Toward Zero Net Energy
Buildings: Optimized for Energy
Use and Cost
by
Carrie Ann Brown
Bachelor of Science in Architecture
Massachusetts Institute of Technology, 2004
Master of Science in Building Science
University of California Berkeley, 2007
Submitted to the Department of Architecture
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Architecture: Building Technology
at the
Massachusetts Institute of Technology
September 2012
@2012
Signature of Author:
Massachusetts Institute of Technology.
All rights reserved.
C
Department of Architecture
/(2*ugust 10, 2012
S
Certified by:
Leon R. Glicksman
Professor of Mechanical Engineering and Building Technology
Thesis Supervisor
/
Accepted by:
Takehiko Nagakura
Professor of Architecture
Chairman, Department Committee on Graduate Students
Thesis Committee
Leon R. Glicksman
Professor of Mechanical Engineering and Building Technology
Massachusetts Institute of Technology
Thesis Supervisor
John E. Fernaindez
Associate Professor of Architecture and Building Technology and Engineering Systems
Director, Building Technology Program
Massachusetts Institute of Technology
Thesis Reader
Luisa G. Caldas
Associate Professor of Architecture
Faculty of Architecture
Technical University of Lisbon
Thesis Reader
3
Toward Zero Net Energy Buildings:
Optimized for Energy Use and Cost
by
Carrie Ann Brown
Submitted to the Department of Architecture
on August 10, 2012 in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy in Architecture: Building Technology
Recently, there has been a push toward zero net energy buildings (ZNEBs). While there are
many options to reduce the energy used in buildings, it is often difficult to determine which
are the most appropriate technologies to implement. To reach zero energy, some designs
extensively rely on the use of photovoltaics (PV) to meet the building load, without first
exploring the benefits of deep energy efficiency measures.
To minimize energy use in a cost effective manner, a tool has been developed to help
compare distributed generation (DG) alternatives with energy efficiency measures early in
the design process. It was designed to be accessible to non-technical users and to allow them
to set up and run simulations in just a few minutes. The tool was built on top of Design
Advisor, which provides the capability to analyze a suite of energy efficiency measures such
as insulation, window type, schedules, and HVAC types, as well as green and cool roofs.
New modules that have been developed for Design Advisor include: heat pumps, absorption
chillers, PV, cogeneration, and cost. Using capital cost above baseline as the independent
variable, the tool outputs the net annual energy use and total cost (capital and energy) for
each case analyzed in the optimization. This allows the user to understand the range of
technologies and costs involved along the path from the basecase to a ZNEB.
Thesis supervisor: Leon R. Glicksman
Title: Professor of Mechanical Engineering and Building Technology
5
Acknowledgements
First of all, I would like to thank my advisor, Prof. Leon Glicksman. Leon has been a
mentor to me for over a decade and I am grateful to have received so much knowledge,
support, and inspiration from him over the years.
I thank my readers, Profs. John Fernandez and Luisa Caldas for all of their valuable
insights and feedback to the development of this project. Their fresh perspectives
and constructive advice were invaluable to me.
Thank you to my family and friends, who were always there with love and support.
I also want to thank my labmates, especially Jaime, Steve, and Brian for their tireless
help with Design Advisor. And again, Jaime, for her wonderful teamwork, support,
and coding expertise when we started this tool as a class project in 16.888.
Turning my MATLAB code into a working website simply could not have happened
without the generous programming assistance of Isaac Cambron.
Finally, I also want to thank the MIT Department of Architecture and the MIT
Portugal Program for funding my graduate education at MIT.
7
Contents
1
19
Background
1.1
Introduction . . . . . . .
1.2
Zero Net Energy (ZNE)
1.3
1.4
19
. . . . . . . . . . . . . . .
20
1.2.1
ZNE Definitions . . . . . . . . . . . . . . . .
21
1.2.2
Other ZNE Issues . . . . . . . . . . . . . . .
24
1.2.3
Project Scope . . . . . . . . . . . . . . . . .
25
Existing tools . . . . . . . . . . . . . . . . . . . . .
25
1.3.1
BEopt (Building Energy Optimization) . . .
26
1.3.2
O ptEPlus
. . . . . . . . . . . . . . . . . . .
26
1.3.3
G enO pt
. . . . . . . . . . . . . . . . . . . .
26
1.3.4
DER-CAM
. . . . . . . . . . . . . . . . . .
27
1.3.5
H om er . . . . . . . . . . . . . . . . . . . . .
27
1.3.6
EnergyPlus Multivariate Regression . . . . .
27
1.3.7
Analysis of existing tools . . . . . . . . . . .
28
/
Renewable Energy Sources
29
1.4.1
Distributed Generation (DG) Overview . . .
29
1.4.2
Brief History of DG . . . . . . . . . . . . . .
30
1.4.3
DG Benefits . . . . . . . . . . . . . . . . . .
34
1.4.4
DG Challenges: Technical and Regulatory
.
35
1.4.5
Overview of DG Technologies
. . . . . . . .
36
1.4.6
W ind . . . . . . . . . . . . . . . . . . . . . .
36
1.4.7
Solar: PV and Thermal
. . . . . . . . . . .
38
Distributed Generation
9
1.4.8
Combined Cooling Heating and Power (CCHP)
. . . .
38
1.4.9
Fuel C ells . . . . . . . . . . . . . . . . . . . .
. . . .
39
. . . . . . . . . . . . . . . . . . . . .
. . . .
40
1.4.11 Clean-Energy Finance Overview . . . . . . . .
. . . .
42
1.4.10 E ngines
2
Methodology
45
2.1
Design of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . .
45
2.2
Photovoltaic Energy Module . . . . . . . . . . . . . . . . . . . . . . .
48
2.2.1
Operating Temperature
. . . . . . . . . . . . . . . . . . . . .
48
2.2.2
Solar Radiation Intensity . . . . . . . . . . . . . . . . . . . . .
50
2.2.3
Geographical Variation . . . . . . . . . . . . . . . . . . . . . .
50
2.2.4
PV Module Validation . . . . . . . . . . . . . . . . . . . . . .
53
2.3
Combined Cooling, Heating, and Power Energy Module . . . . . . . .
54
2.4
Hot W ater . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
2.5
Cost M odule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
2.5.1
Energy Efficiency . . . . . . . . . . . . . . . . . . . . . . . . .
58
2.5.2
Distributed Generation . . . . . . . . . . . . . . . . . . . . . .
62
2.5.3
Energy Costs and Total Cost
. . . . . . . .. ... .. .. ..
64
Optimization Methods . . . . . . . . . . . . . . . . . . . . . . . . . .
65
2.6.1
Energy Simulation: MIT Design Advisor . . . . . . . . . . . .
65
2.6.2
Initial M odel
. . . . . . . . . . . . . . . . .. ... .. .. ..
66
2.6.3
Comparison of Optimization Algorithms . . . . . . . . . . . .
70
2.6.4
Latin Hypercube Monte Carlo . . . . . . . . . . . . . . . . . .
71
2.6
2.7
2.8
Building Model Emulator
. . . . . . . . . . . . . ... .. .. ... .
71
2.7.1
Training Runs . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
2.7.2
Initial Multivariate Regression . . . . . . . . . . . . . . . . . .
72
2.7.3
Multistep Multivariate Regression . . . . . . . . . . . . . . . .
73
2.7.4
Multicity Multivariate Regression . . . . . . . . . . . . . . . .
78
Interface Design
82
10
3
4
Results
87
3.1
Initial Runs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
3.2
Final CCHP Runs
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
3.3
Final PV Runs
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
3.3.1
San Francisco . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
3.3.2
Milwaukee . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
101
107
Summary of Results
109
5 Next Steps
6
5.1
ZNE Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.2
Hourly Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.3
Neighborhood Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.4
Distributed Generation Options . . . . . . . . . . . . . . . . . . . . . 110
5.5
Other Building Types. . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.6
Expanded Options
5.7
Retrofits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .111
. . . . . . . . . . . . . . . . . . . . . . . . . . . .111
113
Conclusions
115
A Regression Coefficients
11
12
List of Figures
1-1
World map weighted by population (Worldmapper.org) . . . . . . . .
29
1-2
World map weighted by CO 2 emissions (Worldmapper.org) . . . . . .
30
1-3
US CO 2 emissions. (recycled-energy.com).
. . . . . . . . . . . . . . .
31
1-4
Average efficiency of delivered electricity has been stagnant since the
50s (C asten 2009).
1-5
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
2/3 of the fuel used to generate electricity is lost as heat (recycledenergy.com ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
1-6
The declining cost of wind. . . . . . . . . . . . . . . . . . . . . . . . .
37
1-7
Global Clean-Energy Projected Growth (Clean Edge 2008). . . . . . .
44
2-1
The effect of cell temperature on the efficiency of a single silicon solar
cell (M alik 2003)
2-2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The effect of solar radiation intensity on the efficiency of a single crystal
silicon solar cell (M alik 2003)
2-3
49
. . . . . . . . . . . . . . . . . . . . . .
50
Annual variation of the relative conversion efficiency of c-Si modules
in Athens, Ispra, Wroclaw, and Loughborough (Huld 2008) . . . . . .
52
. . . . . . . . . . . . . . .
55
2-4
Overview of a CCHP system (Kong 2005)
2-5
Breakdown building energy use in commercial buildings (DOE 2009)
56
2-6
Wall insulation cost with increasing R-value. . . . . . . . . . . . . . .
58
2-7
Window cost with increasing R-value. . . . . . . . . . . . . . . . . . .
59
2-8
Roof insulation cost with increasing R-Value . . . . . . . . . . . . . .
60
2-9
Thermal mass cost with increasing thermal capacitance . . . . . . . .
60
2-10 HVAC cost with increasing SEER . . . . . . . . . . . . . . . . . . . .
61
13
2-11 Summary of Incremental Costs by Building Type (Avery 2011).
62
2-12 Installed cost trends over time, by PV system size (Ren2l 2007).
63
2-13 Net installed cost of commercial PV over time (Ren2l 2007). . .
63
2-14 CCHP installed cost vs. capacity. . . . . . . . . .
64
. . . . . . . . . .
2-15 Design Advisor process (Urban 2005). . . . . . . .
66
2-16 Emulator training runs .. . . . . . . . . . . . . . .
73
2-17 Multivariate regression: No splits . . . . . . . . .
74
2-18 Data is split and recalculated with new regression coefficients.
75
2-19 Data is split into tiers with overlapping divisions.
76
2-20 Multivariate regression: 3 splits. . . . . . . . . . .
76
2-21 Boston results:
Tier 1.
Design Advisor output in red, regressi on
prediction in blue .. . . . . . . . . . . . . . . . . .
2-22 Boston results:
Tier 2.
77
Design Advisor output in red, regressi on
prediction in blue .. . . . . . . . . . . . . . . . . .
2-23 Boston results:
Tier 3.
.. 77
Design Advisor output in red, regressi on
prediction in blue .. . . . . . . . . . . . . . . . . .
.. 78
2-24 Interface design: City, goals, variable, and baseline inputs . . . . . . .
83
2-25 Interface design: Cost inputs . . . . . . . . . . . . . . . . . . . . . . .
83
2-26 Model output: 25%, 50%, 75%, and 100% reduction from baseline goals
and variable results
. . . . . . . . . . . . . . . . . .
2-27 Model output: Net primary energy vs. capital cost
84
. . . . . . . . . .
84
2-28 Model output: Total cost vs. capital cost . . . . . . . . . . . . . . . .
85
2-29 Model output: Net primary energy vs. total cost . . . . . . . . . . . .
85
3-1
Simulation results for a low-rise commercial building in Boston. Assumed
$4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
3-2
. . . . . . . . .
Simulation results for a low-rise commercial building in Boston. Assumed
$4/W PV, $0.1520/kWh, 8% discount rate, 5 years . . . . . . . . . .
3-3
88
89
Simulation results for a low-rise commercial building in Boston. Assumed
$4/W PV, $0.1520/kWh, 8% discount rate, 10 years. . . . . . . . . .
14
89
3-4
Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . . . . . . .
3-5
Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . . . . . . .
3-6
98
Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years. . . . . .
3-9
96
Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
3-8
93
Simulation results for a low-rise commercial building in San Francisco.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
3-7
90
100
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . . 102
3-10 Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . . 104
3-11 Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years. . . . . . 106
15
16
List of Tables
1.1
Overview of Various Distributed Generation Technologies . . . . . . .
37
1.2
Summary of Major Differences of the Fuel Cell Types (EG&G 2004) .
41
1.3
Status of Renewables Technologies. Adapted from Renewables Status
Report 2005 and 2007 (REN21 2007). . . . . . . . . . . . . . . . . . .
43
2.1
Design Advisor inputs
. . . . . . . . . . . . . . . . . . . . . . . .
47
2.2
2010 2-4 Story Office Building Construction Costs (MREA 2010) .
57
2.3
Initial optimization design variables.
. . . . . . . . . . . . . . . .
69
2.4 Weather Inputs for 30 Design Advisor Cities. . . . . . . . . . . . .
79
2.5
80
Emulator error in various cities
. . . . . . . . . . . . . . . . . . .
2.6 Emulator error in various cities for inefficient buildings . . . . . .
3.1
Baseline low-rise commercial building .. . . . . . . . . . . . . . . . . .
3.2
Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . . . . . . .
3.3
95
97
Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years. . . . . .
3.6
92
Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
3.5
91
Simulation results for a low-rise commercial building in San Francisco.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
3.4
81
99
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
17
101
3.7
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years. . . . .
3.8
103
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years. . . . . .
A. 1 Regression coefficients for weather parameters
. . . . . . . . . . . . .
105
116
A.2
Regression coefficients for decision variables (1 of 2)
. . . . . . . . . .
117
A.3
Regression coefficients for decision variables (2 of 2)
. . . . . . . . . .
117
18
Chapter 1
Background
1.1
Introduction
Recently, there has been a push toward zero net energy (ZNE) buildings. While
there are many options to reduce the energy used in buildings, it is often difficult to
determine which are the most appropriate technologies to implement. To reach zero
energy, some designs extensively rely on the use of photovoltaics (PV) to meet the
building load, without first exploring the benefits of deep energy efficiency measures.
This research topic focuses on using optimization techniques to quantitatively
assess technology choices in the built environment. Design space exploration determined
a set of distributed generation (DG) and energy efficiency choices that can be effectively
implemented in Portugal and in the United States.
To minimize energy use in a cost effective manner, an online simulation tool has
been developed to help compare DG alternatives with energy efficiency measures.
Using capital cost above baseline as the independent variable, the tool outputs the
net annual energy use and total cost (capital costs and net present value of energy
costs) for each case analyzed in the optimization. This allows the user to understand
the range of technologies and costs involved along the path from the basecase to a
ZNE building.
The goal of this research project is two-fold:
1. This tool is meant to act as a sketch model early in the design process. It has
19
been designed to be accessible to non-technical users.
2. This tool was also used to explore several case studies to provide general recommendations
for moving toward zero net energy systems.
The framework for this tool is Design Advisor (http://designadvisor.mit.edu), an
online simulation tool that allows non-technical users to set up and run annual energy
simulations in just a few minutes. Design Advisor provides the capability to analyze
a suite of energy efficiency measures such as insulation, window type, and schedules
as well as green and cool roofs. New modules that have been developed for Design
Advisor include:
1. Heat Pumps
2. Absorption Chillers
3. Photovoltaics
4. Combined Cooling Heat and Power (CCHP)
5. Cost: Capital and Total
6. Model emulator: Approximates Design Advisors output using pre-calculated
multivariate regressions. This allows annual energy use to be calculated in a
fraction of the time.
A number of existing building design optimization tools do exist. However, this
research project will be unique in its ability to perform a multi-objective optimization
of buildings with user-defined design variables and parameters in a real-time online
setting.
1.2
Zero Net Energy (ZNE)
The concept of zero net energy buildings (ZNEBs) has been developing for several
decades, but it has become increasingly popular in the last few years. A number of
state and national targets have been set. For example, in the United States, in 2008
the Department of Energy set the goal for all new construction to be ZNE by 2030,
half of all commercial buildings to be ZNE by 2040, and all commercial buildings to
be ZNE by 2050 (Goldstein 2010). Similarly, according the Zero Net Energy Action
20
Plan, California aims for all new residential buildings to be ZNE by 2020 and all new
commercial buildings to be ZNE by 2030. In Massachusetts, the ZNEB Task Force
set the goal to "put the private sector on a path toward (1) broad marketability
of zero net energy commercial and residential buildings by 2020 and (2) universal
adoption of zero net energy practices for new commercial and residential construction
by 2030" (MA ZNEB Task Force 2009).
According to the 2011 Integrated Energy Policy Report (IEPR)
A ZNE building has zero net energy consumption... .The goal is to
minimize energy use as much as technologically possible through costeffective efficiency measures, and then generate the balance of the buildings
energy needs with onsite renewable electricity generation such as solar
photovoltaic systems or wind-driven electricity generators. The substantial
energy efficiency improvements built into ZNE buildings contribute also to
maintaining and improving the buildings comfort and functionality (CEC
2011).
As part of this study, we wanted to understand whether going all the way the
ZNE is the best choice right now, from a cost and energy perspective. We believe
this is a very strong goal and it certainly can point the building industry in the right
direction. But, its possible that given financial constraints, several buildings could
be designed to reach goals somewhere along the path for the same amount that it
would cost to have a single building achieve full ZNE. So, for the same cost, it could
be possible to save more energy across several low-energy buildings.
1.2.1
ZNE Definitions
Despite this increased attention, there is little agreement as to the exact definition of
ZNE. ZNE definitions generally fall into four categories centered around:
1. Site Energy
2. Source Energy
3. Cost
21
4. Emissions
Each of these categories has its own advantages and disadvantages, which have been
previously documented and summarized below (Goldstein 2010, Torcellini 2006).
Site Energy
One approach to ZNE is to look at site energy. Here, over the course of a year, a
building produces as much onsite renewable energy as it uses. One advantage of this
definition is that is it simple to quantify. In general, this approach assumes that the
building is connected to the grid, which can be used to balance energy. Extra energy
can be sold to and bought from the grid as needed (Torcellini 2006). At current levels,
this is not unreasonable to implement, but as more buildings approach ZNE, energy
storage systems may needed. For example, buildings may produce extra energy in the
afternoon when solar energy is plentiful, but buy energy from the grid at other times.
If photovoltaics reach high market penetration, the surplus supply would occur at the
same time, requiring storage to make that energy available at other times.
One disadvantage of looking at site ZNE is that it does not differentiate between
fuel types. Here, a unit of electricity and an unit of gas are interchangeable. This may
lead to technology choices that are less than optimal past the site level. For example,
an electric heater is "100% efficient", meaning that all of the electricity supplied to
the heater is converted to heat.
In contrast, gas furnaces typically see efficiencies
ranging from about 80-95%. Under the site ZNE definition, it would seem that all
heat should be electric. However, when source energy is considered, a typical grid mix
requires about 3 units of energy to provide 1 unit of electricity due to transmission
and distribution losses. So, really an electric heater is closer to 33% efficient when
considering source (or primary) energy. In contrast, gas transmission and distribution
losses are minimal, so a gas furnace's site efficiency is also representative of the source
efficiency.
The field has yet to reach consensus about how to appropriately deal with gas use
in ZNE buildings.
review:
For gas use, three options have surfaced through the literature
all electric buildings, offsetting gas use with excess PV production, and
22
ignoring gas use.
Source Energy
A second approach to ZNE is to consider the source energy, rather than the site energy.
In this case, over the course of the year, the building produces as much renewable
energy as it uses from a source energy standpoint. Here, there is a differentiation
between fuel types. However, this method becomes more difficult to quantify since it
requires detailed information about site to source conversions for the grid.
Cost
A third definition is look at zero net cost. In this case, the building can sell as much
energy as it buys from the grid. Like the site energy definition, zero net cost is also
simple to quantify. The main drawback to this approach is that zero net cost does
not necessarily correlate to decreased site energy use or emissions.
Emissions
The fourth main category of ZNE is to consider zero net emissions. Here, over the
course of the year, the building would produce enough renewable energy to offset
any energy bought from the grid. As with the source energy case, quantifying this
approach requires detailed site to source grid conversions.
Time-Dependent Valuation (TDV)
In the previous definitions, the time of use of energy is not explicitly considered.
Time-dependent valuation (TDV) is one approach to address this issue. With TDV,
"the value of electricity differs depending on time of use (hourly, daily, seasonally)
and the value of natural gas differs depending on season. TDV is based on the cost
for utilities to provide energy at different times" (CEC 2011). TDV also accounts for
differences in efficiency and emissions by considering the time of use.
Also noted in the 2011 IEPR,
23
While the ZNE idea is straightforward, translating the policy into standards,
guidelines, and incentive structures requires collaboration between agencies
and stakeholders. To maximize the alignment of ZNE with California
energy system reliability and policy goals, the Energy Commission recommends
the use of metrics that account for the societal value of energy, including
the critical impact of avoiding peak demand and the value of avoided
carbon emissions, and other energy system costs. These components are
well-addressed in the time-dependent valuation of energy concept used by
the Energy Commission for its efficiency standards and the CPUC for its
valuation of efficiency program savings (CEC 2011).
1.2.2
Other ZNE Issues
Neighborhood Density
Another issue to consider when looking at the future of ZNEBs is the role of neighborhood
density. At one extreme, creating denser neighborhoods with shared energy resources
may make it easier to achieve source ZNE. Looking at neighborhood scale distributed
energy allows for new options beyond building integrated resources, such as PV and
wind farms and larger scale cogeneration. However, higher density also means that it
may be more challenging for individual buildings to reach ZNE by themselves. This
is especially important in cities with taller buildings. The percent of commercial floor
area able to reach zero net energy decreases exponentially with the increase in the
number of floors (Torcellini 2006). This is because daylighting and solar potential
decreases while plugs loads increase relative to heating and cooling needs.
Transportation
However, denser neighborhoods can also allow people to travel less. When looking
at ZNEBs, defining building boundaries is still under debate. For example, should
denser neighborhoods be credited for decreased user transportation? And how can we
handle the introduction of electric vehicles (EVs) to the grid? On one hand, if charged
24
at home EVs can significantly add to the building load. On the other hand, EVs also
have the potential to provide energy storage as distributed generation increases.
1.2.3
Project Scope
As shown above, defining ZNE is a complicated issue that has yet to be resolved.
Moving forward, reaching a consensus of the definition will have interesting ramifications
on the way that ZNE develops.
Using TDV and accounting for the societal value of energy seems very promising.
However, for the purposes of this research project, we have decided that this is out
of scope. TDV is currently used in California, but values have not been developed
for other climate zones. Additionally, dealing with hourly data rather than annual
numbers would significantly increase the complexity of the tool.
Instead, for the development of our tool, we have chosen to look at annual net
source zero energy. We assume that net metering is available, so the grid can be used
for energy balances. To reach ZNE with gas use, we also assume that PV production
must offset any gas used in the building from a source perspective.
With this tool, users can begin to understand the effect that various technologies
have on both cost and energy use. While the user can find the cost optimal solution
for a ZNE building, it is important to note that they can also look for solutions at
any point along the path the ZNE. Current ZNE goals are very aggressive and will
likely change the way that buildings use energy. It is a strong target to move toward,
but at this point in time, from a cost perspective, the best solution may be to only go
part of the way to ZNE. Building one completely ZNE building can cost the same as
building several buildings that use 50% less energy than average and we have designed
the tool to look at this issue.
1.3
Existing tools
There are a number of existing tools with similar objectives and features such as
BEopt, OptEplus, GenOpt, DER-CAM and Homer.
25
1.3.1
BEopt (Building Energy Optimization)
This software, developed by NREL, is designed to find cost-optimal residential building
designs on the way to a ZNE building. The user selects from predefined discrete
options for various building technologies. For example, for a wall, the user can look
at options such as wood stud and double stud, and each option has a variety of
insulation levels. The user manually selects the desired technologies, which are then
used as an input for the automated optimization.
Energy savings are calculated from a Building America Benchmark or a userdefined basecase. The software outputs the minimum building cost designs at various
energy saving levels. A "sequential search technique"is used to find the most cost
effective combination of energy efficient measures and PV (Christensen 2005). Specifically,
"efficiency options are employed until the marginal cost of saved energy for these
options equals the cost of producing PV .... From that point on, energy savings are
solely a result of adding PV capacity until ZNE is achieved" (Christensen 2006).
1.3.2
OptEPlus
OptEPlus is a tool developed by NREL to run parameterized Energy Plus runs. Here,
the user is able to modify parameters in an xml file, rather than individually altering
the EnergyPlus input files, which can be very time intensive. While this is not an
optimization tool, it is in the same family in that it can aid in the comparison of
numerous design options (Flager 2008).
1.3.3
GenOpt
GenOpt is a generic optimization program developed at LBNL that has implemented
a number of optimization algorithms such as generalized pattern search and particle
swarm. It is a stand-alone optimization program that is designed to be coupled with
any simulation program that requires a text input. As in most building simulation,
GenOpt is designed to work with programs where the derivative of the cost function
is not available. GenOpt can handle both continuous and discrete variables and some
26
constraints (Wetter 2004). In addition to the existing library of algorithms, custom
algorithms can be developed by the user and run with the interface.
1.3.4
DER-CAM
The Distributed Energy Resources Customer Adoption Model (DER-CAM) is an
economic model that helps users find the least cost DG and/or CHP option for their
site (in combination with the grid) (Siddiqui 2005).
With distributed generation
(DG) power is generated with numerous smaller plants at a building or neighborhood
scale, rather than having large central plants. Various DG technologies are explained
in more detail below.
1.3.5
Homer
HOMER, developed by NREL, is an optimization model for distributed power. A
number of supply options can be modeled, such as solar photovoltaics, wind turbines,
microturbines, and fuel cells. In addition, there are a number of storage options such
as battery banks and hydrogen.
This tool can be used to find the least cost option to meet electrical and thermal
loads with sensitivity analysis.
However, this tool can only help to make choices
between distributed generation choices and cannot optimize the design to reduce the
thermal load.
1.3.6
EnergyPlus Multivariate Regression
A recent Building and Environment paper by Hygh et al. utilizes a methodology very
similar to our work (Hygh 2012). Like our project, they have been developing a way
to quickly perform energy modeling early in the design process. In both Hygh's work
and the work developed in this dissertation, datasets of full energy simulations were
used to create regression equations used by emulators (also known as meta-models)
to quickly approximate the full simulation.
27
Hygh et al ran a series of EnergyPlus simulations generated from a Monte Carlo
sampling. Using this dataset, they then created a regression model to predict building
energy use. This methodology was implemented for four US cities. In our methodology,
we performed a full factorial analysis to create a dataset from Design Advisor runs in
seven US cities. Then, regression analysis was also performed on this dataset to create
an emulator to quickly predict energy use. This emulator is used in a multiobjective
optimization that uses Monte Carlo sampling to explore the solution space.
In contrast to the Hygh approach, in our model, we used five weather parameter
inputs for the regression analysis. So, our regression equations can be used in multiple
locations, rather than being tied to one city. A more thorough explanation of this
process is included in the methodology.
1.3.7
Analysis of existing tools
While the existing tools are quite useful in certain applications, none of them has
successfully balanced quickly optimizing both efficiency measures and distributed
generation alternatives in a simple, quick online setting.
BEopt is probably the
closest, but it is used for residential and solar applications with discrete choices
in a stand alone tool. Additionally, optimization runs typically take between one
and four hours.
OptEPlus could provide the framework for comparing efficiency
measures, but again the development of EnergyPlus models is time intensive, as well
as computationally expensive to implement. GenOpt is quite useful for aiding in the
optimization side of the problem, but still requires some other building model to run.
DER-CAM and HOMER extensively cover the distributed generation field, but do
not address the building efficiency side.
Therefore, a fast, online optimization tool that can be used by non-technical users
to balance energy efficiency measures with DG options in commercial buildings would
be a new addition to the field.
28
1.4
Distributed Generation
/
Renewable Energy
Sources
Design Advisor provided the framework for the efficiency measures included in the
tool, but it did not have any distributed generation (DG) or renewable energy options.
For our tool, we wanted to explore the effect of both DG and efficiency measures on
cost and energy use.
DG includes power production that is distributed, rather than being provided by
a central plant. DG encompasses many renewable energy sources, such as wind and
solar, but also includes non-renewable options, such as gas-powered engines.
A significant literature review was conducted to determine which technologies to
include in the new ZNE design tool. The following section provides a brief history of
distributed generation and an overview of the technologies that were considered for
tool development.
1.4.1
Distributed Generation (DG) Overview
Despite some level of continued controversy, most of the scientific community believes
that we need to drastically reduce our CO 2 emissions. This is especially important in
the United States. As seen in Figures 1-1 and 1-2, when the world map is weighted
by population, the US represents a small portion of the map. However, when we look
at this weighted by CO 2 emissions, the US is one of the top offenders. It represents
only 4% of the worlds population, yet uses 30% of its energy.
Figure 1-1: World map weighted by population (Worldmapper.org)
29
Figure 1-2: World map weighted by CO 2 emissions (Worldmapper.org)
As seen in Figure 1-3, in 2005, 69% of these US emissions came from supplying
heat and electricity. Figure 1-4 shows that the average delivered efficiency of this
electricity has been stagnant since the 1950s, making little use of the available waste
heat. In fact, as seen in Figure 1-5, about 2/3 of the fuel used to generate electricity
is lost as heat in the US. It seems that one step toward generating our electricity in
a more efficient, less polluting way would be to harness this waste heat into useful
energy.
1.4.2
Brief History of DG
In the beginning of the 20th century, distributed generation was the norm rather
than the exception. However, with the advent of alternating current and large-scale
steam turbines, the electric power industry began to develop. "Technical advances,
economies of scale in power production and delivery, the expanding role of electricity
in American life, and its concomitant regulation as a public utility, all gradually
converged to enable the network of gigawatt-scale thermal power plants located far
from urban centers that we know today" (DOE 2007). During this time, some forms of
DG did develop. However, this was mostly in the form of backup power for facilities
that needed reliable power, such as hospitals.
In the wake of the energy crisis in the 1970s, the Public Utility Regulatory
Act (PURPA) 1978 was passed, and marked a turning point for DG. Section 210
introduced a new class of "qualifying facilities" and gave financial incentives to develop
cogeneration and small power production (DOE 2007).
30
PURPA also resulted in
US C02 Emissions Sources, 2005
Heat &Power = 69% of ALL fossil fuel
Cot emissions
Figure 1-3: US CO 2 emissions. (recycled-energy.com).
31
Pow industry
Effidenc
Puntal
Eff iincy
+
-0
50% -
09L
I gg ukO-qnt AM
aCmbbW MK &Wpow pints
Figure 1-4: Average efficiency of delivered electricity has been stagnant since the 50s
(Casten 2009).
Jf'
L,
oil
Figure 1-5: 2/3 of the fuel used to generate electricity is lost as heat (recycledenergy.com).
32
research in renewable resources, cutting down the cost of PV and wind.
"These
smaller-scale generation technologies challenged the established paradigm of the utility
industry (and many other industries) that previously relied on large-scale equipment
to produce economies of scale.... If non-utilities could produce power as cheaply as
could utilities, than the big power companies no longer deemed recognition as natural
monopolies. And if they were not natural monopolies, they no longer deserved special
status as noncompetitive entities that required regulation" (Hirsh 2006).
As the role of regulation was being questioned, the Gulf War highlighted the
cost of security of our energy supplies. The Energy Policy Act of 1992 "sought to
employ competitive forces to increase domestic fuel production and to improve the
efficiency of energy use" (Hirsh 2006). The next decade marked a period of massive
restructuring of the utility system. California was a leading state in this deregulating
process. However, the blackouts of 2000 and 2001 called into question this increased
competition and decreased control, slowing the restructuring process.
The Energy Act of 2005 authorized investment into DG projects and research and
made several revisions to PURPA 1978. While PURPA is credited with accelerating
renewable generation, it was also blamed for "requiring utilities to purchase expensive
power generated from often low quality sources" (Malmedal 2005). Under PURPA,
utilities were required to buy power from qualifying facilities at the "rate of their
avoided cost for producing this power" (Malmedal 2005).
The Energy Policy Act
of 2005 removed this requirement for new qualifying facilities.
Instead, they will
now sell power to the wholesale market (unless otherwise specified by State law).
In addition, the bill requires utilities to allow net metering (and smart metering
if requested) in addition to interconnection, based on IEEE Standard 1547: IEEE
Standard for Interconnection Distributed Resources with Electric Power Systems.
While this standard does provide some guidance, many interconnection details are
left to utility interconnection rules, which often vary from utility to utility.
Currently, there is about 200 GW of DG installed across the country. However,
most of this capacity is in the form of units for backup/emergency power.
33
1.4.3
DG Benefits
As outlined in The Potential Benefits of Distributed Generation and Rate-Related
Issues That May Impede Their Expansion, DG has a number of potential benefits:
1. Decreased emissions: When using CHP or REs, there are decreased emissions
as compared to the typical grid generation.
2. Increased electric system reliability:
DG can increase system reliability by
introducing more system redundancy. This is especially important to industries
that require uninterrupted service, such as biotechs, because short outages can
results in the loss of days of product. It is estimated that power outages and
quality disturbances cost American businesses $119 billion each year (Hinrichs
2002).
3. Reduction of peak power requirements: Utilities generally invest based on peak
demand. Since DG can help to reduce the baseline load, and thus the resulting
peak, the utilities can invest less in transmission, distribution, and generationfreeing up assets for other purposes.
Furthermore, by providing additional
power during peak periods, DG can help to reduce network congestion, one
of the causes of the 2003 blackouts (Hirsh 2006).
4. Provision of ancillary services, including reactive power:
Ancillary services
include voltage support, regulation, operating reserve, and backup supply.
5. Improvements in power quality: To avoid equipment damage and system downtime,
power quality is critical.
"In certain instances, DG can be used to address
power quality problems particularly when the systems involve the use of energy
storage, power electronics, and power conditionings equipment" (DOE 2006).
6. Reductions in land-use effects and right-of-way acquisition costs
7. Reduction in vulnerability to terrorism and improvements in infrastructure
resilience
34
1.4.4
DG Challenges: Technical and Regulatory
Although DG has numerous potential benefits, there are also a number of barriers that
have inhibited greater expansion. "The greatest impediments to DG technologies are
both physical and immaterial. Lingering monopoly rules continue to favor the design,
financing, construction, and use of large-scale generators. Managerial practices and
methods within the industry continue to rely on fossil-fueled technologies.
Even
more pervasive, deeply held social attitudes regarding a technological society, material
standards of living, and consumption motive consumers to use more electricity, but
also impede the creation of generators near population centers" (Sovacool 2008). It
should be noted that many distributed generation technologies do use fossil fuels as
well, but they can be used more efficiently than the current grid mix.
DG technologies tend to have a higher capital cost per installed kilowatt. However,
when you consider the levelized cost of electricity (total lifetime cost per kWh electricity
produced), many DG technologies are actually cheaper than the grid (Sovacool 2008).
DG systems can be "much more difficult to control for the same reason that it
is easier to drive a large bus with 100 people than to track or monitor 100 separate
automobiles" (Sovacool 2008). Connecting to the grid is one of the biggest technical
challenges that DG faces. "This interconnection challenge really constitutes a set of
related smaller problems: voltage control (keeping voltages within a certain range),
the balancing of reactive power (using power to regulate the flow of alternating
current to maintain proper grid synchronization), and safety (ensuring the adequate
protection of people working on the grid)" (Hirsh 2006).
Many utilities feel that adding DG to the grid would make control unnecessarily
complicated.
With a lack of motivating measures, they make it very difficult for
DG to connect to the grid through high fees for interconnection and stranded costs.
Furthermore, the process for interconnection is often quite complicated, with long
timelines and expensive safety studies, varying from utility to utility. DO already
suffers from high capital investment per kW. Adding more fees exacerbates this issue.
In addition, the Clean Air Act of 1970 grandfathered in old coal-fired plants, allowing
35
them to operate more cheaply than most new technology. This is one example of the
lingering monopoly rules that impede the increased use of DG.
With centralized generation, many consumers have an out of sight, out of mind
mentality. With electrical generation located away from high population centers, it is
easier to underscore the environmental cost of generation. Therefore, some customers
are less willing to seek alternatives (Pasqualetti 2000).
1.4.5
Overview of DG Technologies
A number of distributed generation technologies exist. The following sections provide
a brief overview of some of the technologies that were considered for development for
this tool. A summary of the fuel source, thermal efficiency, capacity, challenges, and
advantages of each technology is included in Table 1.1 and explained in more detail
in the following sections.
In the end, photovoltaics (PV) and combined cooling, heat, and power (CCHP)
were chosen to be developed into new modules for the tool.
Many factors were
considered, but the main determining issues were the ability to be integrated into a
building, cost, and availability.
1.4.6
Wind
Wind energy has experienced rapid growth in recent years, with cumulative global
installed capacity growing from 5,000 MW in 1995 to 80,000 MW in 2007 (Pernick
2008). During this period of rapid growth, the cost of wind power has also dramatically
decreased. These trends, which do not include standby fees, are illustrated in Figure
1-6.
However, the intermittent supply of wind is still a major drawback. In addition,
some of the windiest regions are also the least populated and least grid developed.
So, an expansion of our transmission network is necessary to bring wind power from
rural and off-shore areas (NCEP 2004).
36
Table 1.1; Overview of Various Distributed Generation Technologies
Thermal
Fuel
Capacity
Challenges
wind
NA
1kW-5MW
sunlight
7-17%
1W-10kW
Wind
PV
Comparative
Advantages
Efficiency
intermittence,
distance from
transmission
high, cost of
materials, low
capacity
wide availability of
fuel, declining costs
modularity,
integration into
architecture
factors
plant
Biomass
40%
20-50MW
low energy
density of fuel,
wide availability of
fuel
air pollution
natural gas,
hydrogen
Fuel Cells
1kW-100M
W
17-22%
limited use,
high cost
low emissions,
modularity, long
operating lifetime
Microturbines
natural gas,
digester gas
25-30%
varied
>50%
CHP
20kW-500k
W
low capacity,
high capital
cost, efficiency
light-weight and
compact, modularity,
long operating lifetime
cost, proximity,
matching
electric and
heating
demand
efficient, use for
waste heat
50-
30,000
40
24,000
230
18,000
: 20
12,0U0
6,0M
10 -
0
10
1990
installed Capacity
Cost
The Energy Foundatio, 204
Figure 1-6: The declining cost of wind.
37
1.4.7
Solar: PV and Thermal
The solar industry is growing rapidly. In 2004, it reached 1 GW of total manufacturing
output. This increased to 1.5 GW in 2005, 2 GW in 2006, nearly 3 GW in 2007, and
is projected to grow by more than 30% annually for the foreseeable future (Pernick
2008).
However, grid-connected solar continues to be more expensive than other
grid-connected forms of electricity. "The key challenge for solar PV systems remains
to be reducing costs through improvements in the materials used to manufacture
PV modules. For solar thermal systems, technology improvements, increases in scale,
and cost reductions resulting from higher production volumes are needed to accelerate
deployment" (NCEP 2004). However, in many cases solar thermal is a competitive
choice already. This may especially be the case in milder climates, where domestic
hot water represents a significant portion of the energy demand. Another advantage
of solar is its modularity and ability to be integrated into architecture, making it an
appealing choice for ZNE designs.
1.4.8
Combined Cooling Heating and Power (CCHP)
CCHP, also known as trigeneration, is a process in which one plant produces both
electricity and useful heat.
Typical power plants generate electricity and release
the waste heat. Since this waste heat is harnessed in cogeneration, primary energy
efficiencies are significantly improved.
CCHP can be installed with a variety of
electricity generating technologies, such as fuel cells and microturbines, which are
explained in more detail below.
Installing such a facility introduces a number of design challenges. For example,
the energy must be consumed close to where it is produced to use the heat. However,
this translates in additional efficiency increases because of the decreased transmission
and distribution losses as compared to the grid. Electric needs must be matched
to heating needs to be as efficient as possible. Some systems implement absorption
chilling or desiccant dehumidification to make use of the waste heat year round. In
addition, if installed within a building, some special structural needs may need to
38
address noise and vibration. Finally, users must come to an agreement with the local
utility provider, which often results is high standby charges.
Determining primary energy savings is a complicated process with cogeneration.
First, the efficiency of the electricity generation needs to be compared to that of
the grid. Secondly, the use of the waste heat should be calculated against traditional
heating and cooling systems. Finally, since the energy is in the form of both electricity
and heat, a simple first law analysis is not sufficient (although commonly cited).
Examples:
1. Biogen IDEC. Cambridge, MA: This biotech campus had decided to start
producing its own steam when the local supplier went bankrupt. However, for a
biotech, since short electrical outages can destroy 24-36 hour batches of product,
electrical reliability is very valuable. Since they were already building a new
building on campus, they decided that the need for reliability made cogeneration
a good solution. The capital cost of their plant was about $12 million and it has
been saving about $4 million and 25,000 tons of carbon annually (SourceOne).
2. Cogeneration in the University of Perugia, Italy (Bidini 2001): The facility was
sized to meet 80% of the electrical demand and 100% of the thermal demand.
Fueled by natural gas, the plant reduced carbon emissions by more than 20%
as compared to local alternatives.
1.4.9
Fuel Cells
"Fuel cells are electrochemical devices that convert chemical energy in fuels into
electrical energy directly, promising power generation with high efficiency and low
environmental impact" (EG&G 2004).
reductions (SO
2
For fuel cells, order-of-magnitude emissions
and NO 2 ) are projected as compared to the national average grid
(Zogg 2006).
For most combustion based DO applications, fuel costs represent over 70% of the
overall cost of generating electricity. Fuel cells have the potential for higher efficiencies
and have therefore been developed for DG applications (TIAX 2005). They also have
the potential to "be quieter, more reliable, and have lower maintenance costs than
39
most technologies used for DG" (Zogg 2006).
However, reaching favorable market
production and installation costs proves to be a challenge in fuel cell technology.
A number of fuel technologies exist, which are outlined in Table 1.2. Due to their
high operating temperatures, some fuel cells have potential to be used in cogeneration
to increase system efficiency. For the heating and cooling needs in our buildings,
temperatures close to 150 0 C would be useful, like the operating temperatures seen
in phosphoric acid fuel cells (PAFCs). In contrast in solid oxide fuel cells (SOFCs)
and molten carbonate fuel cells (MCFCs) operating temperatures may actually be
too high to be easily implemented, whereas polymer electrolyte membrane fuel cell
(PEMFCs) operating temperatures would be too low for cogeneration.
1.4.10
Engines
Engines include a wide range of DG technologies that typically use fossil fuels.
These can be separated into internal and external combustion engines. With internal
combustion engines (ICEs), combustion of a fuel in a chamber causes gas to expand.
The force from this expansion moves a piston or turbine, creating mechanical energy.
With external combustion engines, such as a Stirling engine, the combustion occurs
externally, rather than in a chamber and heat is passed through a heat exchanger.
Again, expansion creates mechanical energy.
ICE technology maturity, proven performance in DG applications, rapid startup
and shutdown, relatively high efficiencies, and low costs make them competitive with
newer technologies for many DG applications. However, DG technologies generally
require longer life, higher efficiency, lower emissions, and longer maintenance intervals
(Zogg 2007).
Current ICE technologies can meet or exceed grid efficiencies and
research is underway to increase efficiencies. CHP can increase this efficiency now,
but the installed cost increases 30 to 40%. ICEs reject about 50% of the recoverable
waste heat through exhaust steam and the other half through engine jacket cooling
systems.
Some of the higher temperature systems require a pressurized cooling
system, which usually limits cogen cooling to single-effect absorption or desiccant
regeneration. Research is also being conducted to lower emissions, however current
40
Table 2.1: Design Advisor inputs
Jaime Lee and Carrie
Brow
Master Table
16.88Q
Name
annual energy
annual heating energy
annualc iengy
annual lhung energy
primary emissions
'I.BUILDMi PIEOP~ERTICLIKKM
Region _
6 s.location
Cty
OCNBD
AClN8
EQUIP30off
2._
Symbol
1 Qtotal
2 Qheat
3
Qcool
4
Qlht
5 Emissions
a
Ocupancy Type
Units
Type
Objecnve
kWIgm2
k m2
kWhim2
kWhym2
/m2eve
USA, Europe, etc
Parameter
Parameter
Cty
begins
8
9
10
1
s.endHour
s._occVancyLoad
s.l.g.nLoad
s._egupmentLoad
ends
Person-densty
IUghting Requirement
Equpmmnt load
12 s.li ghngControol
3.V~ffWMrON SYSTEM
HVAC
13 s.yentllatIon pg
14
1
16
17
s.maxTempMax
s.mlnTempOccuped
s.maxTempUnoccupied
s.mlnTemp4unoapled
Occuled
Mn Occupied
Max Unoccup.
1en
Unoccup
USA
'MABoston"
Office Building, retail,
_etc
Parameter
0-24 hours
Parameter
0-24 hours
peop_/m2
lux
W m2
alwaysan', Ineffic!en
It', or *ecen
Parameter
Parameter
Parameter
Parameter
Occupancy Schedule
7 s.startHour
starting value notes
=Qheat+QooI+ght
*m
ical*_
esgn Vanable
eter
_
ree C
C
degree C
degre C
percent In decimal
Dgn Variable
Design Variable
Variable
Design Variable
Office Building sets default value for below
If endHour<startHour,
74occupancy is overniht.
If endHor= =startHour, no
20 occupancy
0.1
500 minimum lux
5
inefficent'=dims together,
inefficent' eftclent'=dlms ndependently
'mechanlca*
26 add 273.15 for degree K
20
2_8
18
1_8tsrelumidyflMax
Max RH
form
Parameter
0.6 0-1. 0=0%,
19 s-airChangecued_
20 s.alrChangesUnoccupid
4.
Air Change Rate
Total ACh Unoccupied
roomfuls per hour
roomfuls per hour
Parameter
Parameter
1.8
0
21 s.thermalMass
5. SIeMA1100t1'YPE
Thermal Mass
'high", 'low, or
zero"
Desen Variable
22 s.typeOfSm
23 sbuildingNSlength
24 s.buildlngEWlenth
Simulation TYe
N-S Length
E-W Lengh
'fouridedunmixed"
"one_slded", or
Parameter
four-sidedmixed"
Parameter
m
Parameter
m
'four sided-unmixed"
25
25
Parameter
Parameter
Parameter
5
5
3
I6.
0014 OTMEWIsWDS
25 s.roomwldth
26 s.roomoepth
27 s.roomfielght
Width
Depth
Heiht
m
m
M
iEML
norti,
'south","easr, or
25 snorentation
Vwidow Orientaton
29 s.windowArea
Glaing to wall ratio
WIndow lVpe (choose
o)
n
typology
"sgu
tnb*
'guinb"
32 s coaqM
ae pgDegth
33 so
8..sWALL
OLRiiaue
34 1s~insulationRiValue
No Winds
Gla
Overhang
W
>Wall
aVi
R-Value
west
%
_'nom
.
Design Variable
_
% of wall
10-9iW
ang
Desgn Variable
'dunb
'clear" or "low-e"
m
Deslen Variable
Desl Variable
"clear
0
I|(m2-K)/(W)
Design Variable
IISingle ae
Doubegalzed
47
rag:
-
1=100%
ICE DG systems do not meet California restrictions (Zogg 2007).
Diesel engines, a type of ICE, are generally more efficient and less expensive than
other ICE choices.
However, a major drawback is that the exhaust contains high
emissions and toxic air contaminants. So, they are often permitted to only run for a
limited number of hours (US EPA 2012). Advancements in biodiesel may make this
a better option in the future.
Microturbines, another subset of ICEs, are promoted as light-weight, compact,
low noise/vibration, low emissions, and fuel flexible compared to competing DG
technologies (such as reciprocating engines) (Zogg 2007).
Typical efficiencies of
microturbines on the market range from 23% to 26% (as compared to the natural grid
average of 31.5% in 2004). However, DOE's Advanced Microturbine Project aims to
reach efficiencies of at least 36% with reduced emissions.
But, for now efficiencies
of microturbines are only comparable to that of the grid when used in cogeneration.
Also, these efficiencies drop when the ambient temperature is above 59'F, which is
problematic during peak summer operation.
Power only microturbines cost about $2600/KW for 30 KW systems and about
$1800/KW for 100 KW systems. To widen microturbine applications, there will need
to be decreased cost and emissions as well as increased efficiency. However, they are
currently useful in applications where fuel flexibility or the light-weight, compact, low
noise/vibration features are needed (Zogg 2007).
1.4.11
Clean-Energy Finance Overview
Investment in renewable energy sources has been growing and is projected to continue
over the next decade, as shown in Figure 1-7. In addition, the cost of renewable energy
has been generally declining. A more detailed analysis of costs and trends is included
in Table 1.3 and explained in the cost sections of the Methodology.
42
Table 1.3: Status of Renewables Technologies.
Report 2005 and 2007 (REN21 2007).
Table 0: Status of Renewables Technologki.
Adapted from Renewables Status
Adapted from Renewables Status Report 2005 and
2007.
Status of Rnewables Technolos-Ch rct.,lk aendCosts
Technology
Power Generation
Typical Characteristics
Energy COst
(US
cents/kWh)
Cost Trends and Potential for Cost Reduction
Costs nave cecne Dy 12 -18% with each coubling of global
capacity. Costs are now nalf those of 1990. Turbine size
nas Increased from 600-000 kW a decade ago. Future
O-Shore Wind
Turbine size: 1-3 MW Blade
ciameter: 60-100 meters
SoLar PV (module)
Cell type ano efficiency: single,
crystal: 17%, polycrystalline:
15%, thin film: 10-12%
restrictions from sie optimizationt, improved blade/generator
design. and electronics.
5-8
tave oedined by 20% for each doubling of istalled
capacity or by aoout 5% per year Costs rose in 2004 due to
rriarket factors, Future cost reicalons oGe to materials,
design, process, effioency, and scale
Contimning dechires CL* to lower solar PV module Costs and
improvements In inverters and balance-of -system
comnponents.
Costs
Rooftop Solar PV
sdar kIsolation
Solar Ther mal Power
(CSP)
Hot Water/ Heating
Peak capacity: 2-5 kW
2,500 kWh/m2/year
1,500 kWh/m2/year
1,000 kWh/M2/year
Plant size: 1-100 MW Type:
tower, dish, troigh
20-40
30-50
50-80
(typical of Southern Europe)
[typical of UK)
Costs nave fallen from about 44 centslkWh for the first
plants in tre 19W0s. Future reductIons due to scale and
20-40 (trough) technology
Sixe: 2-5 m2; type: evacuatec
Soltar hot water/
Ieating
Costs stable of noderately lower due to ecotornmes of scale,
tube/flat-plate, Service: qot
Qludiv
improvemients
water, space heating
Rural (off-grid) Energy
Turbine size; 3- 100 kW
Small wind turbine
2-25
ntew naterials, laNer collectors, and
15-25
Mocerately declining witm technology advances,
Household wind turDtne Turbine size: 0.1 1 kW
System size; 10-1,000 kW;
Options: oattery backup or
15-35
Mocerately declining wttn technology aovances-
25-100
40-60
DechninQ with recucbons. in solar and winid component costs.
Declinin with reouctions in solar componenLt COsts.
Village scale mini-grid
ciesel
Solar home svstern
System sue: 20-1OW
43
Global Clean-Energy Projected Growth
$254.5
Total
Fuel Cells
$16.0
$74.0
Solar Pour
$834
Wind Power
$81.
Blofuels
$0.0
$50.0
$100.0
$150.0
$200.0
$250.0
($US Biions)
32007 U2017
Figure 1-7: Global Clean-Energy Projected Growth (Clean Edge 2008).
44
Chapter 2
Methodology
The following sections describe the methodology implemented throughout this research
project. The overall goal of this project was to create an online tool that would help
non-technical users understand the range of technologies and cost involved along the
path from a basecase building to a ZNE building.
To feasibly implement such a problem in an online environment, the scope of
variables considered needed to be limited. Design of experiments was used to find the
most influential variables to be used in the tool.
Next, a number of modules also needed to be developed, including a PV energy,
CCHP energy, and cost for DG and energy efficiency measures. Then, a variety of
optimization methods were explored to determine which should be implemented. To
perform calculations quickly enough for an online tool, a Design Advisor emulator
was developed. Finally, an interface was designed to allow the user to interact with
the new modules and optimization.
2.1
Design of Experiments
The source code for Design Advisor contains hundreds of variables to consider when
modeling a building. However, looking at all of them in an optimization problem
would be prohibitively expensive. In order to identify the most influential variables,
design of experiments was used.
45
We were unsure which variables would have the strongest effect, so a parameter
study was conducted first. Here, one variable is changed at a time with three or
four levels, leaving the other variables at their base level. For each experiment, we
calculated the annual energy use in kWh/m 2 . We then calculated the effect of each
level and the overall effect of each variable. These were then sorted by effect. From
these results, window to wall ratio and wall R-value were two of the most influential
variables.
Next, a one-at-a-time analysis was used, stepping through variables from the
highest to lowest effect.
optimal value.
At each successive step, the variables were left at their
Then, the variables with the highest effects were chosen to move
forward in the optimization process. Table 2.1 includes a list of the Design Advisor
inputs that were considered.
The efficiency measures chosen to be implemented in the optimization were:
1. Window typology: single, double, triple glazed
2. Glazing type: clear, low-e
3. Building length
4. Orientation: N/S, NW/SE
5. Wall insulation R-value
6. Roof insulation R-value
7. Roof type: bitumen, green, cool
8. Thermal mass: zero, low, high
9. HVAC: mechanical, joint
10. Overhang length
11. Lighting control: always on, dim independently
46
Table 2.1: Design Advisor inputs
lalme Lee and Carrie
Brown
16.-8M
Master Tate
a Symbol
1
2
~oa
ca
_Qhat
QpigQt
51 Emissions
'I. WULMILNG
PROEfM
4
6 s. location
2. _CCUPAWYEANU
annual energy
Jjdivarm2_
annualtealng_nergyJ
2
ann a i
_!!M~m2
_
annual lghngenhrgy_
2
~leneo
mar emisons
19 s.arChanesOccied
20 s.airChange~ocpe
4. TH-EUML 0M4W
_
__m_O
_
startine value notes
Qheat+Qool+Qliht
_
_
_
_
Parameter
Parameter
USA
'MABostn
beg1is
ends
Person-density
Uhtng Requrement
load
ro
Max Occuped
Ibn Occuped .
Max Unoccup.
Unoccup
bn
etc
Parameter
0-24 hours
Parameter
0-24 hours
pple/m2
lux
Wfm2
"alwayson*'Inefficien
tuojnln
or *elcient'
Parameter
Parameter
IParameter
Parameter
r
'mechanlcal
Design Variable
. Parameter
deree C
dege C
d gree C
degree C
percent in decimal
Max RH
Air Change Rate
foem
Total ACh Unoccupied
roomfuls per hour
Design Variable
Deslen Variable
Design Variable
Deinsribe1
tr
roomfuls per hour
Office Building sets default value for below
It endHour<startkoW,
7 occunpcy is ovenght.
It endHour==startHour, no
20 occupancy
0.1
500 minimum lux
5
lneffint'=dlms together,
inefficent' jefflclent'adms independently
*mechancal'
26 add 273.15 for dearee K
20
28
0.6 0-1. 0=0%, 1=100%
Parameter
TParaee
21 .therna[MaThetrlnal Mass
Design Variable
22 s.ypeOfSn
23 s.buildingm!ligth
'foursdedunmlxed'
, "one_sided", or
"foursidedmxed
m
Parameter
Parameter
m
Parameter
25
Parameter
Parameter
Parameter
5
S
3
25 s.roomWdth
26 s.roomDepth
27 s.roomHeiaht
Simulation Type
N-S Length
E-W Langh
idth
Depth
eht
m
m
north',
"south","eas", or
west'
28 s.orientation
Window Orientation
7WM0WDSCRIT0
29 s.windowArea
Glazingto wall ratio
%
Window 7pe (choose
30 stypogy __
su~b*
No Blinds
|Sinle
'Dule
'du. n~b
on
*tyn"__
__.........
_lective
'high-, '1w', or
'wro
24 s.buildinjng
ngith
6. ROOMOtNOMIONS
_
Office Building, retail,
12 s.lightngcnontrol
3. VENT1ILMOMN SYSTEN4
13 s.ventilationWe
HVAC
18 srelHurmjditya
_
Oty
City
7 s.startlour
14 s.maxTempcped
15 s.mInTempccupLed
16 s.maxTem pnoccupied
|.minTemnoccupied
_
Europe, etc
_USA,
Region
Occupancy Schedule
s.endlour
s.occupancytoad
s:it
s.eguwiEqetLoad
_
i
CL1E
4Occupancy Tpe
8
9
10
11
T e
[Oblecdve
Inite
Naeme
_Tnj
esgn Variable
Desg V
Design Variable
gazed
gae
"hjh"
'foursded unmixed"
25
'nort'
10-90 %of wall
*dgunb"
Ilze
32 s.coaona
33 s.overhang0epth
Glass Ctg
Overangm
'clear" or "low-e*
Desi Varable
Design Variable
*clear"
0
F34 snsulationRValue
I
Wal R-Value
(m2-K)/(W)
Deslon Variable
range: 1-7
47
2.2
Photovoltaic Energy Module
A photovoltaic (PV) energy module was developed to be added to Design Advisor.
This module is used to predict annual PV energy production. In this model, the
electricity provided by the PV cells is used for lighting, heating, and cooling. For
heating, an electric heat pump is used. The coefficient of performance (COP) for
both the chiller and heat pump was assumed to be 3. Typical ranges for COP for
both a chiller and heat pump range from about 2 to 4, so we picked a mid-range value.
Future work could explore this as a decision variable as well (ANSI/AHRI 2011).
Modeling PV energy output can be quite detailed, especially when the time of day
is considered. However, for the scope of this project, we were interested in the net
energy use per year. So, we wanted to build a simple model that could predict the
annual electricity production. When predicting PV output, two major geographical
parameters that need to be considered were temperature and solar radiation intensity.
This PV module outputs annual energy production given the system size and
basic weather data. It utilizes a method developed by Huld, Suri, and Dunlap in a
paper entitled, "Geographical Variation of the Conversion Efficiency of Crystalline
Silicon Photovoltaic Modules in Europe." This study inputs annual average daytime
temperature and total irradiation data to determine the relative PV module efficiency
for a given location (Huld 2008). In turn, this efficiency is used to predict the annual
energy output from the cells.
2.2.1
Operating Temperature
The open-circuit voltage, the short-circuit current, and the fill factor are the three
main parameters that affect the performance of p-n junctions in solar cells, and the
resulting efficiency and power output of the cells. The fill factor is equal to the
idealized power output divided by the product of the short-circuit current and the
open-circuit voltage.
"The short-circuit current (Isc) is not strongly temperature dependent. It tends
to increase with increasing temperature. This can be attributed to increased light
48
absorption, since semiconductor band gaps generally decrease with temperature. The
other cell parameters, the open-circuit voltage (Voc) and the fill factor (FF) both
decrease.The dominant variation is that of Voc. This causes the power output and
For silicon, the power output
efficiency to decrease with increasing temperature.
decreases by about 0.4 to 0.5% per 'C" (Green 1998).
Other studies also found that cell efficiency decreased linearly with increasing cell
temperatures, but they predict variation as low at 0.06% per 'C, as shown in Figure
2-1 (Malik 2003).
Graph tImpertur vs efialecy
65
U
a m
a
U-o
1:
80
0
N
N
a
a
a
a
U
a
a
U
a
a
S
a
3
4
5
6
7
8
9
t0
11
emdfncy (%
Figure 2-1: The effect of cell temperature on the efficiency of a single silicon solar cell
(Malik 2003)
49
2.2.2
Solar Radiation Intensity
As seen in Figure 2-2, efficiency increases linearly with decreasing solar radiation
intensity (although the overall power output still decreases). This is really because the
operating temperature increases with increasing solar radiation. In other words, "the
output power of a monocrystalline solar cell is reduced as irradiance is reduced but
at a different rate from the reduction of irradiance" because decreased solar radiation
intensity is swapped over by a decrease in operating temperature (Malik 2003).
Goo SOW radin0 hII1 0111 v 1
1200
nMy
1100.
900
U'
U
700-
1600
300
0
4
i
400
Figure 2-2: The effect of solar radiation intensity on the efficiency of a single crystal
silicon solar cell (Malik 2003)
2.2.3
Geographical Variation
In "Predicting annual PV output from yearly total irradiation and yearly average
temperature," Huld explores the effect of temperature and irradiation on PV output.
50
"It is found that the geographical variation in ambient temperature and yearly
irradiation causes a decrease in overall yearly PV performance from 3 to 13% relative
to the performance under Standard Test Conditions, with the highest decrease found
in the Mediterranean region. Based on the above results we developed a simplied
linear expression of the relative PV module efficiency that is a simple function of
yearly total irradiation and yearly average daytime temperature.
The coefficients
to the linear expression are found by fitting to the map resulting from the abovementioned analytical approach. The prediction of total yearly PV output from this
linear fit deviates less than 0.5% from the more detailed calculation, thus providing
a faster and more simplied alternative to the yield estimate, in the case when only
limited climate data are available" (Huld 2008).
Figure 2-3 shows the annual variation in actual efficiency for various European
locations. You can see that the efficiency is the lower in the summer months, when
the temperature is higher.
Based on this data, the simplified linear expression developed by Huld of the
relative PV module efficiency in a given location is:
T/rel
=
7relo
+
1(H
-
(2.1)
Ho) + A(T - TO)
where:
7reI is the relative efficiency
Treno
is the relative efficiency for a location with HO and To
0.912
= 4.02 x 10-6(kWh year-' m- 2 )-1
H is the total yearly irradiation in the plane of the PV modules
HO total global radiation at optimum angle = 1400kWh year- 1 m
A = -0.00242
2
0 K-1
T is the annual average daytime temperature
To = 16 'C
The relative efficiency, as compared to the efficiency under standard testing conditions,
51
a0M
0.9
0.0
Jon
Feb
.a
Ap
Bay
Jun
Jl
Aug
tep
Figure 2-3: Annual variation of the relative conversion efficiency
Athens, Ispra, Wroclaw, and Loughborough (Huld 2008).
52
Oct
Nw
De
of c-Si modules in
is a function of the module temperature and the in-plane irradiance. Ho, To, K, and
A are all constants derived from a previous study by Kenny et al (Kenny 2006).
The simplified linear expression for total energy produced over the year in a given
location is:
Etot
PnomH(Treo + K(H - Ho) + A(T - To))
(2.2)
where:
Etat is the total energy produced over the year
Pnom is the nominal power of the system in kW
These findings are especially useful for our model. We assumed that net metering
would be installed. Therefore, we only need to know the total amount of energy that
the cells can produce in a year and compare that to the annual energy requirements of
the building. From the weather files, we can determine H and T for a given location.
Therefore, we can directly calculate the annual energy output with the only variable
being the system size (correlated to Pnom in this system) for a given location.
2.2.4
PV Module Validation
To validate our PV module, which was based on Huld's equations, we compared
our results to NREL's PVWATT, a commonly used tool in the industry.
In our
comparisons, we assumed that the modules would be tilted and facing south.
In
Boston, for a 25 KW system, the Huld method predicted 31034 kWh annually, while
PVWATT predicted 30911 kWh. The error between the results is less than 0.1%. In
Los Angeles, the Huld model predicted 3.4% higher output. This is most likely due to
the fact that the Huld regression analysis was mostly conducted for cities in Europe,
where the solar radiation is lower.
The Huld paper also compared their results to PVGIS, a solar radiation database
based on meteorological data developed for the European Commission. This database
provides information about modeled solar radiation and PV generation.
Here, less
than 0.2% error was seen in Southern and Eastern Europe. In Western, Central, and
53
Northern Europe, errors ranged from 0.5-1%.
2.3
Combined Cooling, Heating, and Power Energy
Module
A combined cooling, heating, and power (CCHP) module was also developed. Here,
electricity is generated locally and the waste heat is used to heat and cool the
building. Figure 2-4 provides an overview of the process.
In the system, a gas
powered microturbine generates electricity that is used to power lights, plug loads,
and a chiller. The waste heat from generation is used for heating and cooling. On the
heating side, waste heat is run through a heat exchanger. This is supplemented by a
gas boiler to meet the heating load. On the cooling side, waste heat is run through
an absorption chiller to provide cooling. This is supplemented by an electric chiller to
meet the cooling load. The chiller can be run with electricity from the microturbine
or from the utility grid.
According to SOLAIR, for absorption chillers "the required heat source temperature
is usually above 80'C for single-effect machines and the coefficient of performance
(COP) is in the range from 0.6 to 0.8. Double-effect machines with two generator
stages require driving temperature of above 140 0 C, but the COPs may achieve values
up to 1.2.
In this model, we assumed 26% efficiency for the electrical generation (Zogg 2007).
To compare gas to electricity, 1 kWh equals 3412 Btus. At 26% efficiency, to generate
1 kWh of electricity, 13,137 Btus of gas is needed. From a first law analysis, at 26%
efficiency means that 74% of the gas energy is waste heat, or 9721 Btus.
In CCHP, a typical rate of waste heat is 7,000 Btu per kWh produced. Our model
produces slightly more waste heat because the microturbines are less efficient than
some of the larger engines. In our model, we assumed that 50% of this heat could
be used in heat recovery to heat and cool the building. For cooling, we modeled
absorption chillers with a COP of 0.8. Currently, if there is no heating or cooling
54
demand, the waste heat is not used. This may lead to a less efficient system in some
climates, especially in the spring and fall.
Utility Electricity
Natural Gas
Figure 2-4: Overview of a CCHP system (Kong 2005)
2.4
Hot Water
We considered adding a hot water module, especially for use in combination with
CCHP. However, according the the 2009 Building Energy Data Book, water heating
only represent about 2% of the total building energy use in commercial buildings.
Therefore, we decided not to include this in the new modules. Figure 2-5 shows the
breakdown of building energy use.
55
1000 BTU/sf
Space heating
32.8
Cooling
8.9
Ventilation
5.2
Water heating
2.0
Ughting
23.1
Cooking
0.3
Refrigeration
2.9
Office equipment
Computers
2.6
9.0
Other
Other
7%
Refrration Computers
3% -%\10~---~~~
Office eqwpment
Space heatin
sa34%
Cook
No-coo"n
Water heatg
2%
6.1
10%
Venlanon
s%
Figure 2-5: Breakdown building energy use in commercial buildings (DOE 2009)
2.5
Cost Module
Calculating the cost of a building is generally a complicated process involving factors
such as material cost, installation cost, predicted energy use and prices, discount
rates, etc. For the purposes of this tool, we wanted to create a simple cost function
that would illustrate the often opposing objectives of energy use and building capital
cost.
To get estimates of the additional cost of energy efficient materials in new commercial
buildings, we utilized the Database for Energy-Efficient Resources (DEER) and RSMeans
data, in addition to manufacturer prices. The DEER is a resource developed by the
California Public Utility Commission that provides information about the cost and
potential energy savings of various energy efficient measures (DEER 2011). RSMeans
is another resource that provides building cost information, including construction
and maintenance costs (Reed Construction Data 2010).
This model was developed to compare an energy efficient building to a basecase.
All of the cost equations are based on a change in a variable, rather than set prices.
For example, the cost of adding wall insulation is found by multiplying a constant
times the surface area times the change in R-value.
The cost module outputs both the capital cost above baseline and the total cost,
56
both in dollars per meter squared. The total cost includes the capital cost, as well as
the net present value of annual energy and maintenance costs. A discount level can
be set when the optimization is started.
According to data from RS Means the average cost of a 2-4 story office building
in 2010 was $176/sf or $1891/m
2
(MREA 2010). For reference, Table 2.2 outlines the
2010 construction costs of 2-4 story office buildings in major US cities.
Table 2.2: 2010 2-4 Story Office Building Construction Costs (MREA 2010)
Atanait
$153
$1,4
Baltim
Boston
Chicago
Clevaland
Dallas
Dener$165
$161
$205
$201
$172
$1.7
$2,2
$2.1
$1,352.
$148
it
D
"osteni
City
lansas
$179
$150
$17
$1,594
$1,7
$1,s
$,1
$1,913
i
Mia'innireapolis
ew Orleanis
New York City
Phiadelphia
Phoenim
Pittsburg
neles
orland
WW.LoI
San iego$180
anFrancisco
Seatle
Washington D.C.
57
$156
$195
$153
$231
$200
$154
$174
16
$175
$178
$1,9
$2,9
$1,6
$2M
$2,.15
$1,65
$1,1174
$$1,88
$1,911
$1.,3
$213
$183
$171
$2,29
$1,972
$1,24
2.5.1
Energy Efficiency
To predict the cost of efficiency measures, we looked at both industry prices and
gathered data from RS Means. For each measure, we looked at the additional cost as
compared to a baseline value.
Wall Insulation Cost
As seen in Figure 2-6, wall insulation cost increases with increasing R-value. These
values were gathered from RS Means data.
$/sqft vs R values
$9.00
9
9
9
$8.50
$8.00
*
9
1
1$7.50
* $/sqft
$7,00
999
$6.50
$6.00
I
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
Rvalues
Figure 2-6: Wall insulation cost with increasing R-value.
Window Cost
As seen in Figure 2-7, window cost increases with decreasing U-value. These values
were gathered from RS Means data and DEER.
58
Window $/sqft vs U-Value
$30.00
$25.00
$20.00
$15.00
$RMI Report
EER*
'7-
-DEER
$10.00
$5.00
$0.00
0.000
0.100
0.200
0.300
0.400
0.500
U-Values
0.600
0.700
0.800
0,900
1.000
Figure 2-7: Window cost with increasing R-value.
Roof Cost
For the roof options, we looked at increased insulation, green roofs, and cool roofs.
The current cost of extensive green roofs varies from $8-20 per square foot.
For
intensive green roofs, the costs vary from $15-25 per square foot.
For cool roofs, reflective acrylic paint costs about $0.75 per square foot. Another
option, PVC single-ply membranes cost around $3 per square foot.
As seen in Figure 2-8, roof insulation cost increases with increasing R-value. These
values were gathered from RS Means data and DEER.
Thermal Mass Cost
As seen in Figure 2-9, thermal mass cost increases with increasing thermal capacitance.
These values were gathered from RS Means data and Home Depot product information.
HVAC Cost
As seen in Figure 2-10, HVAC cost increases with increasing SEER. These values
were gathered from RS Means data and acforsale.com.
59
Ceiling Insulation
$2.00
-
$1.80
$1.60
$1.40
* RS Means
@DEER
$1.20
$1.00
$0.80
$0.60
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
R Value
Figure 2-8: Roof insulation cost with increasing R-Value.
Thermal Mass
$1.40
$1.20
51.00
$0.80
*RS Means
1WHonie Depot
$0.60
$0.40
$0.20
$0.00
0.00
0.20
0.40
0.60
0.80
1.00
Thermal Capacitance (Stu/F*st)
1.20
1.40
Figure 2-9: Thermal mass cost with increasing thermal capacitance.
60
AC unit cost vs. SEER
$1,800.
$1,600
$1,400
$1,200
$1,000
$800
adorsale.com
6
GDEER
a $600
$400
$200
so
$0
U
!
-$200
-$400
SEER
Figure 2-10: HVAC cost with increasing SEER.
Lighting Control Cost
In commercial buildings, lighting often represents the largest single energy use. Adding
lighting control can help to reduce this energy use. However, historically, the incremental
cost of advanced lighting systems has been a barrier to wider implementation. However,
the cost of electronics has been decreasing, so these technologies are becoming more
appealing.
"Controllable light sources enable a wide range of lighting controls strategies
capable of significant energy savings such as daylighting, scheduling, tuning, adaptation
compensation, workstation-specific control and manual dimming" (Avery 2011).
Figure 2-11 shows the incremental cost of dimmable lighting systems by building
type.
These represent the low-end estimate of the capital cost of the hardware,
installation, and commissioning. For offices, the low end estimate was $0.61/sf, while
the high-end estimate was $0.75/sf. For our tool, we used an average value of $0.68/sf
(Avery 2011).
61
Prototype
Measure cost per square foot
Office
$0.61
Retail (ASHRAE Type 1)
$0.56
Retail (ASHRAE Type 2)
$1.04
Retail (ASHRAE Type 3)
$2.00
Foodstore
$0.66
School
$0.65
School (Year-Round)
$0.65
Warehouse
$0.37
Hospitality
$0.85
Figure 2-11: Summary of Incremental Costs by Building Type (Avery 2011).
2.5.2
Distributed Generation
Photovoltaic Cost
PV system costs have dropped over the past decade. However, when grouped by
system size, the net installed cost has stayed relatively constant in recent years, as
seen in Figure 2-12. In commercial applications, we see similar trends, as illustrated
in Figure 2-13.
For our model, we focused on crystalline cells and did not consider thin film
technology. We assumed that the net installed cost with state and federal investment
tax credits (ITC) was $4/W and $8/W without tax credits. In addition, we assumed
that net metering would be installed, so any surplus would be sold back to the grid
at retail price. So, the electricity bought from the grid is equal to the total building
electricity usage minus the PV electrical production.
CCHP Cost
CCHP costs vary widely based on the generation type and the system size. Figure
2-14 provides an overview of the range of installation cost few varies generation types
and capacities. For the purposes of this tool, we were interested in building integrated
62
IA
I
$14
$12
$10
,M--
18
0-kW
$
I
--
I
$4
$2
1S.OkW
t0-100 kW
-ee- 100-800kW
---
1998
0kW
2000
logo
F Are 2-2sho:n Inst
2001
2004
2002
2003
Inseiandn Year
d ost endssnove
tiemaae y Pm
ssem csigae
Figure 2-12: Installed cost trends over time, by PV system size
-
I
II I
ntewnd cast
Gm0s
OC uteandor Fedo
-e ad Cos W TC
SateoWLuIY
H)
200
2005
(Rn
l
2007
2007).
(Ren21 2007).
w*
n. (dr-tux)
Nkimsied COst vo f'C
112
110
I
$4
$2
$0
198
.A5
0.1
is"
asio
OV 03WW
2000
2001
ts14
OA W
nE41
12 MiN
2003
2004
ftu168 a,3S8
52 MN 19 BH
n=619
2 MW
2002
inu
2006
2006
2007
nI107
nWOO n*1030
38. MN 511 WY 83 W
ftmn %o.r
Figure 2-13: Net installed cost of commercial PV over time (Ren2l 2007).
63
systems around 15 kW. For tool development, microturbines were chosen because they
are light-weight and compact, and they have low noise/vibration, low emissions, and
fuel flexibility. From a cost perspective, diesel engines would also be an attractive
option, but they have relatively higher emissions. For future applications, biodiesel
fueled engines may be a good choice.
CCHP Installation Cost vs. Capacity
400
-----
----
3000
2000
Ssag t in i
b
1500
1000
m rotivr nes
500
0
1
10
100
1000
10000
100000
capoar
tw
Figure 2-14: CCHP installed cost vs. capacity.
2.5.3
Energy Costs and Total Cost
The tool outputs the annual heating, cooling, and lighting source energy use. To
calculate energy costs, these are then converted back to site energy.
To calculate energy costs, the tool keeps track of gas and electricity that is bought
from the grid. For example, in the case of a building with CCHP, the tool will
determine how many therms of gas were needed to run the microturbine, in addition
to how much gas and electricity was needed from the grid to supplement onsite
production.
64
User supplied gas ($/therm) and electricity prices ($/kWh) are then use to calculate
annual energy costs. Finally, the net present value of the energy costs are calculated
over a 30 year period with a user defined discount rate.
The net present value of the of energy costs are then added to the capital cost
above baseline of efficiency measures and distributed generation choices to determine
the total cost of the building design in dollars per meter squared.
2.6
2.6.1
Optimization Methods
Energy Simulation: MIT Design Advisor
The MIT Design Advisor (Urban 2006) is a web-based building energy simulation tool
developed by MITs Building Technology Lab (http://designadvisor.mit.edu/design/).
Design Advisor aims to help reduce building energy demand early in the design
process. The tool was designed to allow a non-technical user to quickly develop models
and compare alternatives. This was a key shift in available modeling tools because
energy modeling is usually too cumbersome to be included in the initial design phasewhen major capital costs and energy loads are determined. Although the required
inputs have been simplified, the results are still very accurate. The daylighting results
were found to agree within 10% normalized error to LBLs Radiance (Lehar 2007).
Using steady-state heat flow calculations, U-Values and Solar Heat Gain Coefficients
were compared to LBLs WINDOW5 for a variety of glass types. While a few cases
differed by 5 to 10%, must agreed to better than 1% (Urban 2007).
Calibrated
comparisons for heating and cooling loads were compared DOEs EnergyPlus with
good agreement, although the models had to be simplified since EnergyPlus does
offer many more options (Urban 2007).
As seen in Figure 2-15, this process used by this tool is inherently multidisciplinary.
The user input, along with the appropriate weather file, is passed to the Daylight and
HVAC Loads models. The Daylight Model calculates the available daylighting and
passes artificial lighting demand needs to the HVAC Model.
65
These models then
output information about energy, comfort, and daylighting to the user.
Figure 2-15: Design Advisor process (Urban 2005).
The MIT Design Advisor calculates the HVAC loads through the energy balance
method, accounting for heat exchange due to heating, cooling, ventilation, internal
loads, and thermal mass, as well as heat exchange through the envelope of the
building. The energy output results give the annual heating, cooling, and lighting
needs. These numbers are given in terms of primary energy, so efficiency differences
between gas and electric are accounted for.
Therefore, in the optimization, we
analyzed the sum of the heating, cooling, and lighting needs or the total annual
energy use.
2.6.2
Initial Model
An initial model was developed using Design Advisor, MATLAB optimization scripts,
and custom modules to optimize building energy use and cost. At this initial stage, the
only implemented distributed generation option was PV. Initial single objective runs
were completed with a combination of genetic algorithms and local search methods.
However, due to extensive computation time, the subsequent multi-objective runs
were completed with simpler parametric methods.
To begin to understand the solution space, we performed single objective optimizations
for energy use and cost. In both cases, a genetic algorithm was used with varied
66
population sizes and generations.
As expected from the formulation, the additional capital cost above baseline
converged at $0/m2 when using a population size of 100 running for 50 generations.
The sensitivity of the optimal solution to all variables was 0(10-3) or less, therefore
we expect that our optimal solution is stable (Lee 2008).
Due to extensive computational requirements, for the annual energy use optimization,
we reduced the population size to 30 with 40 generations.
This converged to 95.51
kWh/m2 using only energy efficiency options (no DG) (Lee 2008, Brown 2010).
To evaluate solutions of both objective functions simultaneously, we sampled a
set of points between the solution points for each function.
The Pareto front was
approximated by determining the set of points which were not dominated by any
other point in the sample set.
Parameters
Site parameters include data about the site itself that cannot be changed.
These
include annual weather data (including average daily temperatures and average solar
radiation) for Boston and Lisbon, the latitude of the site, and the total area of the
plot available. The building type (office) is also a parameter, affecting the minimum
lighting levels required for office use and the average power density of electrical office
equipment.
Occupancy parameters include the total number of occupants using
the building, their schedule, and the internal heat gains they produce.
Setpoint
temperatures for summer and winter are fixed at typical values. The values for these
parameters are as follows:
1. Location: Boston, MA or Lisbon, PT
2. Building Type: Office
3. Occupation Density: 0.1 person/m2
4. Occupation Schedule: 7am to 8pm
5. Minimum Lighting: 500 lux
6. Equipment Density: 5 W/m2
7. Square footage per floor: 625 m2 ( 6700 sf)
67
8. Number of floors: 2
9. Average Room Dimensions: 5m x 5m x 3m high
10. Ventilation System: Mechanical Heating and Cooling (Chiller COP = 3, Furnace
Efficiency = 100%)
11. Minimum Temperature: 20 C (68 F)
12. Maximum Temperature: 26 C (78.8 F)
13. Fresh Air Rate: 15 L/s-person (1.8 air changes per hour)
Constraints
The only equality constraint included in the optimization is the total square footage
of each floor of the building. This constraint ensures that North-South faade length
times the East-West faade length must remain constant. Additionally, there were also
upper and lower bounds set on the continuous variables, which are discussed below.
Design Variables
We chose six design variables for the optimization, as shown in Table 2. The window
to wall ratio (xl) describes the percentage of the exterior wall that is made up of
windows.
The upper and lower bounds for this variable illustrate feasible designs.
Assuming a building must have a least some window area, the lower bound was set
to 10%. The upper bound was set to 90% to account for structural requirements and
frames. Most new commercial buildings are on the upper end of this range.
The wall R-value (x2) represents the level of insulation in the wall.
The lower
bound of 1 m-C/W is the worst case scenario, while the upper bound of 7 m-C/W
represents the best currently available materials.
The third design variable is the length of the North-South faade (x3). One of the
constraints was that each floor must be 625 m2, so the upper and lower bounds were
based on feasible dimensions, with two rooms next to each other as the minimum
length (10m). The length of the East-West faade was a dependent variable based on
the value of the North-South length and the desired square footage.
The window glazing type (x4) is a discrete variable that refers to the number of
68
Table 2.3: Initial optimization design variables.
Upper
Units
Variable
Description Lower
Bound Bound
x1
Window-to10
90
%
wall Ratio
x2
x3
Wall R-value
facade
N-S
1
10
7
62.5
m- 0 C/W
m
0
2
single,
double,
length
x4
Window type
triple
x5
Window
0
1
clear, low-e
0
2
0
625
zero, high,
low
m2
coating
x6
x7
Thermal
mass
PV
panes of glass in each frame. Currently, top windows have up to three panes, while
older windows typically have only one.
The window coating type (x5) is also a discrete variable.
While most older
windows have no coatings (clear), low-emissivity (low-e) coatings are now available.
A low-e coating is a type of spectrally selective material which coats a window and
results in lowered heat (and some light) transmission through the window.
Finally, the thermal mass (x6) refers to the construction material and its ability
to absorb and release heat. In the Design Advisor model, this refers to the floor slab
material. Higher thermal mass is typically used in areas with high diurnal shifts.
For example, in the winter, the thermal mass of the building can absorb heat, which
is later released during the cooler nighttime hours. Concrete buildings are often
considered high thermal mass, while structures composed of less massive compressive
materials such as brick may be considered low thermal mass. Frame structures using
wood or steel can be considered zero mass.
For the discrete variables, numerical values were assigned to the choices to be
used in the optimization. For example, in window type, a single glazed window was
assigned 0, double 1, and triple 2. The same was done for window coating and thermal
69
mass.
2.6.3
Comparison of Optimization Algorithms
One of the more challenging tasks in the design of this tool was selecting an appropriate
optimization algorithm.
Many, many options exist and the field is continuously
expanding. For the tool, we were looking for something that could handle a nonlinear solution space with mixed variables (both discrete and continuous). And we
needed something that could be efficiently calculated in an online environment.
We first looked at traditional gradient-based optimization methods. These local
searches are often very efficient at converging on a solution.
However, they are
susceptible to local minima, which is problematic with non-linear solution spaces.
Additionally, such solutions typically require a starting point and can only handle
continuous variables.
We also considered some heuristic global search algorithms, such as genetic algorithms
and simulated annealing. In contrast to the gradient-based methods, these algorithms
tend to be more expensive and they can't guarantee convergence. However, they can
handle discrete variables, they do not require a starting point, and due to their
stochastic nature, they are good at avoiding local minima.
Through initial optimization tests, we decided that neither of these optimization
categories were ideal for our problem.
specifically Monte Carlo.
So, we also considered random sampling,
A Monte Carlo sampling does not have a convergence
method, so it is not traditionally considered an optimization algorithm. However,
such a sampling is useful because it does not require a starting point and it can
show the solution space more efficiently. To understand the paths toward ZNE, we
were interested in understanding both the optimal solutions and the overall solution
space. Providing the user with minimum cost solutions at given energy targets was
the final goal, but it also useful to provide the user with information about the overall
solution space so that they can understand the range in energy and cost with different
technology combinations.
70
2.6.4
Latin Hypercube Monte Carlo
Monte Carlos are a general group of algorithms. There are a variety of methods,
but in general they all contain some stochastic element. They are often implemented
with calculations where there is uncertainty in input variables, interactions between
variables, and/or an interest in doing a sensitivity analysis.
For this tool, we implemented a Latin Hypercube Monte Carlo (LHMC). This
algorithm is already being used in the field to calibrate building simulations to
measured data. While the variables and calculations in our study are very similar to
those used in calibration methods, rather than trying to identify influential variables
and tighten bounds, we were interested in using known influential variables and
bounds (from the design of experiments) within a Monte Carlo framework to efficiently
explore the solution space and approximate Pareto optimal solutions.
The main steps in a LHMC are:
1. Discretize variables: Split the variables into multiple ranges and pick the median
values to set levels to explore.
2. Choose the number of trials.
3. Generate a matrix of size
(# of trials
by
#
of variables).
4. Randomly choose the level for each variable in the matrix.
5. Complete a chi square test to check for randomness.
In the tool, a LHMC is used to generate a set buildings with different combinations
of variables.
Then, we calculate the energy and cost of each building in the set.
Finally, the least cost option is picked at given energy goals. So, the end result is a
Pareto optimal point that cannot decrease energy use without increasing cost and vice
versa. In essence, this is comparable to multi objective optimization that minimizes
energy use and cost.
2.7
Building Model Emulator
While Design Advisor is fast enough to provide quick feedback in an online setting, to
calculate annual energy use takes about a minute. As is, this would be prohibitively
71
expensive to use as an objective function in an online optimization tool. Therefore,
an emulator has been developed to allow objective function calculation time to be
reduced to a fraction of a second.
The emulator uses pre-calculated results from Design Advisor to construct its
rules, without reference to the equations of heat transfer used in the original model.
2.7.1
Training Runs
As seen in Figure 2-16, eleven variables were explored with two or three levels each. A
full-factorial set of training runs were completed, totaling 6912 runs. While some runs
took about a minute, others took longer to account for more complicated calculations.
On average, the 6912 runs usually took about ten days to run.
Of the thirty US cities available in Design Advisor, these training runs were
completed for seven of them.
These cities were chosen to represent a variety of
weather types with respect to latitude, heating degree days, cooling degree days,
relative humidity, and average solar radiation. The seven analyzed were:
1. Anchorage
2. Los Angeles
3. Honolulu
4. New Orleans
5. Boston
6. Buffalo
7. Washington DC
2.7.2
Initial Multivariate Regression
As seen in Figure 2-17, the first regression analysis was completed using the full
dataset. The red line represents the actual output from Design Advisor, while the
blue line represents the predicted output from the regression equation. The grey dots
are the percent error. In this case, the sections in the middle perform moderately
well, with most errors falling at about 20%. However, at the edges, errors reach as
72
Design Advisor Runs
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
Low
Single
Clear
Double
Low-e
High
units
Triple
30
12.5
90%
50 m
1
1
5SI
6SI
NW/SE
N/S
Bitumen
Zero
Mechanical
Green
Low
Joint
Always On
Dim Indep.
Cool
High
1
25 % room depth
sec/run
combinations
sec
min
hrs
days
levels
3
2
2
2
2
2
3
3
2
2
2
65
6912
449280
7488
124.8
5.2
Figure 2-16: Emulator training runs.
much as 50%, which is unacceptably high.
2.7.3
Multistep Multivariate Regression
For the next step in the development of regression equations, the data was split into
three tiers as seen in Figures 2-18 and 2-19. As before, all of the data was used as an
input to the first tier. Then, it was split by top and bottom half of energy use from the
full dataset and new regressions equations were calculated. Finally, new regression
equations were also calculated for a third tier that was splits into quadrants. As seen
in Figure 2-19, when splitting between the top and bottom half, to avoid edge errors,
we used a slight overlap in the middle from 45-55%. The same reasoning was used
for the ranges in the third tier as well.
Rather than trying to predict the total energy in one equation, this process was
completed three separate times for heating, cooling, and lighting energy, resulting is
separate regressions for each. This is because high energy use in one does not mean
high energy use in the others. For example, a building could have high lighting and
cooling use, but low heating use.
Now, as seen in Figure 2-20, when this system of walking through three tiers of
regression equations is used, the highest errors are closer to 10%. Since this calculation
73
400.
360.00
300.0%
40A%
180.0
100.0
~
0.0~~~~~
1 9 17 25 33 41 49
57
66 73 61
8U 97
20--.N---IN1----%i
106113 121129137 146153161 1T7186193 201 20 217
-- Design As$or-e--Prodied
+
%Enor-P(%Evr)
Figure 2-17: Multivariate regression: No splits
runs in a fraction of a second instead of 30-60 seconds, we felt this small error was a
reasonable tradeoff for speed. Figures 2-20, 2-21, and 2-22 show a detailed analysis
of the regression runs in Boston. Note there is a change in scale in the y-axis and
errors are especially high at the edges. While mid-range errors were only 20% when
looking at the full set, errors in the low and high energy cases were as high as 50%.
In contrast, when split into 3 tiers, the new regression equations were typically 10%
or lower.
74
/
z
Split and recalculate
with new regression
coefficients
/
e--A
Figure 2-18: Data is split and recalculated with new regression coefficients.
75
Tier 1
0-100%
Tier 2
45-100%
0-55%
Tier 3
0-27%
23-52%
48-77%
73-100%
Figure 2-19: Data is split into tiers with overlapping divisions.
22.0
000%
200.0
800%
178.0
400%
I
100
A
125.0-
.200%
I
100.
or
'
+
,
*
-
,
7M
1
3
5
7
9 11 13 15 17 19 21 23 25 2
-+-DesignmAdsor--
2f 31 33 35 37 3
Pmedd
*%
Enr-
P
41 43 45 47 49 51 53 55
(%Eror)
Figure 2-20: Multivariate regression: 3 splits.
76
100%
C
0'0
6912
run #
Figure 2-21: Boston results:
prediction in blue.
Tier 1.
Design Advisor output in red, regression
run #
3001
600
250
500
200
400
150
300
100,
1000
Figure 2-22: Boston results:
prediction in blue.
200113457
-
3456
Tier 2.
6912
Design Advisor output in red, regression
77
0
1728
Figure 2-23: Boston results:
prediction in blue.
2.7.4
5184
3456
Tier 3.
6912
Design Advisor output in red, regression
Multicity Multivariate Regression
After completing the training runs and developing multivariate regression equations
for the seven cities listed above, a multicity multivariate regression equation was
developed so that regression equations could be used to predict energy use in cities
other than the 7 cities explored in the training runs.
Here, we compiled 48,383 runs from the 7 cities. We looked at the 11 design
advisor variables, in addition to 5 weather parameters and calculated coefficients for
these 16 factors. Table 2.4 outlines these weather inputs.
The 16 weather parameters and design variables include:
1. Latitude
2. Heating degree day (base 65)
3. Cooling degree day (base 65)
4. Relative humidity
5. Average solar radiation
6. Window typology (single, double, triple)
7. Glazing type (clear, low-e)
8. Glazing to wall ratio
9. Building length
10. Orientation
11. Wall insulation R-value
12. Roof insulation R-value
78
13. Roof type (Bitumen, green, cool)
14. Thermal mass (zero, low, high)
15. HVAC (mechanical, joint)
16. Overhang depth
17. Lighting control (always on, dim independently)
Full
iName
Anchorage
Phoenix
San Diego
Los Angelos
San Francisco
Denver
Miami
Atlanta
Honoludu
Chicago
New Orteans
Boston
Detroit
Minneapolis
Saint Louis
Kansas City
New York
Bufalo
Cincinnati
Cleveland
Portland
Philadelphia
Pittsburg
Memphis
San Antonio
Dallas
Salt Lake City
Seattle
Milwaukee
Washingon DC
Latitude
61
33
32
33
37
40
25
33
21
41
29
42
42
44
38
39
40
42
39
41
45
39
40
35
29
32
40
47
42
38
HDD 65
CDD 65
RH
0
10470
1552
2661
1507
352
224
1819
3042
22
6016
529
206
2442
3095
1017
0
2524
506
6127
1465
1597
5630
419
6228
413
8159
454
4750
927
5249
774
4848
639
294
6927
5070
578
6154
489
4366
188
4865
682
5930
443
3227
1234
1570
1677
2290
1533
5983
669
5185
76
7444
336
4921
1113
Ave Solar
72
34
73
73
75
50
72
66
68
70
76
66
71
66
66
68
69
71
74
70
75
66
66
66
67
64
54
74
73
69
881
2523
2099
1760
1770
1975
1454
1580
1853
1292
1435
1335
1226
1502
1527
1699
1280
1197
1198
1130
1163
1410
1136
1585
1629
1782
1814
1156
1344
1197
Table 2.4: Weather Inputs for 30 Design Advisor Cities.
In the regression analysis, heating, cooling, and lighting energy were calculated
separately. Results were also split into three tiers, as explained above. This resulted
in 21 different regression equations.
To approximate energy use, the emulator first calculates heating, cooling, and
79
lighting energy use based on the first tier of equations. Again, these are based on the
entire set of solutions. The emulator calculates whether the predicted solution is in
the top or bottom half of the energy use solutions and moves to the second tier of
equations. If the prediction from the first regression falls within the range of the lower
half of the energy from the full set of training data, in the second tier calculation, the
lower energy regression equation is used. If the predicted energy use falls within the
higher range, the higher energy regression equation is used in the second tier.
Again, heating, cooling, and lighting are calculated separately and can move
through tiers independently. So, a building could have a high cooling energy use and
a low lighting energy use. Based on the second round of calculations, the emulator
picks the closest quadrant of energy use and recalculates heating, cooling, and lighting
energy use for a final solution. Tables with regression coefficients are included in
Appendix A.
For various cities, we spot checked 100 buildings that were not included in the
regression training data. Table 2.5 provides an overview on the results in Boston,
New Orleans, LA, and San Antonio. As you can see, the average error in each was
less than 8%. While the maximum errors were relatively higher in New Orleans and
LA, in both cases, these errors occurred buildings within the highest quadrant of
energy use. These buildings will not fall within our optimal solutions, so we are more
concerned about the accuracy of the lower energy buildings.
To verify that these errors would not be unacceptably high in all cities, we checked
inefficient buildings in each location. The results are outlined in Table 2.6. In the
highest quadrant of energy use, the average error was 17.4%. In contrast, for buildings
that fell in the bottom quadrant of energy use, the average error was only 3.7%.
Boston
min
max
mean
New Orleans LA
San Antonio
0.0%
0.0%
0.0%
0.70%
23.5%
4.3%
30.7%
3.7%
41.1%
7.7%
16.80%
5.20%
Table 2.5: Emulator error in various cities
80
Anchorage
Atlanta
Boston
Buffalo
Chicago
Cincinnati
Cleveland
Dallas
Denver
Detroit
Honolulu
Kansas City
Los Angelos
Memphis
Miami
Milwaukee
Minneapolis
New Orleans
New York
487.6
477.9
431.9
441.1
469.7
454.3
449.2
544.7
449.4
460.8
619.6
488.4
439.6
511.9
608.2
471
520.3
530.2
441.1
689.1
517.56
473.11
545.2
525.11
602.38
515.97
798.13
390.95
529.74
535.58
512.44
432
572.6
689.68
591.22
579.54
526.82
476.87
Philadelphia
457.1
459.99
0.6%
Phoenix
Pittsburg
Portland
Saint Louis
Salt Lake City
San Antonio
San Diego
San Francisco
Seattle
623.5
448.5
381.9
491.5
492
562.5
438.8
401.1
365.9
812.33
469.44
485.9
586.46
423.47
716.84
491.3
317.69
503.89
30.3%
4.7%
27.2%
19.3%
-13.9%
27.4%
12.0%
-20.8%
37.7%
458.7
546.86
19.2%
41.3%
8.3%
9.5%
23.6%
11.8%
32.6%
14.9%
46.5%
-13.0%
15.0%
-13.6%
4.9%
-1.7%
11.9%
13.4%
25.5%
11.4%
-0.6%
8.1%
Washington
DC
Table 2.6: Emulator error in various cities for inefficient buildings
81
2.8
Interface Design
As with Design Advisor, we wanted an interface design that would allow users to
quickly set up simulations. So, we continued with the use of drop down menus and
radio buttons.
As shown in Figure 2-24, the user begins by selecting the city. Then, they select
four energy goals, where 0% is the baseline building and 100% in a ZNE building.
Next, for each of the design variables, the user can choose to include them in the
optimization or to freeze them. For example, they may have a building with a set
orientation, so they can freeze that variable when running the optimization. The user
then selects the baseline values for each of the design variables.
Figure 2-25 shows the default cost values provided by the tool. This is also where
the the user can change cost parameters.
They can set the price of PV, CCHP,
electricity, and gas. They can also set the discount rate that is used when calculating
the net present value of energy cost. Finally, the user can also set an overall cost
multiplier. So, for example, if they lived in a more expensive city, they could look
at how higher costs may affect the solutions. In the current version, the user cannot
input separate costs for each variable, but this would be an area for improvement in
future versions.
Figure 2-26 shows the top of the output screen after running the simulation. The
table provides the total cost optimal solution at the given energy reduction goals.
The values of each of the design variables are provided, in addition to energy, capital
cost, and total cost.
Figures 2-27, 2-28, 2-29 show the graphs that are created after the simulation is
complete. These show net primary energy vs. capital cost, total cost vs. capital
cost, and net primary energy vs. total cost. For each graph, the baseline building is
marked in orange. The four optimal solutions are red. All other buildings modeled
are in blue. A more detailed analysis of these outputs is included in the Results.
82
Coptimizer
#4=adty
Seise Geals
. M.. bn
".Call"s
Good
0,so
%o
ines~poa
0od0aM4esi
aldgye
spma
P Opdadas
Poltasin oWheRao
p
p|
o-i
S
pa
%ime[0pd=in
WImdtin R
a
Pweie O.
-Pu
BuidngLagh
%,between 10-90
Pmm
i. between 123 37.$
lPad
betwe
'(m2-CyW,
|
e
-5
as
%. between 10 - 90
m.between 12.5 - 37.5
between 1- 5
(m2-CW.
[
..
...
e e mM. .W
ro
mga
PV orCP
-t
Ophmim F1Pmew*v
%
depti, betswen I - 3
m depthbetwee
I
:
Figure 2-24: Interface design: City, goals, variable, and baseline inputs
IParmeter
Value
PV Price
sooo
/kW
CCHP Price
is00
$kW
77s
Dicount Rae o
Electricty Price 0.1s20
$/kWh
Gas Price
Multiplier
1.00
Figure 2-25: Interface design: Cost inputs
83
-30
25%
Goal
Net Priaryewgy
22857 kWhft2
Wdw Typoog
Tiple
lw-e
Glaing Type
10.1
ruin to WiRaio
32.1
Buinwg Length
Odentaton
NW-SE
aR InumISIoa R Whoe 2A
RoofInulaina Rt RVlue 3.1
Roof Type
Bineen
Then MISS
zero
HVAC
Mehaical
Ovehang Depth
6.5
Ighting Contml
Dim-ladsp
PV
33
CCHP
5.9
PVorCCHP
PV
$37.4
Capital Cost
TOal
Cost
$4.11
Bocause you've ipocintela PV systm
Goal 2
Net Puimry enegy
wihdow Typolog
Glaing Type
Glazing to Wil Ratio
Bulkng Legh
so%
Goal3
13626 kWVm2 Net Priarty enrg
Wndw
Typogy
10.9
22.0
NW-E
Odentsadon
Wal Innatan Rmalwe 3.0
Root Inslation R lue 3.9
Bit-en
Roof Type
ThOnnaTas NII
Mecbmkal
HVAC
Overbang Depth
29.0
75%
67.61 kWhfm2
Tiple
Glazing Type
Law-e
Glaing to Wal Ratio
Buiding Length
Odentation
20.2
19.6
NW-SE
Wail inatioa R Value 4.6
Roof Insultio RAlue 6.5
Roo(Type
Thenmalu M
HVAC
Overhang Depth
Lihtn Conne
PV
CHP
PV orCCHP
PV
20.1
CCHP
33
PV orCCHP
PV
S119J0
Capita Cost
Total Cost
$25.67
nybitmen oofa are vailabl.
CNapai Cost
B
uen
High
Joint
17.2
Dn-indep
24.3
14,
PV
$225.4
Total Cost
$9727
Ig%
sohins found
God 4
No
Figure 2-26: Model output: 25%, 50%, 75%, and 100% reduction from baseline goals
and variable results
I
0
0
30
0
200
I
0
100
_________________________________________________I
0
-100
0
100
*Agl
usngsa * SdeId &ag
200
* 0..n.
Figure 2-27: Model output: Net primary energy vs. capital cost
84
180
120
0
e
0
40
i
.100
0
*
--
100
CuCoMS Abo
AM bdngs
Sebdad
200
.a(st>
inag
* SN.OW
Figure 2-28: Model output: Total cost vs. capital cost
NAfwwg
ewr
I
c.
*
S
0
C
0
S
200
I
a
S.
0.*
100
0
0
0
ant COM
.ma
* Wsu*CW&n
*MASaas
120
a
Uafu.mns
Figure 2-29: Model output: Net primary energy vs. total cost
85
160
86
Chapter 3
Results
Initial tests of the model were completed on a sample 2-story office building in Boston.
We ran initial tests with PV at $4/W and $8/W (no tax breaks).
In addition, we
looked at net metering prices of $0.1520 and $0.10 per kWh. Discount rates were
varied from 8% to 12%. And finally, we explored 5, 10, and 30 year projections.
3.1
Initial Runs
Figure 3-1 shows the standard output from the simulation runs. The top graph shows
the total cost vs. the capital cost above baseline. The total cost is equal to the capital
cost plus the NPV of the energy cost. In the axis, the baseline is at zero. The plotted
points are separated into buildings that only employ energy efficiency measures, ones
that use just PV, and ones that use both. The bottom graph shows the net annual
energy use vs. the capital cost above baseline. The total cost of the system is low
over 30 years due to the tax breaks on PV and the avoided energy costs over the
years. However, to reach net zero energy (as seen in the points on the far right of
the bottom graph), significant capital costs above baseline must be met. Figure 3-1
and Figure 3-2 show the same building, but over 5 and 10 years, as opposed to 30
years. After 5 years of energy savings, the total cost of the system is still well above
the baseline. However, with the PV tax breaks, Figure 3-3 illustrates that the total
cost does go below baseline after 10 years. In contrast, as seen in Figure 3-4, when
87
you calculate the cost of PV without tax breaks, it takes 30 years to just break even
on the total cost. And that is after investing extremely high capital costs. If the
discount rates are higher, or the net metering buy back rates are lower, one simply
cannot break even over the lifetime of the PV cells.
Law fles coumuutii usen In sehhmn
WIMM U%uWm
100 rV
f 30 yem
-
0
10
10
2oc
11
OWnPn
50
23
4'
Aa
4,
ar
%.LZ
0
W0
100
Ile
$4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
88
0W
11W
300
LOW ima
cOmminO OuS" in Ds~
SUW PV, .ISaWiW,
Iao W
Ma.,
S
5ye
SeS
-se
*
LU
Ilk
*
-1501L
0
1100
cp- CAoet B.inf
JuEE
*Jug PV
2to
itsna2
Figure 3-2: Simulation results for a low-rise commercial building in Boston. Assumed
$4/W PV, $0.1520/kWh, 8% discount rate, 5 years
LaW ies ComMnw Sm"n in8suM
?PV. UWi$Wftk S% SWOUR"
lowes
soo
100
apasenKA
-aco
ewcmmesr
'
*JUNg EE
0So
100
150
20D
250
0
Abenerune sMas
GeW Owasr
Figure 3-3: Simulation results for a low-rise commercial building in Boston. Assumed
$4/W PV, $0.1520/kWh, 8% discount rate, 10 years.
89
Low Oft*~mra
C
OM,
3Wa
11
1MA"Nso
s1
OW
so1eso
WIetf#
anew*u rift
*
0~
~
-
Eand PV
PV
Ica
o
so
100 Joe0
200
250
3W
350
400
400
000
Figure 3-4: Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
3.2
Final CCHP Runs
In the updated online implementation, both PV and CCHP were considered for DG
alternatives. Table 3.1 shows the inputs for a typical baseline 625m 2 2-story office
building. For the costs, we looked at building that had $8/W PV, $1800/kW CCHP,
$0.1520/kWh grid electricity, and $1.25/therm grid gas.
The optimization freely
chooses between PV and CCHP, along with efficiency options. In all cases explored,
the cost optimal solution for the energy reduction goals chose CCHP over PV.
In the simulations, ZNE goals of 25%, 50%, 75%, and 100% reduction from the
baseline building's energy use were set. The baseline building would be considered
0% of the way to ZNE, while 100% would represent a building that achieves ZNE.
Table 3.2 shows the total cost optimal solution for each of these energy goals for
a low-rise building in Boston. A table like this is generated when the tool is run. It
outlines the net primary energy, decision variable solutions, capital cost, and total
cost for the four energy goals.
Three graphs are also generated when the tool is run, as seen in Figure 3-5:
1. Net primary energy vs. capital cost above baseline
90
N... .[.Optin
Optimize
Widow Typology
jfx
sgto W
uiding Length
nR
Freezel
Feeze
Optimize
Freeze
e o Optiz0
Ibeof
4!ype
@Optimize
renna
IHVAC
a
[
erbag Det
lughting Control
PV oCCHP
II
o
between 10 - 90
,between 1235-37.5
.
-
%,between 10 - 90
so
2s
7n,between 12.5 - 37.5
N--S
km2-CYW, between 1 -5
m2-CYW, between 1 -5
Freeze
0Freeze
-amen
---
:-
iz I u Freeze' zro
FeeL| echmc
Opimz
@LOptimnize
Banudne
ue
Freeze singie
*toj.tmze
@Optimize
lOrenttian
Valnau
K
e Freezel
r
zeo
room depth, between 1- 30
1
Mecankc
OptimizeU F=eze Always-on
Jptmizele Freze Pv
%~~7
ro deth bewe 1-30
AbOVS-o-
Table 3.1: Baseline low-rise commercial building.
2. Total cost vs capital cost above baseline
3. Net primary energy vs. total cost about baseline
In each of these graphs, the baseline building is marked in orange, the selected
optimal solutions are red, and all of the other buildings simulated are in blue.
91
Goal
25%
50%
Net Prmary Energ
128.47 kWh/m2
128.47 kWhIm2
39.44 kWh/m2
-6.35 kWh/m2
Single
Single
Double
Triple
Glazing Typ
Clew
Claw
Clear
Clear
Glaing to Won Ratio
80.7
80.7
10.3
13.7
BuNing Length
33.7
33.7
30.2
26.4
Orientation
NW-SE
NW-SE
N-S
NW-SE
Wel insultlon R Value
2.0
2.0
3.0
4.1
Roof Insulaton R Value
2.5
2.5
4.0
5.6
Roof TA
Bitumen
Bitumen
Bitumen
Cool
Thermal Mom
High
High
High
High
HVAC
Mechanical
Mechanical
Mechanical
Mechanical
Overhang Depth
10.9
10.9
28.9
27.6
Ughng Con
AMws-on
Always-on
Always-on
Dim-Indep
py
16.9
15.9
23.0
8.7
CCHP
19.5
19.5
18.4
19.3
PV or CCHP
CCHP
CCHP
CCHP
CCHP
Capital cost
$-27.67
$-27.67
$16.05
$48.63
Total Cost
$-159.26
$-159.26
$-156.66
$-148.41
75%
100%
Table 3.2: Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
92
Not PrWmay EnergyJCapital Cast
450
I
I
em
301
S
a
15D
U
0
0
C#
M
Buidrs
300
150
Cosl Above BeaVw ($r7)
M Selectad
uiunmgs
U
450
Ban
Toal CostAaNaI Cost
300
158
0
0
-300
0 N50
* M
300
uLddngs
U
Selecdd Baukungs
U
40
aen
NEt PrimnyfTaw Cast
450
IR
300
a.
150
0
-150
-150
Total
M A Budns
0
Above 8&mos
10
3W0
(Vm2j
OBasan
U Sci8edd suSnM
Ot
Figure 3-5: Simulation results for a low-rise commercial building in Boston. Assumed
$8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
93
3.3
Final PV Runs
After completing updates and the online implementation, new runs were also done in
San Francisco and Milwaukee to explore both a mild and a cold climate. Again, Table
3.1 shows the inputs for a typical baseline 625m 2 2-story office building. However,
this time the DG option was frozen on PV, forcing the algorithm to only consider PV
in the DG options. The maximum PV system size is limited by the roof area. We
wanted to understand the effect of changing the cost of PV and grid energy prices,
so for both cities, three simulations were explored:
1. $8/W PV price (full price), $0:1520/kWh (current retail price)
2. $4/W PV price (price with tax breaks), $0.1520/kWh (current retail price)
3. $4/W PV price (price with tax breaks), $0.30/kWh
The following sections show the results of these simulations.
3.3.1
San Francisco
In Table 3.3 and Figure 3-6, you can see that ZNE can be achieved in San Francisco.
When PV is $8/Watt and the grid electricity price is $0.1512/kWh, this would require
a capital cost of $255/m
2
above baseline and a total cost $76.22/m 2 . As noted above,
the average cost of a 2-4 story office building was $1891/m
$255/M
2
2
in 2010. An additional
would mean a 13.5% increase in capital cost.
Over 30 years, the additional cost of the measures will not be outweighed by the
energy savings. In contrast, the 50% ZNE goal can be achieved at a total cost of
$0.03/sf, meaning that you would approximately break even. Anything above the
50% would not be cost effective from a total cost perspective.
In the next set of simulations, we looked at a PV price of $4/W, which is possible
to achieve now with tac breaks. In this case, as seen in Table 3.4 and Figure 3-7,
reaching ZNE would require an additional $122/m
2
in capital cost. But, when PV
is half price, the total cost of the ZNE building is $-45/m 2 . This means that the
energy savings over 30 years would outweigh the additional cost of the measures. It is
also interesting to note that the cost optimal solution at each goal has the roof nearly
94
covered in PV. In contrast, at the real PV price, the first three goals were met without
using all of the roof space for PV. When PV price is $4/W and the grid electricity
price is increased to $0.30/kWh, ZNE can be reached with a total cost of $-166.72.
In this case, the first three goals were all met by the same building, meaning that it
was as cost effective to go 75% of the way to ZNE as it was to go 25% of the way
there.
Goal 1
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R
Value
Roof Insulation R
Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
25%
206.78
kWh/m2
Triple
Clear
14.5
28.8
N-S
1.8
2.1
Bitumen
Low
Mechanical
17.8
Always-on
8.0
9.0
PV
$40.13
$-19.63
Goal 2
50%
149.48
kWh/m2
Window Typology
Triple
Glazing Type
Clear
Glazing to Wall Ratio 21.3
Building Length
30.0
NW-SE
Orientation
Wall Insulation R
1.9
Value
Roof Insulation R
2.4
Value
Roof Type
Bitumen
Thermal Mass
LOw
HVAC
Mechanical
Overhang Depth
1.4
Always-on
Lighting Control
15.1
PV
13.4
CCHP
PV
PV or CCHP
$89.21
Capital Cost
$0.03
Total Cost
Net Primary energy
Goal 3
75%
65.42
kWhm2
Window Typology
Triple
Low-e
Glazing Type
Glazing to Wall Ratio 10.2
Building Length
26.0
Net Primary energy
Orientation
N-S
Wall Insulation R
Value
4.0
Roof Insulaton R
Value
Roof Type
Thermal Mass
5.5
Overhang Depth
Lighting Control
Bitumen
Zero
Mechanical
3.6
Always-on
PV
24.2
CCHP
PV or CCHP
Capital Cost
Total Cost
3.3
HVAC
Goal 4
100%
-25.81
kVA/m2
Window Typology
Triple
Glazing Type
Low-e
Glazing to Wall Ratio 11.1
Building Length
19A
Orientation
N-S
Wall Insulation R
4.5
Value
Net Primary energy
Roof Insulation R
Value
Roof Type
Thermal Mass
HVAC
PV
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
$152.02
Capital Cost
$19.69
Total Cost
6.2
Bitumen
Low
Joint
14.5
Dim-Indep
293
11.1
PV
$255.39
$76.22
Because you've specified a PV system, only Bitumen roofs are available.
Table 3.3: Simulation results for a low-rise commercial building in San Francisco.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
95
Nt P*nay Energoc~ap
Cost
460
I
a
300
150
I
ai
3M0
Cap
ABueONVs
Cast Abow &MWa
U 3sece
(A72)
&esnsv
tesnes
Towa coesCapa CoC*
150
ep
100
31
a
.50
-100
100
0
200
*A nO ouns
*Se3$edu sA
300
Q2)
Capta Cast Ab"o BnAns
e
NBeosee
Not Primaryffosl Coo
450
I
300
S
0
0
0
150
I
0
Si
a
0
-150
100
50
R7aW
U MI
&Os
150
Cost AbovBasrnOm (&w2
U seWd & nps
U
M"nee
Figure 3-6: Simulation results for a low-rise commercial building in San Francisco.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
96
Goal 1
25%
Net Primary energy
88.72 kWh/m2
Window Typology
Single
Glazing Type
Clear
Glazing to Wall Ratio 273
Building Length
25.1
Orientation
NW-SE
Wall Insulation R Value 1.1
Roof Insulation R Value 1.1
Roof Type
Bitumen
Thermal Mass
Zero
HVAC
Mechanical
Overhang Depth
20.5
Lighting Control
Always-on
PV
28.5
CCHP
5.2
PVorCCHP
PV
Capital Cost
$57.90
Total Cost
$-62A7
Goal 2
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
50%
88.72 kWh/m2
Single
Clear
27.3
25.1
NW-SE
1.1
1.1
Bitumen
Zero
Mechanical
20.5
Always-on
28.5
5.2
PV
557.90
$-62.47
Goal 3
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
75%
7939 kWh/m2
Single
Clear
19.4
28.4
NW-SE
3.6
4.8
PV
26.3
6.9
CCHP
PV or CCHP
Capital Cost
Total Cost
Goal 4
Net Primary energy
100%
-1.78 kWh/m2
Window Typology
Double
Glazing Type
Glazing to Wall Ratio
Low-e
14.0
Building Length
32.2
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
NW-SE
4.4
6.2
Bitumen
Zero
Mechanical
PV
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
23.3
Dim-Indep
30.0
13.0
PV
$64.49
$-60.67
Capital Cost
Total Cost
$121.70
Bitumen
High
Mechanical
21.6
Always-on
$-45.12
Because you've specified a PV system, only Bitumen roofs are available.
Table 3.4: Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
97
Not Pdmry EnrpCapital Cao
450
I
U.
.*,
.
150
ISO
3.
150
70
C*p&CosW
Abow Ue4W (5bn2)
SAl Buddves MSeiectd Adng
MBaiseine
Toi
140
210
CouVCaptM Cot
100
to
50
e;
0
.4 .
V
0
.3
-
Y
0
a
-70
70
Ce*P&CoaW AbMo 60ONe (Sm2)
UMAlIWaOr
MeU
8 A*qd
s
Mft.sle
40
210
Nt Pfum)rv/TOal Cost
450
I
U.
3WQ
0
150
-
0
100
50
0
TOW&
Cot
SAlldOdes
A0"
50
100
WeMAe "w2
MrWscdeuA*ung
MOasene
Figure 3-7: Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
98
25%
Goal 1
44.76 kWh/m2
Net Primary energy
Double
Window Typology
Clear
Glazing Type
Glazing to Wall Ratio 15.4
Building Length
35.6
Orientation
N-S
Wall Insulation R Value 2.5
Roof Insulation R Value 3.2
Roof Type
Bitumen
Thermal Mass
Low
IIVAC
Mechanical
Overhang Depth
21.0
Lighting Control
Always-on
PV
CCHP
29.0
9.6
PV or CCHP
PV
Capital Cost
Total Cost
$82.20
S-199.91
50%
Goal 2
44.76 kWh/m2
Net Primary energy
Double
Window Typology
Clear
Glazing Type
Glazing to Wall Ratio 15.4
35.6
Building Length
N-S
Orientation
Wall Insulation R Value 2.5
Roof Insulation R Value 3.2
Bitumen
Roof Type
Low
Thermal Mass
Mechanical
HVAC
21.0
Overhang Depth
Always-on
Lighting Control
29.0
PV
9.6
CCHP
PV or CCHP
PV
$82.20
Capital Cost
Total Cost
$-199.91
Goal 3
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
75%
44.76 kWh/m2
Double
Clear
15.4
Building Length
35.6
Orientation
N-S
Wall Insulation R Value 2.5
Roof Insulation R Value 3.2
Bitumen
Roof Type
Thermal Mass
Low
Mechanical
HVAC
21.0
Overhang Depth
Always-on
Lighting Control
29.0
PV
9.6
CCHP
PV
PV or CCHP
$82.20
Capital Cost
$-199.91
Total Cost
Goal 4
Net Primary energy
100%
-4.59 kWh/m2
Window Typology
Triple
Glazing Type
Glazing to Wall Ratio
Building Length
Low-e
13.6
12.6
Orientation
NW-SE
Wall Insulation R Value 4.3
Roof Insulation R Value 5.9
Bitumen
Roof Type
Thermal Mass
Low
HVAC
Overhang Depth
Lighting Control
PV
CCHP
Joint
6.8
Dim-Indep
27.0
4.0
PV or CCHP
PV
Capital Cost
Total Cost
$165.39
$-166.72
Because you've specified a PV system, only Bitumen roofs are available.
Table 3.5: Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years.
99
Nt Prkmary
EnergyOCapita Cast
450
ii
0
%
*
~*
ISO
9
S
0
140
0
70
210
Cap.laf Cot Aove BaslsAe (S*M
UA sse
M
U
ftisdree
s a 80asr-#
Total COtUCapital Coat
70
0
.70
4
*
%@tooS
0
-140
-210
0
,70
140
210
CapW Cos Abt 5e ase (MM"4
* AllWOOV M Se.c4 Buldw~e U PaeWi
Ntt Primaryffloai Cost
450
S
24
300
0
150
a
M&A
PM1% -ToWe
-
0
,210
-140
-70
7b?.W
Co"t Abo' feteWe (Sen"
mAll sAW&
U SeW~saOue. U es..r
0
Figure 3-8: Simulation results for a low-rise commercial building in San Francisco.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years.
100
3.3.2
Milwaukee
The same set of runs were completed in Milwaukee and the results are included below.
In these runs, the building was unable to reach ZNE with onsite PV. At $8/W PV
price and $0.1512/kWh electricity price, you can reach 75% of the way to ZNE, but
at a capital cost of $201/sf and a total cost of $60.29/m 2 . If the PV price is reduced
to $4/W, these go down to $104/m
2
and $-25/m 2 , respectively. If the grid price is
increased to $0.30/kWh while PV remains at $4/W, 75% of the way to ZNE can be
achieved at a capital cost of $121/m
2
and a total cost of $-145/m
2
Goal 1
25%
Goal 2
50%
Goal 3
75%
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
228.26 kWh/m2
Triple
Clear
17.2
31.7
NW-SE
3.6
49
Bitumen
Low
Mechanical
14.8
Always-on
7.7
7.5
PV
$46.83
$-,91
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PVorCCHP
Capital Cost
Total Cost
134.71 kWh/m2
Triple
Clear
103
33.9
N-S
4.2
5.8
Bitumen
Low
Mechanical
8.8
Always-on
19.5
14.5
PV
$123.14
$2638
Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Roof Insulation R Value
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
47.19 kWh/m2
Triple
Low-e
11.6
28.1
NW-SE
4.8
6.7
Bitumen
High
Mechanical
16.9
Dim-Indep
27.1
2.7
PV
$201.97
$60.29
Goal 4 100%
No solutions found
Because you've specified a PV system, only Bitumen roofs are available.
Table 3.6: Simulation results for a low-rise commercial building in Milwaukee.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
101
400
*I
%
I
3W0
200
a -
6b0
0
100
0
0
100
0
-100
200
Capw&CoaW Ao~w6ftmMo ("2)
E AM8OddeV
Sfette O&4ngs
U
TotWlCosCpit
Bseine
Cost
10
120
60
,a
100
CapW Coat AoN e 86.SM' (WM2)
U M AOdegs USeected 83Wngs U B00s0600
-100
Not PrnmryffToaI CGt
0
I
*
*
S
0e
U
0
S
S
S
I
200
*
C
U
Sg
100
40
0
60
Cost AD"
M8U4z4 Selctd
WU
Total
120
emfie ($M
Suebngs
18D
2)
BasetOne
Figure 3-9: Simulation results for a low-rise commercial building in Milwaukee.
Assumed $8/W PV, $0.1520/kWh, 8% discount rate, 30 years.
102
Goal 1
Net Primary energy
Goal 2
25%
200.73 kWh/m2 Net Primary energy
Window Typology
Glazing Type
Glazing to Wall Ratio
Building Length
Orientation
Wall Insulation R Value
Single
Clear
37.5
24.7
N-S
2.7
Roof Insulation R Value 3.6
Bitumen
Roof Type
Thermal Mass
Low
HVAC
Mechanical
Overhang Depth
1.2
Always-on
Lighting Control
PV
17.7
CCHP
9.0
PV
PV or CCHP
Capital Cost
$24.31
Total Cost
$-38.56
50%
116.56 kWh/m2
Window Typology
Triple
Low-e
Glazing Type
Glazing to Wall Ratio 12.8
Building Length
30.5
N-S
Orientation
Wall Insulation R Value 3.8
Roof Insulation R Value 5.3
Bitumen
Roof Type
Thermal Mass
HVAC
Overhang Depth
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
Zero
Mechanical
10.4
Always-on
22.5
8.6
PV
$70.61
$-35A7
75%
Goal 3
68.11 kWh/m2
Net Primary energy
Triple
Window Typology
Low-e
Glazing Type
Glazing to Wall Ratio 21A
14.9
Building Length
NW-SE
Orientation
Wall Insulation R Value 4.1
Roof Insulation R Value 5.7
Goal 4 100%
Bitumen
Roof Type
No solutions found
Low
Thermal Mass
Mechanical
HVAC
10.2
Overhang Depth
Always-on
Lighting Control
PV
CCHP
PV or CCHP
Capital Cost
Total Cost
29.4
62
PV
$104.99
$-25.96
Because you've specified a PV system, only Bitumen roofs are available.
Table 3.7:
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
103
Not Pdmry EUrYcalmpaI coal
400
I
a
a
200
100
*
.
0
100
-100
300
200
CePAWCot Above atmie ($p"?)
N Au SUWOgs
SelecOd Buldt-g
U lsselns
TowI C.sVcapt
Cast
120
s0
U
10
40
0
0s 00
40
y a,
qs1 .qJ
e
too
.100
"-O--
Nt PrnmArY/To"aI
CMe Cog ADo"etase
(Sn12)
0 soww &AWtu
BsmaI--
ct
400
I
300
*
%
0*
0
0
S
*
40
0 40
T7taWCoO
0 AhJ
9ulfrrj
40
AbO 01&e (4W2)
E Sei0ed 8.,hlln
0
120
N Bmsne
Figure 3-10: Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.1520/kWh, 8% discount rate, 30 years.
104
50%
Goal 2
Goal 1
25%
91.80 kWh/m2
Net Primary energy
91.80 kWh/m2 Net Primary energy
Double
Window Typology
Window Typology
Double
Low-e
Glazing Type
Low-e
Glazing Type
Glazing to Wall Ratio 10.5
Glazing to Wall Ratio 10.5
193
Building Length
Building Length
19.5
N-S
Orientation
Orientation
N-S
Wall Insulation R Value 2A
Wall Insulation R Value 2.4
Roof Insulation R Value 3.0
Roof Insulation R Value 3.0
Bitumen
Roof Type
Roof Type
Bitumen
High
'hermal Mass
Thermal Mass
High
Mechanical
HVAC
HVAC
Mechanical
7.8
Overhang Depth
Overhang Depth
7.8
Always-on
Lighting Control
Always-on
Lighting Control
28.7
PV
PV
28.7
14.4
14A
CCHP
CCHP
PV or CCHP
PV
PVorCCHP
PV
$81.92
$81.92
Capital Cost
Capital Cost
$-15232
Total Cost
Total Cost
$-152.52
Because youve specified a PV system, only Bitumen roofs are available.
Table 3.8:
75%
Goal 3
60.30 kWh/m2
Net Primary energy
Window Typology
Triple
Low-e
Glazing Type
Glazing to Wall Ratio 22.2
27.2
Building Length
Orientation
N-S
Wall Insulation R Value 3.7
Roof Insulation R Value 5.1
Goal 4 100%
Bitumen
Roof Type
No solutions found
High
Thermal Mass
Mechanical
HVAC
10.0
Overhang Depth
Dim-Indep
Lighting Control
28.5
PV
6.7
CCHP
PV
PV or CCHP
$121.26
Capital Cost
$-145.11
Total Cost
Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years.
105
Net Prknmy EneryIrapiaI Ca
400
.
I
0
*aO
0
200
100
0
!40
70
Cp'& Co*tAbo
U
Mi Buddfe.
210
6eftMe (Sm2)
estd Bunds
UBa*ene
Tab! cosWCapi Cost
0
a
,a
a
S
0
2
P.S.
*0
'3
0
)*..
40
-120
-100
140
70
0
.70
210
CapW&C4 Above 8ftee
(5M2)
mA Mie.
0 Um*ded fat*-,
N %*.."
Nut PrmrTotaI
Cost
400
I
S
0
*
S
S
S
S
S
* a.
200
100
S
-10
-120
-40
Above 8000099 $42)
uMf*sse N Si.ded 8"o
U
Tbta
0
80
Cogt
.i.ne
Figure 3-11: Simulation results for a low-rise commercial building in Milwaukee.
Assumed $4/W PV, $0.30/kWh, 8% discount rate, 30 years.
106
Chapter 4
Summary of Results
From the initial Boston runs, as expected, it is clear that it is still quite expensive to
achieve ZNE when PV is the only DG choice. Even with tax breaks over 30 years,
the overall cost of the building is more expensive than the basecase. When energy
efficiency measures are implemented, then PV is added, the buildings become more
affordable. With tax breaks, the payback takes about 10 years (30 years real price).
However, these results are highly dependent on factors such and the kWh price and
discount rate.
In the updated online implementation, we explored both PV and CCHP as DG
choices. In all cases, CCHP is chosen as the cost optimal solution for energy goals.
At current prices, CCHP is more cost effective than PV. Furthermore, as seen in the
Boston runs, CCHP is more cost effective than some efficiency options. In Boston,
both the 25% and 50% reduction from baseline energy goals had the same solution.
This means that the least cost option that meets the 25% goal also meets the 50%
goal. In this case, a 19.5 kW CCHP was picked. And in this case, the CCHP option
was less expensive than higher insulation and better windows. To reach the ZNE goal,
in addition to CCHP, the system also implemented better windows, higher insulation,
and independently dimming lights. In all cases, the total cost was negative.
As seen in the final PV runs in San Francisco and Milwaukee, the total cost optimal
solutions of the energy goals vary significantly. In San Francisco, it was possible to
reach ZNE, whereas solutions could not be found in Milwaukee.
107
Milwaukee has a
more severe climate than San Francisco, with both higher cooling and heating loads.
At base 65, San Francisco has 3042 heating degree days and 22 cooling degree days.
In contrast, Milwaukee has 7444 heating degree days and 1113 cooling degree days.
In addition, San Francisco has higher average solar radiation than Milwaukee. So, it
is not surprising that ZNE is more difficult to achieve in Milwaukee.
Furthermore, in San Francisco, even after 30 years, the energy savings would not
outweigh the additional capital cost required to reach ZNE at a PV price of $8/W.
In Milwaukee, at this price, only the 25% goal would result in a negative total cost.
In Milwaukee, it is interesting to note that the 25% solution varies significantly
between these three runs. At higher PV costs and lower grid prices, the 25% goal is
met with a 7.7 KW system. With the decreased PV price, the goal is met with a 17.7
KW system. When PV costs are decreased and the grid electricity price is increased,
the most cost effective way of meeting the 25% goal includes a 28.7 KW system.
This means that at current prices, the energy efficiency measures implemented in the
decision variables are more cost effective than PV. But, as PV costs go down or grid
prices go up, it may actually be cheaper to install more PV than it would be to invest
in more efficiency measures.
108
Chapter 5
Next Steps
5.1
ZNE Definition
In this tool, we chose a ZNE definition that looked at zero net source energy. However,
as mentioned previously, a clear definition of ZNE is still being debated within the
field. In future versions of the tools, it would be interesting to be able to optimize
around different definitions.
It is likely that a ZNE definition that relies on site
energy, cost, or emissions would have different solutions and it would be informative
to explore these differences.
Furthermore, within a given definition, more site specific data could be implemented.
For our purposes, we defined general grid efficiencies and site to source conversions.
In reality, these vary by location and utility. A more exact calculation would include
data the can vary by city.
5.2
Hourly Calculations
Developing an emulator to quickly predict heating, cooling, and energy use was a
critical step to allow this tool to run in a live, online environment. However, to do
this, we needed to aggregate the hourly energy outputs that Design Advisor normally
provides. It would be informative to explore the differences between hourly and annual
optimization runs. This was out of the scope of this study, since the energy emulator
109
and DG modules needed to be developed at an annual scale for calculation efficiency.
However, in doing so, some analysis is more difficult. For example, matching CCHP
electricity production with waste heat use would likely be more accurate with an
hourly output.
Also, as mentioned in the background, California is exploring the use of Time
Dependent Valuation (TDV) in their ZNE goals. The hope is that TDV can provide
some measure of the societal cost of energy (CEC 2011). Values vary seasonally, daily,
and hourly. So, again hourly calculations would be necessary. Furthermore, TDV has
only been developed in CA, so further research would be needed to create rates in
other areas of the country.
5.3
Neighborhood Scale
In our tool, we explored ZNE for single buildings. Another interesting area to consider
would be neighborhood scale ZNE, rather than single building scale. ZNE will be
easier to achieve in some building types than others. For example, given the roof
area and relatively low loads, it's likely that unrefrigerated warehouses could be net
producers. In contrast, reaching ZNE in high-rises, shaded areas, or hospitals will be
more challenging. Looking at ZNE from a neighborhood scale could allow these types
of buildings to balance one another.
5.4
Distributed Generation Options
For this tool, we explored PV and CCHP, but further developments could look at other
options, such as solar domestic hot water. Also, considering ZNE at a neighborhood
scale could expand the number of viable options. Building integrated wind does exist,
but is limited. At a neighborhood scale, a wind farm (or a PV farm) could be a viable
option. Also, with CCHP, district heating would be a good way to use larger, more
efficient generation while still having a use for the waste heat.
110
5.5
Other Building Types
This study analyzed low-rise commercial buildings. A similar methodology could be
implemented for the other building types in Design Advisor, such as high-rise and
low-rise residential and retail. The same procedure could be repeated to create the
regression equations for the emulator, but new cost models would need to be created.
5.6
Expanded Options
In addition to adding new DG options and other building types, other variables and
parameters could also be expanded. For example, the tool currently only looks at 2story buildings. ZNE becomes increasingly hard to achieve as buildings get taller, so
exploring the number of stories would be a useful addition. Also, the tool currently has
an option to add a cost multiplier to account for more or less expensive scenarios. The
user can also vary the PV, CCHP, electricity, and gas prices, along with the discount
rate. But, they cannot change the price of individual variables such as windows or
insulation. Adding more input options is another area for future improvement.
5.7
Retrofits
Finally, ZNE retrofits represent another area that warrants further study. In the
current tool, a retrofit can be approximated by adjusting the baseline building and
cost multiplier. But, a more exact analysis would require a new cost module for
retrofits. Additionally, the possible variables would be slightly different. This will
become an increasingly important area in the coming years, as ZNE goals begin to
be applied more extensively to existing buildings.
111
112
Chapter 6
Conclusions
As stated above, the goal of this research project was two-fold:
1. This tool is meant to act as a sketch model early in the design process. It has
been designed to be accessible to non-technical users.
2. This tool was also used to explore several case studies to provide general recommendations
for moving toward zero net energy systems.
We aimed to help users understand the path from a basecase building to a ZNE
building. Coming into the project, we were interested in looking at several aspects
of this process. First, to reach ZNE, some buildings rely heavily on PV, without first
exploring deep energy efficiency measures. Choosing between many design options
can be challenging and we wanted to better understand the effects of various inputs
and the resulting tradeoffs between cost and energy.
Furthermore, while there has been a recent push toward ZNE goals, we wanted
to understand whether going all the way the ZNE is the best choice right now, from
a cost and energy perspective. We believe this is a very strong goal and it certainly
can point the building industry in the right direction. But, it's possible that given
financial constraints, several buildings could be designed to reach goals somewhere
along the path for the same amount that it would cost to have a single building
achieve full ZNE.
Several conclusions can be drawn here:
1. From a source perspective, in some climates, it may not be possible to achieve
113
ZNE with onsite generation.
2. At current prices, efficiency measures are generally more cost effective than PV.
3. At current PV and grid prices, in San Francisco, the incremental cost of six
buildings that use 25% less energy than a typical low-rise commercial building
is less than the incremental cost of one ZNE building. In Milwaukee, 100%
could not be reached onsite, but reaching 75% requires more than 4 times the
capital cost of a 25% reduction.
4. At current PV and grid prices, in San Francisco, even after 30 years, for any
buildings that have more than 50% reduction, the total cost is greater zero. In
Milwaukee, this point is closer to 25% reduction.
5. While ZNE is an exciting goal, it is unlikely to be cost effective unless PV prices
decrease and/or grid prices increase. And, even if this occurs, building are likely
to still see additional capital costs to reach ZNE designs.
As explained in the Next Steps section, there are many additions that can be added
to make this a more robust tool. Additionally, a number of decisions still need to be
made about the future of ZNE goals-such as choosing a uniform definition. While
this tool will not provide perfect information because improvements can always be
made and there is uncertainty in the field, we hope that it will help users to start to
understand the many tradeoffs involved in building design.
114
Appendix A
Regression Coefficients
Below we include the current version of the regression coefficients. These values may
be updated as the tool progresses and we will include updated tables on the site.
Under the "split" column, "1" refers to the first tier, which includes the entire set
of buildings.
"2" refers to the second tier that splits the buildings into the top and
bottom half of energy use. Here, "a" is the lower half and "b" is the upper half.
Finally, "3" is the third tier that was split into quadrants.
from the lowest to the highest quadrants of energy use.
"a" through "d" range
Each of these also has a
heating, cooling, and lighting value.
The other columns include the coefficients from the regression analysis. For the
weather data, the tool matches the coefficient with the corresponding value from the
table of city weather data. For the decision variables, the tool matches the coefficient
to the building inputs for that run.
To approximate energy use, the emulator first calculates heating, cooling, and
lighting energy use based on the first tier of equations. Again, these are based on the
entire set of solutions. The emulator calculates whether the predicted solution is in
the top or bottom half of the energy use solutions and moves to the second tier of
equations. If the prediction from the first regression falls within the range of the lower
half of the energy from the full set of training data, in the second tier calculation,
the lower energy regression equation is used (2a).
If the predicted energy use falls
within the higher range, the higher energy regression equation is used in the second
115
tier (2b).
Again, heating, cooling, and lighting are calculated separately and can move
through tiers independently. So, a building could have a high cooling energy use and
a low lighting energy use. Based on the second round of calculations, the emulator
picks the closest quadrant of energy use and recalculates heating, cooling, and lighting
energy use. The three values are then summed for a final solution.
Table A. 1: Regression coefficients for weather parameters
Splits
x1
1 heating
1 cooling
1 lighting
2a heating
2a cooling
2a lighting
2b heating
2b cooling
2b lighting
3a heating
3a cooling
3a lighting
3b heating
3b cooling
3b lighting
3c heating
3c cooling
3c lighting
3d heating
weather data
Latitude
HDD
CDD
RH
Solar
1.49783934 0.03942049 0.02852081 1.13859327
0.0602968
1.07262203 -0.0065254 0.09734636 5.48441266 0.02747747
0.26784741 0.00014835 0.00148066 0.00276836 0.00513496
17.3064754
-0.0293987 0.04944797 -5.3703862
-0.0612535
-0.1428151
-0.0024077 0.05354676 0.85732447 0.01578923
0.53569481 0.00029669 0.00296132 0.00553673 0.01026992
2.32385283 0.02393693
-0.0281488
0 -0.0090745
1.07054783
-0.0116189 0.09982865 6.29114249 0.00187175
-2.47E-13
4.07E-16
-1.151E-15
9.43E-14
1.25E-15
0 0.01967668 0.00905872
0 -0.0096923
-0.7428046 0.00274354 0.02817919
0 0.04256095
1.33097666 0.00016787 0.00757993 0.37343699 0.02158263
0 0.04916075 0.14207249 -11.160922 0.41249631
1.4411317
-0.0100558 0.04915586 1.56630322
-0.0293858
0.50412558 0.00072337 0.00262882
-0.1146S2
0.0128162
0 0.02473447
-0.0172229
0 0.03394065
0.69317746 0.00191737 0.09636216 8.36061605 0.07516372
-0.0027076
-1.80E-06
-2.02E-05
-0.0017989
-4.30E-05
1.57336543 0.03030268
-0.027628
0 -0.0176835
13.3551434
-0.0110454 0.13336518
9.4651094 0.24125224
-7.62E-15
-3.61E-16
1.08E-15
2.OOE-14
-1.21E-15
3d cooling
3d lighting
116
Table A.2: Regression coefficients for decision variables (1 of 2)
x2
glazing to
wall
0.5409093
0.67382609
-0.0258173
0.13698278
0.19485428
-0.0471774
0.89442433
1.02847684
-0.0060054
0.02273649
0.08861643
-0.0383271
0.20725984
0.23746708
-0.0656204
0.44884917
0.61976286
-0.0001421
1.28061577
1.34595599
1.34E-15
x4
x5
ns wall
roof r value
-0.0063235
-0.0878879
0.00022131
-0.0045232
-0.0265389
0.00043132
-0.0099686
-0.1281121
9.50E-05
-0.0001584
-0.0143318
0.0004843
-0.0072536
-0.0310236
0.00070637
-0.0138304
-0.0685187
1.36E-05
-0.0003447
-0.1757554
1.OOE-15
-2.256201
-0.3830623
-3.76E-14
-0.7781751
-0.167128
0.00023567
-3.9560452
-0.6042969
0.00034506
-0.1892281
-0.0884182
-2.87E-06
-1.1376441
-0.2220142
0.00101388
-2.748318
-0.4598024
0.00023163
-4.3803323
-1.047754
1.74E-14
x6
overhang
depth
0.23905205
-0.8952989
0.00373556
0.08621987
-0.2563366
0.0074939
0.38935672
-1.371064
0.0011482
0.01983576
-0.0895969
0.00336944
0.11523644
-0.3404612
0.01123729
0.26255116
-0.7611993
0.00093022
0.49528245
-1.7456785
-1.45E-14
x7
window
glazing
-19.337893
-2.5893968
0.22684148
-4.2267032
0.30016791
0.41589671
-31.894864
-4.8542541
-0.0353285
-0.6495992
0.06594401
0.25165407
-6.3643999
0.34413507
0.613038
-14.79418
-1.7735088
0.02143048
-45.822093
-6.6670647
2.03E-14
x8
window
coating
-13.851214
-6.2627407
0.07149367
-3.1890924
-0.7109755
0.13607227
-22.696064
-10.18239
-0.0081769
-0.3384008
-0.3836971
0.09649993
-4.928993
-0.8959897
0.22612701
-9.5423719
-4.665459
0.00404474
-32.799361
-13.408024
-2.57E-14
Table A.3: Regression coefficients for decision variables (2 of 2)
x9
x10
thermal mass roof type
-0.0618967
-32.916798
-1.02E-13
-2.9567556
-11.150852
0.03741733
2.74676743
-48.094296
0.00058218
-2.440905
-5.3068946
-1.29E-05
-2.8893744
-13.510099
0.02099204
0.54063139
-31.907481
0.00154627
4.965854
-57.865873
-9.94E-14
0.39065692
-1.0759975
1.78E-13
0.2525479
-0.4031988
8.06E-04
0.55011977
-1.6527577
-0.0001941
0.07683596
-0.1833716
-6.46E-06
0.32028542
-0.5392703
1.09E-04
0.46280208
-1.0071181
8.37E-05
0.68911901
-2.8004961
4.66E-14
x12
x11
ventilation
type
orientation
0.83185024
-20.909764
-5.99E-14
0.47039291
-9.7495777
0.00604767
1.21868607
-26.641791
0.00103489
0.16663535
-4.2144678
-1.72E-05
0.59475043
-12.080727
0.05130947
1.0449854
-19.828471
0.00129125
1.18441178
-22.182529
-5.72E-14
117
0.1243864
0.19771986
-0.0312892
0.11867121
0.06396117
-0.0538555
0.17534321
0.41278551
0.01866656
0.02044207
0.0391581
-0.0630301
0.17902823
0.07160533
-0.0573668
0.2801421
0.27080679
-0.0020907
-0.122847
0.75902782
7.28E-14
x13
lighting
control
2.69122983
-4.0700581
-32.485854
1.12643321
-1.6388837
-33.070634
4.28956865
-6.0823017
-28.733901
0.31009518
-0.7673241
-29.460435
1.34925835
-2.2695074
-32.951559
3.20307053
-4.2308758
-27.650411
5.46774575
-9.1680875
0
intercept
-266.22279
-309.48143
101.910117
0
0
84.9373689
0
-282.44741
116.997911
0
0
0
0
0
88.8999511
0
-610.65653
115.767985
0
-1266.9536
87.8703704
118
Bibliography
[1] J. E. Allison and J. Lents. Encouraging distributed generation of power that
improves air quality: can we have out cake and eat it too?".
Energy Policy,
30:737-752, 2002.
[2] ANSI/AHRI
(Approved
American
National
Heating, and Refrigeration Institute).
conditioning
Standard/
Air-Conditioning,
Performance rating of unitary air-
air-source heat pump equipment.
Technical Report Standard
210/240 with Addendum 1, ANSI, AHRI, 2011.
[3] Arthur D. Little. Cooling, heating, and power for commercial buildings benefits
analysis. Technical report, April 2002.
[4] Doug Avery, J. Benya, M. Neils, F. Rubinstein, and D. Neils. Requirements
for controllable lighting: 2011 california building energy efficiency standards.
Technical report, Southern California Edison, March 2011.
[5] E. Bilgen. Exergetic and engineering analyses of gas turbine based cogeneration
systems. Energy, 25:1215-1229, 2000.
[6] Bluestein, J. et al. Combined heat and power white paper. Technical report,
Western Governors' Association, 2006.
[7] S. Brooks, M. Eldridge, and R. Elliott. Combined heat and power: Connecting
the gap between markets and utility interconnection and tariff practices (part
ii). Technical Report IE063, ACEEE, August 2006.
119
[8] S. Brooks, B. Elswick, and R. Elliott. Combined heat and power: Connecting
the gap between markets and utility interconnection and tariff practices (part i).
Technical Report IE062, ACEEE, March 2006.
[9] Burdick, J. et al. Temperature dependence of photovoltaic cell, modules, and
systems. 25th PVSC, May 1996.
[10] California Energy Commission. 2011 integrated energy policy report. Technical
Report CEC1002011001CMF, 2011.
[11] California Energy Commission. Deer - database for energy efficicient resources,
June 2012.
[12] S. Carter. Breaking the consumption habit: Ratemaking for efficient resource
decisions. The Electricity Journ, December 2001.
[13] T. Casten. Turning Off the Heat. Prometheus Press, 1998.
[14] T. Casten. future electricity generation
renewable energy. Technical report,
Recycled Energy Development, December 2009.
[15] G. Chicco and P. Mancarella. Trigeneration primary energy saving evaluation
for energy planning and policy development. Energy Policy, 35:6132 6144, 2007.
[16] C. Christiansen and R. Anderson.
Beopt software for building energy
optimization: Features and capabilities. Technical Report NREL/TP-550-39929,
US Department of Energy: Energy Efficiency and Renewable Energy, August
2006.
[17] Christiansen, C. et al. Beopt: Software for identifying optimal. ISES 2005 Solar
World Congress, 2005.
[18] M. De Paepe and D. Mertens. Combined heat and power in a liberalised energy
market. Energy Conversion and Management, 48:2542-2555, 2007.
[19] DOE (US Department of Energy). 2006 Building Energy Data Book. 2006.
120
[20] DOE (U.S. Department of Energy).
The potential benefits of distributed
generation and rate-related issues that may impede their expansion. Technical
report, 2007.
[21] DOE (U.S. Department of Energy). 2009 Buildings Energy Data Book. 2009.
[22] M. Drunsic, D. White, and R. Horby. Impacts of distributed generation on
wholesale electric prices and air emissions in massachusetts. Technical report,
Synapse Energy Economics, March 2008.
[23] R. Elliot and A. Shipley. Chp five years later: Federal and state policies and
programs update. Technical Report IE031, ACEEE, January 2003.
[24] Flager, F. et al. Multidisciplinary process integration and design optimization
of a classroom building. volume 0. ITcon, 2008.
[25] W. Gans, A. Shipley, and R. Elliot. Survey of emissions models for distributed
combined heat and power systems. Technical Report IE071, ACEEE, January
2007.
[26] David B. Goldstein, L. Burt, J. Horner, and N. Zigelbaum. Zeroing in on net-zero
buildings: Can we get there? how will we know when we have arrived? pages
125-137. ACEEE Summer Study on Energy Efficiency in Buildings, 2010.
[27] M. Green. Solar Cells. 1998.
[28] B. Griffith, P. Torcellini, N. Long, D. Crawley, and J. Ryan. Assessment of
the technical potential for achieving zero-energy commercial buildings. Number
NREL/CP-550-39830. ACEEE Summer Study on Energy Efficiency in Buildings,
2006.
[29] D. Hinrichs, S. Conbere, and M. Lobash.
Building Operating Management, July 2002.
121
Taking control of power supplies.
[30] R. Hirsh and B. Sovacool. Technical systems and momentum change: American
utilities, restructuring, and distributed generation technologies.
Journal of
Technology Studies, 32(2), 2006.
[31] T. Huld, M. Suri, and E. Dunlop. Geographical variation of the conversion
efficiency of crystalline silicon photovoltaic modules in europe.
Progress in
Photovoltaics: Research and Applications, 16:594-607, 2008.
[32] KEMA.
Market potential of combined heat and power in massachusetts.
Technical report, June 2008.
[33] R. Kenny, E. Dunlap, H. Ossenbrink, and H. Mullejans. A practical method for
the energy rating of c-si photo- voltaic modules based on standard tests. Progress
in Photovoltaics: Research and Applications, pages 155-166, 2006.
[34] X.Q. Kong, R.Z. Wang, and X.H. Huang. Energy optimization model for a cchp
system with available gas turbines. Applied Thermal Engineering, 25(2-3):377391, February 2005.
[35] M. Kushler, D. York, and P. Witte.
Aligning utility interests with energy
efficiency objectives: A review of recent efforts at decoupling and performance
incentives. ACEEE, 2006.
[36] J. Lee and C. Brown. Optimization of energy performance and capital cost of
an office building in boston. MIT 16.888 Final Report, 2008.
[37] M. Lehar. A branchingfuzzy-logic classifierfor building optimization. PhD thesis,
Massachusets Institute of Technology, 2005.
[38] M. Lehar and L. Glicksman. A heuristic tool for the rapid evaluation of building
performance.
[39] M. Lehar and L. Glicksman. Rapid algorithm for modeling daylight distribution
in office buildings. Building and Environment, 42(8):2908-2919, 2007.
122
[40] R. Madlener and M. Wickhart. The economics of cogeneration techology and
adoption diffusion:
A deterministic model.
Technical report, CEPE ETH,
December 2003.
[41] Makrides, G. et al. Outdoor efficiency of different photovoltaic systems installed
in cyprus and germany.
[42] A. Malik and S. Damit. Outdoor testing of single cell crystal silicon solar cells.
Renewable Energy, 28:1433-1445, 2003.
[43] K. Malmedal, B. Kroposki, and P. K. Sen. The energy policy act of 2005 and its
impact on distributed generation.
[44] A. Martens. The energetic feasibility of chp compared to the separate production
of heat and power. Applied Thermal Engineering, 18:935-946, 1998.
[45] Massachusetts Zero Net Energy Buildings Task Force. Getting to zero: Final
report of massachusetts zero net energy buildings task force. Technical report,
March 2009.
[46] E. Meyers and M. G. Hu. Clean distributed generation: Policy options to promote
clean air and reliability. The Electricity Journal,January 2001.
[47] Midwest CHP Application Center. Does cogeneration make sense for me? the
first step in examining the cogeneration opportunity. Technical report.
[48] E. Minciuc, 0. Corre, V. Athanasovici, and M. Tazerot. Fuel savings and co2
emission for trigeneration systems. Applied Thermal Engineering, 23(1333-1346),
2003.
[49] Minciuc, E. et al. Thermodynamic analysis of tri-generation with absorption
chilling machine. Applied Thermal Engineering, 23(1391-1405), 2003.
[50] C. Mone, D. Chau, and P. Plelan. Economic feasibility of combined heat and
power and absorption refrigeration with commercially available gas turbines.
Energy Conversion and Management, 42:1559-1573, 2001.
123
[51]
L. Montoya.
The road still not taken: How combined heat and power can
contribute to a sustainable energy future inmassachusetts.
Master's thesis,
Massachusets Institute of Technology, 2008.
[52] MREA (Medical Real Estate Advisors). Office building construction costs: U.s.
cities compared.
[53] NNEC
(Network
grid:
Best
policies
and
for
and
New
worst
Energy
Choices).
practices
interconnection
standards.
in
Freeing
state
net
the
metering
Technical
report,
http://www.newenergychoices.org/uploads/FreeingTheGrid2008,eport.pdf, 2008.
[54] H. Onovwiona and V. Ugursal.
Residential cogeneration systems: review of the
current technology. Renewable and Sustainable Energy Reviews, 10:389-431, 2006.
[55] Optimal Energy, Inc.
Economically achievable energy efficiency potential in new
england. Technical report, Northeast Energy Efficiency Partnerships, Inc., May 2005.
[56] Reed Construction Data. Building Construction Cost Data 2010 Book. 2010.
[57] REN21. Renewables 2007 global status report. Technical report, Renewable Energy
Policy Network for the 21st Century, 2007.
[58] W. Ryan. Economics of cogeneration. ASHRAE Journal,pages 34-39, October 2002.
[59] A. Siddiqui, C. Marnay, 0. Bailey, and K. LaCommare. Optimal selection of onsite generation with chp applications. International Journal of Distributed Energy
Resources, 1(1):33-62, 2005.
[60] SOLAIR, July 2012.
[61] B. Sovacool. Distributed generation and the american electric utility system: What
is stopping it? Journal of Energy Resources Technology, 130, March 2008.
[62] Sustainable Buildings Industry Council. Energy-10: A Powerful Energy Simulation
Tool for Buildings and Homes, 2007.
124
[63] TIAX. Technology review and assessment of distributed energy resources: Distributed
generation.".
Technical Report Product ID: 053828, Electric Power Research
Institute.
[64] P. Torcellini, S. Pless, M. Deru, and D. Crawley. Zero energy buildings: A critical
look at the definition. Number NREL/CP-550-39833. ACEEE Summer Study on
Energy Efficiency in Buildings, 2006.
[65] B. Urban and L. Glicksman. The mit design advisor - a fast, simple building design
tool. 1st IBSPA-Germany Conference, October 2006.
[66] B. Urban and L. Glicksman. A rapid building energy model and interface for nontechnical users. 10th ORNL Thermal Performance of the Exterior Envelopes of Whole
Buildings International Conference, December 2007.
[67] US EPA (U.S. Environmental Protection Agency).
Combined heat and power
partnership.
[68] M. Wetter. Simulation-Based Building Energy Optimization. PhD thesis, University
of California, Berkeley, 2004.
[69] D. Wu and R. Wang. Combined cooling, heating and power: A review. Progress in
Energy and Combustion Science, 32:459-495, 2006.
[70] Y. Li X.H. Huang X.Q. Kong, R.Z. Wang. Optimal operation of a micro-combined
cooling, heating, and power system driven by a gas engine. Energy Conversion and
Management, 50(3):530-538, March 2009.
[71] R. Zogg, K. Roth, and J. Broderick. Using chp systems in commercial buildings.
ASHRAE Journal,47(9):33-35, September 2005.
125