, 2 x 10 16" O.C. 19
or compression of 22
insulation. R- 30
Value for wood is
based on fir, pine, 24" O.C. 19 96% 18.29
and similiar
22 95% 20.87
softwood.
30 91O h 27.37
If the insulation is - -- -- -- -.-. . - - -- -. ---. ..-....... --- -- --- --- ----
. --
compressed, 2 ,x 12
WOO~ 16" O.C. 22 95% 20.82
adjust the
insulation R- Value 30 90% 26.94
in accordance 38 85% 32.47
with manufactures
recommendations. 24" O.C. 22 97% 21.28
30 94% 28.09
Metal truss values
are based upon 38 91Oo/ 34.47
-.
. -- -- -- -- -.- - . . . ...---- .-- - .. - -
metal trusses with
4 ff spacing that Wood, 2 x 14 16" O.C. 22 97% 21.35
penetrate the 30 93% 27.83
insulation, and
0.66 in diameter
crossmembers
every 1 ft.

Metal truss
APPENDIX I1 ENERGY USE INDICES
NOTES ON ENERGY CONSUMPTION BY END-USE TABLE.
The table shows the range of electric energy use intensities (EUIs) by end-use for new commercial build-
ings in the Pacific Northwest. The values shown are based both on simulations of prototypical buildings
with characteristics typical of new buildings and on measured data (for samples of 2 to 17 buildings per
building type).
End-use EUIs vary greatly from building to building, even with buildings of the same type. This variation
can be explained by climate building characteristics (such as size, shell, and types of lighting and HVAC),
occupancy, and type of activities. Most of the data are based on all-electric buildings although some gas
and steam use was reported for some of the buildings. There are some unique charateristics of the data for
several building types:
Office. Data are from sources (1) (n=7 for the ELCAP data) and (2). Variations in the HVAC EUIs are
explained by differences such as climate, occupancy, building size, and equipment type. Differences in
heating are explained by differences among data sources. Prototypical buildings represent resistance heat
while the ELCAP sample and the forecast have a fraction of buildings that use heat pumps (heat pumps are
more efficient than electric resistance).
Retail. Data are from sources (1) and (2). As reflected in the table, the most important end-use in retail
space is lighting. Lighting use affects HVAC related end-uses. The range in heating EUIs also reflect the
fact that 7 of the buildings in the sample uses gas as their primary heating source.
Grocery. Data are from sources (2) and (3) (n=6). The high values for cooling (8 kBTU/sq.ft-yr) and hot
water (13 kBTU/sq.ft.-yr) end-use EUIs are from the ELCAP measured data. The low values are from
simulations (cooling had a value of 0.34 kBTU/sq.ft.-yr that was rounded up to 1.0 kBTU/sq.ft.-yr). The
large difference in the hot water EUI may be due to differences in kitchen use within groceries.
Restaurant. Data are from sources (2) and (3) (n=6). The wide range for some of the end-use EUIs in
restaurants is explained by the mix of fast-food and sit-down restaurants. Fast-food restaurants tend to be
more energy intensive due to relatively smaller seating area. The high value in heating EUI is from simula-
tions while the low value is from metered data (where half of the buildings had gas as their primary heating
source).
Warehouse. Data are from sources (2) and (3) (n=12). The range in warehouses end-use EUIs is small.
School. Data are from sources (2) (n=2) and (3) (n=2). The low heating EUI comes from metered data
and is explained by the fact that both buildings in the ELCAP sample use gas as their primary heating fuel.

Data for the table came from the following sources:

(1) Northwest Power Planning Council, "Northwest Conservation and Electric Power Plan", Vol. 11-Part
1, p. 356, May 1991. Data are from forecasts of 1989 construction and from a subset of ELCAP
buildings (post-1979) obtained from source (3). ELCAP EUIs have been adjusted to account for
non-electric heating. Forecasts are based on all-electric buildings only and on data from public utili-
ties.
(2) SBW Consulting Inc., "Analysis of Commercial Model Conservation Standards. Conservation
Analysis Summary", Final Report, Vol. 1, November 1990. Simulated data from prototypical build-
ings. All electric with resistance space heating and Seattle weather.
(3) Taylor Z.T., Pratt R.G.,"Description of Electric Energy Use in Commercial Buildings in the Pacific
Northwest. End-Use Load and Consumer Assessment Program (ELCAP)", Pacific Northwest Labor-
tatory, DOE/BP-13795-22, p. 356, December 1989. End-use metered data from Post-1980 buildings.
Both all-electric and fossil-fuel heated buildings are included in ELCAP. The EUIs have not been
adjusted to account for non-electric hating. Office buildings average less than 50,000 square feet.
APPENDIX I11 OCCUPANCY SCHEDULES
 't't't +'tu-
000 +Icy. 000 't't't
0 0 0 0 0 0
m o o 000 't'tY-
N Icy.+

I%% C0 C C
0 0
000 0 0 0
c
000
N
0 0 0
C'tu- m o o 000
-
-
0 -0
I c0

c:% C C C
0 0 0
000
M M
000 000
N
C+Ic
o '0
t ' t0
m o o
C-
000
M N

esg C C C
0 0 0
o m m
99
000
f-
000
M
CIcu- goo 000
MN
O 0
r *0

Kg% H i 5 000 000

M
000
M
CIcIc
o +0
'+ 0
000
N
000
VIN-

C C C 000 m o o 000 C CIc 000

kc:% 0 0 0 Ch- m 0 0 -0 N
000
m 9 N
m o m C C C m 00 m o o o m 0 C C r 000 000
k m 9 0 0 0 Chr 0 o r 4- rcme
C C C m m o o o m 0 C c't
kcS3 0 0 0
00
0.7 0.- C o't
000
M-
000
CQmU
C C C m o o m o o o m 0 C c't
Kg% 0 0 0 0.7 0.- 0 OIc
000
M-
000
IC-03-

CZ3 sE5 m o o m o o
0.-
o m 0
0.-
C C't
0 o r
o m 0
m-
000
,mu
C C C m o o m o o o m 0 C C't m m o
ICE% 0 0 0 *M cO- 0 O '0
t VI-
000
bcO4
m o o C C C m o z m o o 000 C CIc m o o 000
,VIM 0 0 0 M N 4.M ChM 0 O '0
t .4fN L03N

e:z C C C
0 0 0
m m m m o o
ChM
000
ChM
C
0 o
cu-
r
m m o
M-
000
ma^

C C C m m m C CU- m o o
KZZ 0 0 0 0 o't M N
000
Lnm-
000 C C C 000 C CIc 000 000
U M M 0 0 0 0 0- M- N N

-
0
000 C C C 000 t Cu- m o o 000
4'MM 0 0 0 E g o 32" 0 0 - r r 7

--
0
000 C C C 000 000 Q O O c c't 000 000
u 0 0 0 0 O't
0
000 CIc't 000 000 000 't't't
0 -0
'40 Ic + Y-
0 0 0
000 't+'t 000 't't't
.-
3
'tu-u-
0 0 0
'4-Y-Y-
0 0 0
rn 4 ooo 000 '+ICY- 000 Ic'Cr
't 't 4
'- '4-'I-'+
0 0 0 0 0 0
000 't'tIc 006 +'t+
'4-Y-'4- +r't
0 0 0 0 0 0
000 IcIcIc 000 '4-'tY-
'4-Y-Y- 't't't
0 0 0 0 0 0
000 'CIcIc 000 't'tu-
't't'b 't't't.
0 0 0 0 0 0
r r r r
64 r W64L Wr 64 >.
t s L m
64 r
x3u
at' C
3-w
aJWC
Y 3a J$4
WC
S~ 5l 4
cr
a m 3 aJm3 41 (0 3 o m 3
3 m m 3 m m B m m 3 m m
C
0)
.C

C,
C U
rn aJ a
.r-.I
0 2 U
a
-
>s+J
n a
m
X
d

9
2
aJ
E u
QJ
cn aJ
V) E
E .-
't
cn
3
' 't-
U d 4 o
 Table A3-1 (cont.)
B u i l d i n g Schedule Percentage M u l t i p l i e r s

Hour
1 2 3 4 5 6 7 8 9 10 I?12 13 14 15 16 17 18 19 20 21 22 23 24

Retail Lighting Ueekday 0 0 0 0 0 0 0 20 50 90 90 90 90 90 90 90 90 90 60 60 50 0 0 0

& Receptacle Saturday 0 0 0 0 0 0 0 10 30 60 90 90 90 90 90 90 90 90 50 30 30 10 0 0
Sunday 0 0 0 0 0 0 0 0 0 1 0 4 0 4 0 6 0 6 0 6 0 6 0 6 0 4 0 2 0 0 0 0 0 0

R e t a i l HVAC Weekday o f f o f f o f f o f f o f f o f f on on on on on on on on on on on on on on on o f f o f f o f f
Saturday o f f o f f o f f o f f o f f o f f on on on on on on on on on on on on on on on on o f f o f f
Sunday o f f o f f o f f o f f o f f o f f o f f o f f on on on on on on on on on on o n o f f o f f o f f o f f o f f
Retai 1 SWH Weekday 0 0 0 0 0 0 0 10 20 30 40 55 60 60 45 40 45 45 40 30 30 0 0 0
Saturday 0 0 0 0 0 0 0 15 20 25 40 50 55 55 45 45 45 45 40 35 25 20 0 0
Sunday 0 0 0 0 0 0 0 0 0 10 25 30 35 35 30 30 35 30 20 0 0 0 0 0

Uarehouse Weekday 0 0 0 0 0 0 0 15 70 90 90 90 50 85 85 85 20 0 0 0 0 0 0 0
Occupancy Saturday 0 0 0 0 0 0 0 0 2 0 2 0 2 0 2 0 10 10 10 10 0 0 0 0 0 0 0 0
Sunday O O O O O O O O O O O O O O O O O O O O O O O O

Warehouse L i g h t i n g Weekdays 0 0 O 0 0 0 0 40 70 90 90 90 90 90 90 90 90 0 0 0 0 0 0 0
& Receptacle Saturday 0 0 0 0 0 0 0 0 10 25 25 25 10 10 10 10 0 0 0 0 0 0 0 0
Sunday 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

WarehouseHVAC Weekday o f f o f f o f f o f f o f f o f f o f f on on on on on on on on on o n o f f o f f o f f o f f o f f o f f o f f
Saturday o f f o f f o f f o f f o f f o f f o f f o f f on on on on on on on on o f f o f f o f f o f f o f f o f f o f f o f f
Sunday o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f

Warehouse SWH Weekday 0 0 0 0 0 0 0 5 25 35 35 45 55 40 35 40 15 0 0 0 0 0 0 0

Saturday 0 0 0 0 0 0 0 0 0 1 0 1 0 1 5 0 0 0 0 0 0 0 0 0 0 0 0
Sunday 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

School Occupancy Weekday 0 0 0 0 0 0 0 5 75 90 90 80 80 80 80 45 15 5 15 20 20 10 0 0

Saturday 0 0 0 0 0 0 0 0 10 10 10 10 10 0 0 0 0 0 0 0 0 0 0 0
Sunday O O O O O O O O O O O O O O O O O O O O O O O O

School L i g h t i n g Weekday 0 0 0 0 0 0 0 30 85 95 95 95 80 80 80 70 50 50 35 35 30 30 0 0
Saturday 0 0 0 0 0 0 0 0 15 15 15 15 15 0 0 0 0 0 0 0 0 0 0 0
Sunday O O O O O O O O O O O O O O O O O O O O O O O O
 Table A3-1 (cont.)
B u i l d i n g Schedule Percentage M u l t i p l i e r s

Hour
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2

School HVAC Ueekday o f f o f f o f f o f f o f f o f f o f f on on on on on on on on on on on on on on on o f f o f f

Saturday o f f o f f o f f o f f o f f o f f o f f o f f on on on on on o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f o f f
Sunday off off off off off off off off off off off off off off off off off off off off off off off off

School SWH Weekday 0 0 0 0 0 0 0 5 30 55 60 70

Saturday 0 0 0 0 0 0 0 0 0 0 0 0
Sunday 0 0 0 0 0 0 0 0 0 0 0 0

Hotel/Motel Weekday 90 90 90 90 90 90 70 40 40 20 20 20
Occupancy Saturday 90 90 90 90 90 90 70 50 50 30 30 30
Sunday 70 70 70 70 70 70 70 70 50 50 50 30

Hotel/MoteL Weekday 20 15 10 10 10 20 40 50 40 40 25 25
Lighting Saturday 20 20 10 10 10 10 30 30 40 40 30 25
& Receptacle Sunday 30 30 20 20 20 20 30 40 40 30 30 30

Hotel/Motel HVAC Weekday on on on on on on on on on on on on

Saturday on on on on on on on on on on on on
Sunday on on on on on on on on on on on on

Hotel/Motel SWH Weekday 20 15 15 15 20 25 50 60 55 45 40 45

Saturday 20 15 15 15 20 25 40 50 50 50 45 50
Sunday 25 20 20 20 20 30 50 50 50 55 50 50

Restaurant Ueekday 15 15 5 0 0 0 0 5 5 5 20 50
Occupancy Saturday 30 25 5 0 0 0 0 0 0 5 20 45
Sunday 2020 5 0 0 0 0 0 0 0 1 0 2 0

Restaurant Ueekday 15 15 15 15 15 20 40 40 60 60 90 90
Lighting Saturday 20 15 15 15 15 15 30 30 60 60 80 80
& Receptacle Sunday 20 15 15 15 15 15 30 30 50 50 70 70

Restaurant HVAC Weekday on on on o f f o f f o f f o f f on on on on on

Saturday on on on o f f o f f o f f o f f o f f o f f on on on
Sunday on on on o f f o f f o f f o f f o f f o f f o f f on on
 Table A3-1 (cont.)
Bui l d i n g Schedule Percentage Mu1t i p l iers

Hour
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Restaurant SWH Ueekday 20 15 15 0 0 0 0 60 55 45 40 45 40 35 30 30 30 40 55 60 50 55 45 25

Saturday 20 15 15 0 0 0 0 0 0 50 45 50 50 45 40 40 35 40 55 55 50 55 40 30
Sunday 25 20 20 0 0 0 0 0 0 0 50 50 40 40 30 30 30 40 50 50 40 50 40 20

Health Occupancy Weekday 0 0 0 0 0 0 0 10 50 80 80 80 80 80 80 80 80 50 30 30 20 20 0 0

Saturday 0 0 0 0 0 0 0 10 30 40 40 40 40 40 40 40 40 10 10 0 0 0 0 0
Sunday 0 0 0 0 0 0 0 0 5 5 5 5 5 5 5 5 0 0 0 0 0 0 0 0

HeathLighting Ueekday 0 0 0 0 0 0 0 50 90 90 90 90 90 90 90 90 90 30 30 30 30 30 0 0
& Receptacle Saturday 0 0 0 0 0 0 0 20 40 40 40 40 40 40 40 40 40 40 10 0 0 0 0 0
Sunday 0 0 0 0 0 0 0 0 10 10 10 10 10 10 10 10 0 0 0 0 0 0 0 0

Health HVAC Weekday on on on on on on on on on on on on on on on on on on on on on on on on

Saturday on on on on on on on on on on on on on on on on on on on on on on on on
Sunday on on on on on on on on on on on on on on on on on on on on on on on on

Health SWH Ueekday 0 0 0 0 0 0 0 1 5 5 5 6 5 7 5 8 0 7 0 8 0 7 5 7 0 6 0 4 0 1 5 1515 5 0 0

Saturday 0 0 0 0 0 0 0 0 15 25 25 25 20 20 20 20 20 5 0 0 0 0 0 0
Sunday O O O O O O O O O 1 5 1 5 1 5 1 5 1 5 1 5 O O O O O O O O O

(a) Reference: ASHRAE/IES Standard 90.1-1989, Table 13-3.

' (b) Table A3-1 contains m u l t i p l i e r s f o r converting the nominal values f o r b u i l d i n g occupancy, receptacle power density, s e r v i c e hot
water, and l i g h t i n g energy i n t o time s e r i e s data f o r estimating buikding Loads under t h e standard c a l c u l a t i o n procedures.

For each standard b u i l d i n g p r o f i l e t h e r e are three s e r i e s - one each f o r weekdays, Saturdays, and Sunday. There are 24 elements per
series. These represent the m u l t i p l i e r t h a t should be use t o estimate b u i l d i n g Loads from 12 a.m. t o 1 a.m. (series # I ) through
11 p.m. t o 12 a.m. ( s e r i e s element #24). The estimated load f o r any hour i s simply the m u i t i p l i e r from the appropriate standard
p r o f i l e m u t t i p l i e d by t h e appropriate value from the t a b l e s c i t e d above.

(c) The b u i l d i n g HVAC system schedule l i s t e d i n Table A3-1 l i s t s the hours when the HVAC system s h a l l be considered ON o r OFF i n
accordance w i t h Section 13.7.3.2 o f ASHRAE 90.1-1989.
APPENDIX IV SAMPLE CONTROLS SEQUENCE OF OPERATIONS
 ECONOMIZER SEQUENCE OF OPERATIONS
(For a Rooftop Packaged HVAC Unit)

T h i s c o n t r o l sequence i s more e l a b o r a t e than m o s t t h a t

are w r i t t e n . However, we b e l i e v e that this l e v e l o f
detail i s n e c e s s a r y i n o r d e r t o t i g h t l y s p e c i f y t h e ECM
a s funded. A l s o , a commissioning agent c o u l d e a s i l y
t a k e this c o n t r o l sequence and w r i t e a d e t a i l e d
f u n c t i o n a l test plan aimed a t performance v e r i f i c a t i o n .

Intended operation:

The purpose of the economizer is to use outside air for cooling whenever
possible to minimize compressor operating time. When Stage 1of the space
thermostat calls for cooling and the outside air enthalpy is below the
controller setpoint, the outside air damper will move toward the open
position. Simultaneously, the return air damper will move toward the closed
position. The outside air damper will admit a proportionate amount of air
necessary to satisfy the mixed air temperature setting of 55'F. If Stage 2 of
the space thermostat calls for cooling, second stage mechanical cooling will
cycle.

O n a call for cooling from the space thermostat, if the outside enthalpy is
above the setting on the controller, the outside air damper will remain in
the minimum open position. In this event, the space thermostat will cycle
the mechanical cooling equipment.

With power on the economizer motor, the outside air damper will always be
in the minimum air position during the heating season and whenever the
space cooling demand is satisfied. With no power on the economizer motor,
the spring in the motor will completely close the outside air damper.

The minimum outside air setting is determined by the adjustment of the

minimum position potentiometer located on the economizer motor.

During the night setback cycle, the outside air damper is completely closed.

The check-out procedure is intended to ensure: 1) that the economizer

controls and motor function properly; 2) that the dampers perform as
intended without binding; 3) that the mechanical cooling system is locked
out during Stage 1 cooling.
APPENDIX V SAMPLE PERFORMANCE SPECIFICATIONS
This specification is intended to be representative of
a reasonable level of detail in a typical performance
specification. However, it does not contain a l l
hardware, software, and execution direction that might
be required for a specific building project.

PERFORMANCE SPECIFICATION FOR

ENERGY MANAGEMENT AND CONTROL SYSTEM

1. SCOPE O F WORK

Install a DDC energy management and control system (EMCS) t o provide

night setback and optimum start for each of the building's hydronic heat
pumps.
The system shall control operation of each heat pump's supply fan,
compressor, and reversing valve as well as the central electric boilers,
building supply and exhaust fans, cooling tower and hydronic loop valves and
water pumps. Each heat pump shall be controlled on the basis of a user-
input occupancy schedule, user-input heating and cooling space temperature
setpoints for occupied and unoccupied periods, actual space temperatures,
and manual override status. Each of these inputs shall be tracked separately
for each heat pump.

2. HARDWARE

The EMCS shall incIude:

a central processing unit (CPU) capable of running all software and

functions listed in these specifications, and

a central Man-Machine Interface, with CRT, keyboard and dot matrix

printer, to permit input of schedules and setpoints, and to permit
review of actual conditions at each point or create a trend log
(history) of setpoints, schedules and actual conditions at any point on
the system, and

a manual override for each heat pump, located in the space

conditioned by that heat pump.

3. SOFTWARE

The software shall offer the option of assigning a unique occupancy schedule
and optimum start calculation for each heat pump. In addition, the software
shall be programmed as follows:

3.1 Occupied Mode. In the occupied mode, the heat pump supply
fans shall be run continuously to provide ventilation, and the heat
pump shall be operated to meet the occupied heating and cooling
space temperature setpoints for the space served.

3.2 Unoccupied Mode. During the periods a space is scheduled as

unoccupied, the EMCS shall operate the heat pump in the
EMCS PERFORMANCE SPECIFICATION, page 2

unoccupied mode, unless it is in the Override Mode, Optimum Start

Mode, or Heat Recovery Mode, as described below. In the
unoccupied mode, the heat pump supply fans shall be on only as
necessary to meet the unoccupied heating and cooling space
temperature setpoints.

3.3 Override Mode. Each heat pump shall have a manual override
switch located in the area it serves. If the manual override switch is
activated during a period that the space is selected to be unoccupied,
the heat pump operation will go into the occupied mode. The length
of the manua1 override interval shall not exceed 2 hours.

3.4 Optimum Start Mode. It shall be the intent of the optimum start
calculations to bring the heat pumps into warm-up, heat pump by
heat pump, at the latest possible time on a day-to-day basis, while still
meeting the occupied setpoints by the beginning of the scheduled
occupancy period, and while minimizing use of the boilers. Warm-up
shall begin no more than 5 hours before occupancy (except on
Monday mornings). In the Optimum Start Mode, the heat pump
shall be operated to take the space temperature from the unoccupied
temperature setpoint range to the occupied temperature setpoint
range.

3.5 Heat Recovery Mode. If the hydronic loop return temperature

drops below 65 degrees, the controls shall scan the heat pumps to see
if any of the zones is in the unoccupied mode and has an actual space
temperature higher than the occupied space temperature setpoint
range. If so, all such heat pumps shall be brought on in the cooling
mode until the occupied setpoint(s) is (are) reached. Only then shall
the boilers come on, if still necessary.

4. USER-INPUTS

The occupancy schedules input into the EMCS shall agree with the attached
list except where tenant occupancy schedules have changed.

The heating setpoint shall, in all cases, be no more than 63°F for
unoccupied periods, and no more than 70°F for occupied periods.

The cooling setpoint, for occupied periods, shall be no less than 75°F.

5. DESIGN DOCUMENTS

A h l l point list and specifications shall be drawn up prior to installation and

updated after installation to reflect as-built conditions. The specifications
shall give, at least, the control sequences for the boilers, cooling towers, loop
controls, and the heat pumps. A copy of the design documents shall be
provided to <utility name > (1) after a contract has been awarded and prior
to installation, and (2) after they have been edited to reflect as-built
conditions.
EMCS PERFORMANCE SPECIFICATION, page 3

6. OPERATOR'S MANUALS AND MANUFACTURER'S LITERATURE

The customer shall receive from the manufacturer a complete set of manuals
to permit new staff to learn how to operate the system properly, and a full
set of cut sheets describing the installed hardware.

7. TRAINING

The controls contractor shall provide two days of training to the building
operatorts). If the contractor does not have the expertise to conduct the
training himself, he shall subcontract to the manufacturer's representative
to conduct the training. The training shall be provided in two separate one-
day periods. The 1st period shall be conducted when the EMCS is fully
programmed and ready for normal operation. The 2nd period shall be at
least several days later, and shall be arranged per the operator's request.

8. COST PROPOSAL

To receive a <utility name > contract for this project, the customer must
send the <utility name> analyst a Cost Proposal containing at least the
following elements:

a description of the proposed system, including the total number of

points, a list of points for each heat pump, the number of heat pumps
served, and itemization of major components, with model numbers,

the controls sequence of operations,

the total installed cost, broken down into equipment, design, labor,
training, and overhead and profit, and a statement "This cost covers
an installation meeting all of the requirements of the <utility name>
Work Orders for this project, dated ",and

cut sheets for the proposed equipment.

(The point count shall only include physical points where either a physical
condition is measured, or the EMCS operates a physical device. The point
count shall not include calculated values.)

9. CODE COMPLIANCE

This project shall comply with all applicable codes. Where permits are
required, the installation shall be approved by appropriate building officials.
The building owner agrees to maintain compliance documents in their files
and have them available to review if requested.

10. PROJECT COMPLETION

<Utility name> payment is made after the <utility name> analyst has
confirmed, via documentation review and on-site inspection, that the project
meets these Work Orders.
EMCS PERFORMANCE SPECIFICATION, page 4

10.1 Documentation.

Prior to inspection, the customer shall send the <utility name>

analyst:

a signed, approved copy of any permit(s) required for the

project,

an invoice for the full amount of the installation,

a one-week trend log demonstrating that the system is operating

per Work Order,

cut sheets for the installed equipment (if different from cut sheets
in Cost Proposal), and

Work Orders signed by the installer and the customer and edited
by them to reflect as-built conditions.

10.2 On-site Inspection.

The <utility name> analyst will visit the site and confirm that (a) at
least one member of the customer's staff has a full understanding of
the system so that it can be operated properly, and (b) the installed
equipment meets the requirements of these Work Orders.

10.3 Changes.

If the installed project does not meet these Work Order

requirements, delays in payment may be introduced, and the funding
level may change. Approval for a change to the Work Orders is valid
only if it is made in writing by the <utility name> analyst.
This specification is significantly more detailed than
the EMCS specification .
This sometimes is warranted,
especial1y with relatively new technology, where
potential project pitfalls are numerous.

Electronic Ballast Specifications

BIDS

Bids shall give the number of units, the manufacturer and model number, as
well as a description of the number and type of lamps served (e-g., "2 each
F96T12 slimlines").

Bids shall include cut sheets.

EQUIPMENT

The ballasts shall be electronic.

Each ballast shall have a rated input wattage of no more than 75% of their
standard electromagnetic counterparts.

Each ballast shall be listed as U.L. Class P, or be equipped with an internal

fuse.

Each ballast shall have at least a 3 year warranty covering both equipment
and labor costs.

Total harmonic distortion (THD) of the input current shall be no more than
30% of the total input current. The 3rd harmonic shall be no more than
25% of the total input current.

Each ballast shall be able to withstand input power line transients as defined
in ANSI C62.41 (or IEEE Publication 587 Category A) without damage.

Each ballast shall tolerate line voltage variations plus or minus 10 percent.

The power factors shall be no less than 90%.

The ballasts shall be able to start and operate properly with the respective
lamps down to 50°F ambient air temperature.

Each ballast type shall be approved by the ballast manufacturer for use with
the number and type of lamps it is serving. That approval shall be indicated
by the presence of rated input wattage data (for the respective lamp/ballast
combination) in the manufacturer's standard literature.

A list of the compatible lamp combinations, and a wiring diagram, shall be

given on each ballast.
BALLAST PERFORMANCE SPECIflCATION, page 2

The ballasts shall be compatible with all controls controlling the new ballasts
at the time of the installation and inspection.

INSTALLATION

Each ballast shall be installed in accordance with the manufacturer's written

instructions.

The installer shall assure that the neutral wires serving the new ballasts are
large enough to carry the ballasts' triple (3rd, 9th, 15th, etc) harmonics
without creating unreasonable voltage drops.

PAYMENT

The <utility name > inspector will inspect the installation after receiving the
following paperwork:

an itemized invoice (giving number of ballasts and their

manufacturers and model numbers),

an as-built sketch or drawing showing where the ballasts have been

installed,

a copy of the approved electrical permit, and

cut sheets for any ballasts that were installed but not bid.

If the installation does not agree with the bid, funding levels may be
affected. <Utility name> change order approvals are only valid if made in
writing.
APPENDIX VI SAMPLE CALCULATION SPREADSHEET
 HAND CALCULATIONS FOR LOADS NOT INCLUDED IN SIMULATION
SAMPLE OFFICE BUILDING

DOMESTIC HOT WATER

(1 gal/person-day x 260 days x 333,056sf/150sf/person 84,564
x 8.33 x (110-50 deg F)) / 3412 Btuh/kW
EXHAUST FAN
3 0 hp x -746 x 0 . 8 ~ 3120 hrs =
GARAGE EXHAUST
20 hp x ,746 x 0 . 8 ~ 8760 hrs =
TOILET EXHAUST
0.25 hp x ,746 x 0 . 8 ~ 8760 hrs =

HYDRONIC LOOP PUMPS

40 hp x ,746 x 0 . 7 ~ 8760 hrs =
DHW PRESSURE PUMPS
4 hp x -746 x 0 . 7 ~ 8760 hrs =

GARAGE LIGHTS
35.-088 kw x 8760 hrs =
EXTERIOR LIGHTS
6.25 kw x 4380 hrs =
4.5 kw x 8760 hrs =

ELEVATORS:

10,272,000 kwh/yr total bldg x 2% =

TOTAL
% of total bldg energy

FILE: \a-canerstuf\handphoe.wkl
DOVBP-26683-2
May 1992
3C