EPA-600/R-96-130
November 1996
DEMONSTRATION OF THE
ENVIRONMENTAL AND DEMAND-SIDE
MANAGEMENT BENEFITS OF GRID-
CONNECTED PHOTOVOLTAIC POWER
SYSTEMS
by
Edward C. Kern, Jr.
Daniel L. Green berg
Ascension Technology. Inc.
P.O. Box 314
Lincoln Center, MA 01773
EPA Contract 68-D2-0148
EPA Project Officer: Ronald J. Spiegel
Air Pollution Prevention and Control Division
National Risk Management Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Prepared for:
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D.C. 20460
-------
FOREWORD
The U.S. Environmental Protection Agency is charged by Congress with pro-
tecting the Nation's land, air, and water resources. Under a mandate of national
environmental laws, the Agency strives to formulate and implement actions lead-
ing to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, EPA's research
program is providing data and technical support for solving environmental pro-
blems today and building a science knowledge base necessary to manage our eco-
logical resources wisely, understand how pollutants affect our health, and pre-
vent or reduce environmental risks in the future.
The National Risk Management Research Laboratory is the Agency's center for
investigation of technological and management approaches for reducing risks
from threats to human health and the environment. The focus of the Laboratory's
research program is on methods for the prevention and control of pollution to air,
land, water, and subsurface resources; protection of water quality in public water
systems; remediation of contaminated sites and groundwater; and prevention and
control of indoor air pollution. The goal of this research effort is to catalyze
development and implementation of innovative, cost-effective environmental
technologies; develop scientific and engineering information needed by EPA to
support regulatory and policy decisions; and provide technical support and infor-
mation transfer to ensure effective implementation of environmental regulations
and strategies.
This publication has been produced as part of the Laboratory's strategic long-
term research plan. It is published and made available by EPA's Office of Re-
search and Development to assist the user community and to link researchers
with their clients.
E. Timothy Oppelt, Director
National Risk Management Research Laboratory
EPA REVIEW NOTICE
This report has been peer arid administratively reviewed by the U.S. Environmental
Protection Agency, arid approved for publication. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Information
Service, Springfield, Virginia 22161.
-------
Abstract
This project investigated the pollutant emission reduction and demand-side management potential of 16
photovoltaic (PV) systems installed across the country in 1993 and 1994. The project was sponsored by
the U.S. EPA and 11 electric utilities. This report presents analyses of each system's ability to offset
emissions of S02, NOx, C02 and particulates, and to provide power during peak load hours for the
individual host building and the utility. Results of simulations of battery storage systems powered by
each of the PV systems are also presented.
The analysis indicates a very broad range in the systems' abilities to offset pollutant emissions, due to
variation in the solar resource available and the marginal emission rates of the participating utilities. Use
of dispatchable storage would reduce emission offsets due to energy losses in charging and discharging
the batteries. Each system's ability to reduce building peak loads was dependent on the correlation of
that load to the available solar resource. Most systems operated in excess of 50 percent of their capacity
during building peak load hours in the summer months, but well below that level during winter peak
hours. Similarly, many of the systems operated above 50 percent of their capacity during utility peak
load hours in the summer months, but at a very low level during winter peak hours. The addition of
dispatchable energy storage significantly increases each system's peak load matching abililty, raising
capacity factors to 100 percent for most systems during the utility's highest load hours.
This document is copyrighted in its entirety by the author. ©1996 Ascension Technology. Since this
work was, in part, funded by the U.S. Government, the Government is vested with a royalty-free, non-
exclusive, and irrevocable license to publish, translate, reproduce, and deliver this information and to
authorize others to do so.
ii
-------
Contents
Abstract jj
Executive Summary ES-1
Chapter 1 Introduction i-i
References 1-3
Chapter 2 FV Systems 2-1
2.1 System Design 2-1
2.2 System Installation 2-6
2.3 Cost Summary 2-18
Chapter 3 Data Collection, Storage, Retrieval, and Review 3-1
3.1 Data Collection 3-1
3.2 Daily Data Retrieval and Review 3-2
3.3 Monthly Data Processing and Reporting 3-3
Chapter 4 PV System Performance History 4-1
Chapter 5 Dispatchable Battery Storage Mode! 5-1
5.1 Modeling Approach 5-1
5.2 Description of Model 5-2
Chapter 6 Host Building Load Impacts 6-1
6.1 Building Load and PV System Data 6-1
6.2 Data Analysis 6-1
6.3 Results 6-2
6.4 Conclusions 6-10
Chapter 7 Utility Coincident Peak Load Reduction 7-1
7.1 Pittsburgh, NY (EP AO 1) 7-3
7.2 Berlin, CT (EPA02) 7-3
7.3 Pieasantville and Brigantine, NJ (EPA03 and EPA04) 7-4
7.4 White Plains, NY (EPA05) 7-5
7.5 Scottsdale, Peoria, and Flagstaff, AZ (EPA06, EPA07, and EPA13) 7-6
7.6 Ashwaubenon and Denmark, WI (EPA08 and EPA09) 7-7
7.7 Minnetonka, MN (EPA10) 7-8
7.8 San Ramon, CA (EPA11) 7-9
7.9 Austin, TX (EPA 12) 7-9
7.10 Barstow, Edwards AFB, and Palm Desert, CA (EPA 14, EPA 15, and EPA 16) 7-10
7.11 Conclusions 7-11
Chapter 8 Calculation of Emission Offsets 8-1
8.1 Case Study Approach 8-1
8.2 Development of Marginal Emission Models 8-2
8.3 Results 8-3
8.4 Conclusions 8-14
iii
-------
Chapter 9 Conclusions 9-1
9.1 PV System Hardware 9-1
9.2 PV System Operation and Maintenance 9-3
9.3 Host Building Load Impacts 9-4
9.4 Utility Coincident Peak Load Reduction 9-4
9.5 Emissions 9-5
Appendix A. Quality Assurance Project Plan A-l
Appendix B. Quality Control Evaluation Report B-l
Appendix C. Operation of the Rotating Shadovvband Pyranometer C-l
Appendix D. Events Affecting PV System Performance D-l
iv
-------
Figures
Figure ES-1, Four kW PV installation in Brigantine, NJ ES-3
Figure ES-2. Annual S02 offsets ES-4
Figure ES-3. Annual N0X offsets ES-4
Figure ES-4. Annual CO, offsets ES-4
Figure ES-5. Annual particulate offsets ES-4
Figure ES-6. Example utility load duration and cumulative average capacity factor curves ES-6
Figure 2-1. Schematic of a 4-kW PV System 2-2
Figure 6-1. Example of PV Capacity Factor at Peak Building Load 6-2
Figure 6-2. Example of Building Load Duration Curves 6-3
Figure 6-3. Average Building Load Factor and PV Capacity Factor 6-3
Figure 6-4 Building Load Duration Curves with and without PV for Pittsburgh, NY 6-12
Figure 6-5 PV Capacity Factor at Monthly Peak Load for Pittsburgh, NY 6-13
Figure 6-6 Average Building Load Factor and PV Capacity Factor for Pittsburgh, NY 6-13
Figure 6-7 Building Load Duration Curves with and without PV for Berlin, CT 6-14
Figure 6-8 PV Capacity Factor at Monthly Peak Load for Berlin, CT 6-15
Figure 6-9 Average Building Load Factor and PV Capacity Factor for Berlin, CT 6-15
Figure 6-10 Building Load Duration Curves with and without PV for Pleasantville, NJ 6-16
Figure 6-11 PV Capacity Factor at Monthly Peak Load for Pleasantville, NJ 6-17
Figure 6-12 Average Building Load Factor and PV Capacity Factor for Pleasantville, NJ 6-17
Figure 6-13 Building Load Duration Curves with and without PV for Brigantine, NJ 6-18
Figure 6-14 PV Capacity Factor at Monthly Peak Load for Brigantine, NJ 6-19
Figure 6-15 Average Building Load Factor and PV Capacity Factor for Brigantine, NJ 6-19
Figure 6-16 Building Load Duration Curves with and without PV for White Plains, NY 6-20
Figure 6-17 PV Capacity Factor at Monthly Peak Load for White Plains, NY' 6-21
Figure 6-18 Average Building Load Factor and PV Capacity Factor for White Plains, NY 6-21
Figure 6-19 Building Load Duration Curves with and without PV for Scottsdale, AZ 6-22
Figure 6-20 PV Capacity Factor at Monthly Peak Load for Scottsdale, AZ 6-23
Figure 6-21 Average Building Load Factor and PV Capacity Factor for Scottsdale, AZ 6-23
Figure 6-22 Building Load Duration Curves with and without PV for Peoria, AZ 6-24
Figure 6-23 PV Capacity Factor at Monthly Peak Load for Peoria, AZ. 6-25
Figure 6-24 Average Building Load Factor and PV Capacity Factor for Peoria, AZ 6-25
Figure 6-25 Building Load Duration Curves with and without PV for Ashwaubenon, WT 6-26
Figure 6-26 PV Capacity Factor at Monthly Peak Load for Ashwaubenon, WT 6-27
Figure 6-27 Average Building Load Factor and PV Capacity Factor for Ashwaubenon, WT 6-27
Figure 6-28 Building Load Duration Curves with and without PV for Denmark, WI 6-28
Figure 6-29 PV Capacity Factor at Monthly Peak Load for Denmark, WT 6-29
Figure 6-30 Average Building Load Factor and PV Capacity Factor for Denmark, WT 6-29
Figure 6-31 Building Load Duration Curves with and without PV for Minnetonka, MN 6-30
v
-------
Figure 6-32 PV Capacity Factor at Monthly Peak Load for Minnetonka, MN 6-31
Figure 6-33 Average Building Load Fator and PV Capacity Factor for Minnetonka. MN 6-31
Figure 6-34 Building Load Duration Curves with and without PV for San Ramon, CA 6-32
Figure 6-35. PV Capacity Factor at Monthly Peak Load for San Ramon, CA 6-33
Figure 6-36. Average Building Load Factor and PV Capacity Factor for San Ramon, CA 6-33
Figure 6-37 Building Load Duration Curves with and without PV for Austin, TX 6-34
Figure 6-38. PV Capacity Factor at Monthly Peak Load for Austin, TX 6-35
Figure 6-39. Average Building Load Factor and PV Capacity Factor for Austin, TX 6-35
Figure 6-40 Building Load Duration Curves with and without PV for Flagstaff, AZ 6-36
Figure 6-41. PV Capacity Factor at Monthly Peak Load for Flagstaff, AZ 6-37
Figure 6-42. Average Building Load Factor and PV Capacity Factor for Flagstaff, AZ 6-37
Figure 6-43 Building Load Duration Curves with and without PV for Barstow, CA 6-38
Figure 6-44. PV Capacity Factor at Monthly Peak Load for Barstow, CA 6-39
Figure 6-45. Average Building Load Factor and PV Capacity' Factor for Barstow, CA 6-39
Figure 6-46 Building Load Duration Curves with and without PV for Edwards AFB, CA 6-40
Figure 6-47. PV Capacity Factor at Monthly Peak Load for Edwards AFB, CA 6-41
Figure 6-48. Average Building Load Factor and PV Capacity Factor for Edwards AFB, CA 6-41
Figure 6-49 Building Load Duration Curves with and without PV for Palm Desert, CA 6-42
Figure 6-50. PV Capacity Factor at Monthly Peak Load for Palm Desert, CA 6-43
Figure 6-51. Average Building Load Factor and PV Capacity Factor for Palm Desert, CA 6-43
Figure 7-1. Example Utility Load Duration Curve with Cumulative Average Capacity Factors 7-1
Figure 7-2. Utility Load and Cumulative Average PV Capacity Factor for Pittsburgh. NY 7-12
Figure 7-3. Utility Load and Cumulative Average PV Capacity Factor for Berlin, CT 7-14
Figure 7-4. Utility* Load and Cumulative Average PV Capacity Factor for Pleasantville, NJ 7-16
Figure 7-5. Utility Load and Cumulative Average PV Capacity Factor for Brigantine, NJ 7-18
Figure 7-6. Utility Load and Cumulative Average PV Capacity Factor for White Plains, NY 7-20
Figure 7-7. Utility Load and Cumulative Average PV Capacity Factor for Scottsdaie, AZ 7-22
Figure 7-8. Utility Load and Cumulative Average PV Capacity Factor for Peoria, AZ. 7-24
Figure 7-9. Utility Load and Cumulative Average PV Capacity Factor for Flagstaff, AZ 7-26
Figure 7-10. Utility Load and Cumulative Average PV Capacity Factor for Ashwaubenon, WT 7-28
Figure 7-11. Utility Load and Cumulative Average PV Capacity Factor for Denmark, WI 7-30
Figure 7-12. Utility Load and Cumulative Average PV Capacity Factor for Minnetonka, MN 7-32
Figure 7-13. Utility Load and Cumulative Average PV Capacity Factor for San Ramon, CA 7-34
Figure 7-14. Utility Load and Cumulative Average PV Capacity Factor for Austin, TX 7-36
Figure 7-15. Utility Load and Cumulative Average PV Capacity Factor for Barstow, CA 7-38
Figure 7-16. Utility Load and Cumulative Average PV Capacity Factor for Edwards AFB, CA. . . . 7-40
Figure 7-17. Utility Load and Cumulative Average PV Capacity Factor for Palm Desert, CA 7-42
Figure 8-1. Example Offset Chart 8-4
Figure 8-2. Emission Offsets for Pittsburgh, NY (EPA01) 8-17
Figure 8-3. Offsets for Berlin, CT (EPA02) 8-18
Figure 8-4. Emission Offsets for Pleasantville, NJ (EPA03) 8-19
Figure 8-5. Emission Offsets for Brigantine, NJ (EPA04) 8-20
Figure 8-6. Emission Offsets for White Plains, NY (EPA05) 8-21
Figure 8-7. Emission Offsets for Scottsdaie, AZ (EPA06) 8-22
Figure 8-8. Emission Offsets for Peoria, AZ (EPA07) 8-23
Figure 8-9. Emission Offsets for Flagstaff. AZ (EPA13) 8-24
vi
-------
Figure 8-10. Emission Offsets for Ashwaubenon, WI (EPA08) 8-25
Figure 8-11, Emission Offsets for Denmark, Wl (EPA09) 8-26
Figure 8-12. Emission Offsets for Minnetonka, MN (EPA 10) 8-27
Figure 8-13. Emission Offsets for San Ramon, CA (EPA11) 8-28
Figure 8-14. Emission Offsets for Austin, TX (EPA12) 8-29
Figure 8-15. Emission Offsets for Barstow, CA (EPA14) 8-30
Figure 8-16. Emission Offsets for Edwards AFB. CA (EPA 15) 8-31
Figure 8-17. Emission Offsets for Palm Desert, CA (EPA 16) 8-32
vii
-------
Tables
Table ES-1. SYSTEM SIZE AND COST F.S-3
Table 2-1. PV MODULE AND PANEL ASSEMBLY CHARACTERISTICS 2-3
Table 2-2. 4kW PV ARRAY CHARACTERISTICS 2-4
Table 2-3. POWER CONDITIONER CHARACTERISTICS 2-5
Table 2-4. INSTRUMENTATION OF PV SYSTEMS 2-6
Table 2-5. HARDWARE COMPONENT COSTS 2-6
Table 2-6. SYSTEM COST SUMMARY 2-19
Table 4-1. SUMMARY OF EVENTS AFFECTING SYSTEM GENERATION 4-1
Table 8-1. EMISSION RATE CALCULATION METHODOLOGIES . 8-2
Table 8-2. MONTHLY AND ANNUAL OFFSETS FOR THE AVERAGE U.S. PV SYSTEM 8-5
Table 8-3. MONTHLY AND ANNUAL OFFSETS FOR EPA01 8-5
Table 8-4. MONTHLY AND ANNUAL OFFSETS FOR EPA02 8-6
Table 8-5. MONTHLY AND ANNUAL OFFSETS FOR EPA03 8-7
Table 8-6. MONTHLY AND ANNUAL OFFSETS FOR EPA04 8-7
Table 8-7. MONTHLY AND ANNUAL OFFSETS FOR EPA05 8-8
Table 8-8. MONTHLY AND ANNUAL OFFSETS FOR EPA06 8-9
Table 8-9. MONTHLY AND ANNUAL OFFSETS FOR EPA07 8-10
Table 8-10. MONTHLY AND ANNUAL OFFSETS FOR EPA 13 8-10
Table 8-11. MONTHLY AND ANNUAL OFFSETS FOR EPA08 8-11
Table 8-12. MONTHLY AND ANNUAL OFFSETS FOR EPA09 8-11
Table 8-13. MONTHLY AND ANNUAL OFFSETS FOR EPA 10 8-12
Table 8-14. MONTHLY AND ANNUAL OFFSETS FOR EPA11 8-12
Table 8-15. MONTHLY AND ANNUAL OFFSETS FOR EPA 12 8-13
Table 8-16. MONTHLY AND ANNUAL OFFSETS FOR EPA 14 8-14
Table 8-17. MONTHLY AND ANNUAL OFFSETS FOR EPA 15 8-14
Table 8-18. MONTHLY AND ANNUAL OFFSETS FOR EPA 16 8-14
Table 8-19. ANNUAL OFFSETS AND CAPACITY FACTORS BY SITE 8-16
viii
-------
Acronyms and Abbreviations
ACE
Atlantic City Electric
APS
Arizona Public Service
BOS
Balance-of-system
CACF
Cumulative average capacity factor
C02
Carbon dioxide
COA
City of Austin Electric Department
DSM
Demand-side management
ETL
Edison Testing Laboratory
HVAC
Heating, ventilation and air conditioning
kW
Kilowatt
kWh
Kilowatt-hour
LDC
Load duration curve
MW
Megawatt
MWh
Megawatt-hour
NOx
Nitrogen oxides
NSP
Northern States Power
NU
Northeast Utilities
NYPA
New York Power Authority
NYSEG
New York State Electric and Gas
O&M
Operation and maintenance
PG&E
Pacific Gas & Electric
POA
Plane-of-array irradianee
PV
Photovoltaic
RSP
Rotating shadowband pyranometer
SCE
Southern California Edison
S02
Sulfur dioxide
SOC
Standard operating conditions
STC
Standard test conditions
T&D
Transmission and distribution
UL
Underwriters Laboratories
WPS
Wisconsin Public Service
IX
-------
Executive Summary
Introduction
Photovoltaic (PV) conversion of
sunlight to electricity has become
substantially less costly and more
efficient in recent years. Since its first
application in the space program in the
1950's, the cost of PV modules has
fallen approximately 70 percent per
decade, and module manufacturers
continue to make progress in reducing
costs further. Although technological
innovation has been responsible for
much of the decline in costs, an
international market for remote, off-
grid power, growing at the rate of 20 to
30 percent annually, has resulted in
expansion of module production
capacity. This, in turn, has led to
production economies which have
driven module prices down still
further.
Despite these cost reductions, modules
remain the dominant factor in the cost
of a grid-tied PV power system
accounting for approximately 70% of
the total. The power converter
(inverter), necessary for transforming
the DC power output from a PV array
to grid-synchronous AC power, is
another significant component of
system cost, accounting for about 15
percent of total cost. Because the
market for grid-tied AC power from
PV systems has been relatively small,
there has been little progress in
reducing the cost of the inverter.
However this project, and other similar
projects are increasing the demand for
inverters, and will likely result in
technological improvement and cost
reduction. The remaining cost
components of PV systems are the
array mounting structure, wiring, and
switchgear, collectively referred to as
the balance of system (BOS).
Although electricity generated by
photovoltaics remains too expensive to
compete with conventional power
sources in most grid-connected
applications, there is a growing niche
of cost-effective applications (most of
them remote from the power grid)
which will expand as the cost of PV
power falls. A 50 percent drop in
module prices is expected within the
decade which has the potential to
greatly expand the grid-connected
market. Against this background of
falling costs is a heightened public
awareness of the threats to
environmental quality posed by the by-
products of electricity production. The
most notable concern today is the
possibility that emission of carbon
dioxide resulting from the combustion
of fossil fuels may lead to climatic
changes on a global scale. As a result
of this heightened concern regarding
environmental quality, many
consumers have shifted their
consumption patterns, and some are
willing to pay premiums for products
that have lower environmental
ES-1
impacts. Several recent surveys
suggest that about half of electric
utility customers would be willing to
pay a $10 monthly premium for
electricity generated by renewable
resources. Given this context, it is
very likely that the domestic market
for grid-tied photovoltaic power
systems will expand substantially
within the next decade, and continue to
grow rapidly.
The potential environmental benefits
from PV power generation are quite
large. If PV systems were installed
where possible on the roof-tops of the
U.S. inventory of residential,
commercial, and industrial buildings,
they could produce roughly 20% of the
Nation's electricity. Currently, fossil
fuels used for electric power
generation in the U.S. account for
approximately 34% of the carbon
dioxide (C02), 67% of the sulfur
dioxide (S02), and 37% of the nitrogen
oxide (NOJ emissions into the
atmosphere from controllable sources
within the U.S..
In September 1991 the EPA issued a
solicitation for the installation of grid-
tied PV systems with the goal of
measuring their environmental and
demand-side benefits. Ascension
Technology developed a proposal in
response to this solicitation, with the
support and participation of utilities
across the nation. Eleven utilities
-------
supporting the proposal to EPA were
(I) New England Electric System
(NEES) with service areas in Rhode
Island, Massachusetts, and New
Hampshire; (2) New York State
Electric and Gas (NYSEG) in upstate
New York: (3) Northeast Utilities
(NU) with service areas in
Connecticut, Massachusetts, and New
Hampshire; (4) Atlantic City Electric
(ACE) in southern New Jersey; (5)
New York Power Authority (NYPA)
with customers throughout New York
State; (6) Arizona Public Service
(APS) in central and northern Arizona;
(7) Wisconsin Public Serv ice (WPS) in
southeastern Wisconsin; (8) Northern
States Power (NSP) with service areas
in Minnesota, Wisconsin, Michigan,
and the Dakotas; (9) Pacific Gas and
Electric (PG&E), serving most of
northern California; (10) the City of
Austin Municipal Utility (COA); and
(II) Southern California Edison (SCE)
serving much of southern California.
In addition to the geographic diversity
of the service areas represented by
these utilities, their pollutant emission
characteristics also proved to be quite
divergent. Ascension Technology's
partners from the PV industry were
Siemens Soiar Industries, which
provided PV modules, and Omnion
Power Engineering Corporation, which
provided the inverters.
EPA awarded the contract for this
project to Ascension Technology in the
third quarter of 1992. The final system
design effort began shortly thereafter,
and the first system was installed and
operating in April 1993. Ten of the
systems were operating by the end of
August 1993, and the last was
completed by mid-January 1994.
Monitoring of each system began
concurrently with initial system
operation, although the "official data
start date" was delayed at sites where
there were initial technical problems
with either instrumentation or PV
system hardware. At each site, 15-
minute average values of solar
irradiance, ambient temperature, PV
system power output, and building
load were recorded and stored for
subsequent retrieval by modem.
Monitoring of each site (for the
purposes of this study) continued
through September of 1994.
Emission rate and load data provided
by each of the participating utilities
were used in conjunction with the data
collected from each system to conduct
analyses of (1) the emission offsets
resulting from operation of the PV
systems: (2) the ability of each PV
system to reduce the peak power
demand of the building on which it
was installed; and (3) the
chronological correlation of each PV
system's power output to the
respective utility's peak loads. In
addition, a model was developed to
simulate the operation of each system
in conjunction with dispatchable
battery storage. This simulation
shifted each system's daily generation
to the utility's daily peak load hour(s),
thus increasing the peak load
correlation and reducing emission
offsets (due to battery charging and
discharging losses).
Chapter 1 of this report provides a
general introduction to the project.
Chapter 2 describes the design,
installation, and cost of each system.
Chapter 3 describes the data
acquisition system and presents data
collection and review procedures.
System performance history is
described generally in Chapter 4
(details are provided in appendix D).
Chapter 5 discusses the model used to
simulate the behavior of the PV
systems with dispatchable battery
storage. The marginal emission rate
models developed for each of the
participating utilities is described in
Chapter 8, as are the site-by-site
emission offset estimates. Chapter 6
discusses each system's impact on the
load of the building it is installed on,
and utility-level load matching results
are presented in Chapter 7.
Conclusions from this project are
summarized in Chapter 9.
Procedure
System Design
Designs were developed for nominal
4-kW "building block" PV systems for
this project, capitalizing upon project
staffs experience with roof-mounted
PV arrays in prior projects. The
majority of the project's sites use either
one system (4kW) or a group of three
systems (12 kW total). Note that the
nominal system size refers to the
inverter AC rating. The actual power
output of the PV systems under
standard operating conditions (1000
W/ni2 irradiance (full sunlight) and an
ambient temperature of 20°C) is
limited by the PV array to multiples of
3.5 kW AC.
PV arrays were configured using 12
PV panel assemblies. Each assembly
contains seven modules, electrically
wired in series. A PV source circuit is
formed with four PV panel assemblies,
wired in series. A 4-kW PVarrav is
comprised of three PV source circuits,
as shown in Figure ES-1.
ES-2
-------
Figure ES-1. A 4-kWPVinstallation in Brigantine, NJ
Table ES-1. System Size and
Cost
Cost"
Watt
1
NYSEG
12
101.9
9.69
2
NU
4
36.0
10.26
3
ACE
12
105.0
9.98
4
ACE
4
31.1
8.87
5
NYPA
4
35.2
10.04
6
APS
8
64.2
9
7
APS
4
31.3
8.94
8
WPS
12
101.0
9.61
9
WPS
4
31,2
8.91
10
NSP
• 4
38.0
10.83
11
PG&E
12
104.9
9.40
12
COA
12
97,5
9.27
13
APS
4
32.2
9.19
14
SCE
4
30.9
8.81
15
SCE
4
31.1
8.89
16
SCE
12
96.7
9.20
"Thousands of dollars.
Both pitched- and flat-roof installations
utilize Ascension Technology RoofJack
PV array supports, which have been
used to install more than 1 megawatt of
PV systems, PV arrays are held in
place by ballast on flat-roofs; this
approach requires no roof penetrations
for hold-down of the PV arrays.
System design details were developed
in close cooperation with Siemens
Solar Industries of Camarillo, CA. the
PV module supplier, Omnion Power
Engineering was selected as the
supplier of power conditioners. The
PV systems were designed to
accommodate the specifications of the
4kW-rated Omnion Series 2200 unit.
System Installation
System installation began in April
1993, and was complete by the end of
January 1994, although instrumentation
and hardware problems delayed the
initiation of monitoring at some sites.
The systems were installed on a variety
of residential, commercial, and
industrial buildings. Installation costs
for each system varied by system size
(4-, 8-, or 12-kW) and a number of site-
specific factors. Table ES-1
summarizes the size and cost of each
system.
PV System Performance History
Of the sixteen PV systems installed by
this project, all but two suffered events
during the study period which
temporarily limited system output or
prevented generation altogether.
Inverter-related problems were the most
vexing of the generation-limiting
events. In all, 27 inverter-related
events resulted in a generation loss of
12,740 kWh, approximately nine
percent of the combined generation of
these systems over the relevant time
periods.
As a result of the inverter-related
outages experienced in this project, the
inverter manufacturer made several
design changes and increased product
testing across their full line of inverters.
In addition, they extended the product
warrantee for the EPA project
installations.
Snow cover was also a frequent cause
of PV system outages for those systems
located in northern locations or at high
altitude. Of the systems in such
locations, the estimated energy loss as
a result of snow cover ranged from less
than one percent to 16 percent of
measured annual generation.
A variety of other outages occurred
during the study period, not all of
which have identified causes. Of those
"other" outages for which a cause was
identified, the most frequent was, by-
far, fuse failure in the DC disconnect
switch. Such failures occurred a total
of 17 times, at 11 of the sites. It was
determined that the original fuses in
the DC disconnect switches did not
have the proper surge rating. As they
failed, they were replaced by "slow-
blow" fuses which were rated for 600V
ES-3
-------
Annual Sulfur Dioxide Offset
20
^ 15
1 2 3 4 5 © 7 & 9 10 11 12 13 14 15 10
Site Number
Figure ES-2. Annual S02
offsets.
Figure ES-3. Annual NOx
offsets.
DC, None of the replacement fuses has
failed to date.
Battery Storage Model
Where peak loads do not coincide with
peaks in the available solar resource,
the value provided by a PV system can
sometimes be greatly enhanced by the
addition of dispatchable battery
storage. Although some energy is lost
both in charging and discharging a
battery array, the ability to dispatch
energy generated by the PV system
during utility peak loads (or for that
12345S7B9 10 1112 1314 1515
Site Number
Figure ES-4. Annual C02
offsets.
approach taken was to maximize the
contribution of each PV system during
the highest utility load hours of each
day, by simulating the daily operation
of each system's inverter(s) at its
(their) peak capacity for as long as
possible. The duration of operation
each day was determined by the
amount of energy actually generated
each day and the AC rating of the
inverter. The addition of dispatchable
battery storage affects both pollutant
emission offsets (due to battery
charging and discharging losses) and
the system's operation during utility
peak load hours.
I 04
0
H5
3
1 0,2
Mllll
ilk
1 2 3 4 5 8 ? S 9 1011 12 13 14 15 13
Site Number
Figure ESS. Annual Particulate
offsets.
matter the peak loads of a transmission
line or distribution feeder) allows the
PV generation to be used to reduce
generation by a utility's highest
operating cost units, which are
typically used only during peak
periods
To investigate the degree to which
dispatchable battery storage would
improve the ability of each PV system
to offset load during utility peak load
hours, a simple model was developed
to simulate battery charging,
discharging, and dispatch. The
ES-4
Results
Pollutant Emission Offsets
Models of marginal emission rates
(i.e., emission rates of load following
units) were developed for each utility
based on utility-provided data. The
hourly emission rates of sulfur dioxide,
nitrogen oxides, carbon dioxide, and
particulates were then combined with
hourly PV system generation data (and
simulated PV/storage dispatch data) to
determine hourly emission offsets.
Annual emission offsets are presented
in Figures ES-2 through ES-5. Annual
sulfur dioxide offsets ranged from 4
g/kW to 16 kg/kW of system rating
under standard operating conditions
(SOC) (1000 W/m2 irradiance and 20°
C ambient temperature). NOx offsets
ranged from 110 g/kW to 8.7 kg/kW.
The range in annual C02 emission
offsets was from 700 to 2,300 kg/kW
of system rating, and that for
particulates was 20 g/kW to 600 g/kW
annually. The lighter shaded area in
each figure is an estimate of the
pollutant offset achievable by a PV
system with average insolation, using
average U.S. emission rates based on
-------
data collected by the Energy
Information Administration for 1993,
The extreme variability in these results
is due to two factors: (1) variability in
the local solar resource and (2)
variability in utility marginal emission
rates. The second of these is far more
influential than the first, as can be seen
by comparing the range for COz to
those of the other pollutants. Since
there are currently no mitigation
measures in place for CO,, variation in
utility CO, emission rates is due only
to the relatively small (about two to
one) variation in the carbon content of
fuels used and variation in the heat
rates of the power plants. The range of
the highest to lowest annual offset is
relatively small at 3.3. For the other
pollutants, variations in the pollutant
content of the fuel as well as inter-
utility differences in installed pollution
mitigation equipment give rise to the
tremendous differences between utility
emission rates which underlie the
differences in emission offsets
described above.
The results of the PV-powered
dispatchable storage system
simulations indicate that pollutant
offsets would be reduced by at least
25% were storage added to these
systems. This is largely due to energy
losses in charging and discharging
batteries, but is also influenced by
marginal emission rates which are
typically lower during utility peak load
hours when cleaner, more efficient
power plants are often used to follow
load.
Building-Level Load Reduction
Analysis of each PV system's ability to
provide power during building peak
load hours was conducted by
comparing each building's net (of PV
generation) and gross load duration
curve (LDC). The LDC is constructed
by sorting all load values for a given
period in descending order, and
plotting each value against its rank in
the sort. Differences in a building's net
and gross LDC for the highest load
values indicate the PV system's ability
to reduce building peak loads.
As one would expect, reductions in net
building load were generally higher in
the summer months and lower in the
winter months, with the difference
being particularly pronounced for
systems installed in northern states.
Most systems reduced the building's
LDC by more than 50 percent of
system AC rating during the highest
load hours in the second and third
quarters of the year. In the winter
months, PV output during building
peak load hours dropped below ten
percent of rating for some systems,
although many of the systems in the
southern and western states performed
as well or even better during winter
peak load hours.
Two general conclusions may be
drawn from the analysis. The first is
the relatively self-evident conclusion
that if reduction of customer net
demand is the primary motivation for
the installation of a PV svstem, it is
* *
critical to investigate the correlation of
building peak loads to solar irradiance.
The set of host buildings participating
in this project included some with
loads which were very well matched to
the solar resource as well as some for
which the match was very poor. The
systems in Ashwaubenon, WI and
Scottsdale, AZ are examples of
systems which reduced host building
LDCs by a substantial fraction of their
SOC rating. The highest loads in these
buildings occurred during the midday
hours, when the solar resource peaks.
The systems in Barstow, CA and
Denmark, WI, on the other hand had
very little effect on the host building's
LDC, despite ample solar resource.
Many of the highest building loads at
these sites occurred near or after
sunset.
The second general conclusion to be
drawn from the data is that the
generation by a PV system during an
individual building's peak load hour
provides little information regarding
that system's ability to reduce the
building's peak monthly load, or to
reduce demand charges. Even if the
system generates at full power during
the monthly peak, there may be hours
during which building load is slightly
below the monthly peak and during
which the PV system operates at a
much lower level. In such cases there
may be very little change in the
building's net LDC and
correspondingly small changes in
demand charges. The monthly peak
load will have simply been shifted to
another hour.
Utility Coincident Peak Load
Reduction
Each PV system's ability to provide
power during utility peak load hours
was analyzed by simultaneously
sorting hourly PV generation data and
hourly utility load data in descending
order, with utility load level
determining the sort order. The result
was a utility load duration curve with
a value of PV generation for each
corresponding hour on the LDC. A
"cumulative average PV capacity
factor curve" (CACF curve) was then
created by dividing each hourly PV
ES-5
-------
generation value by the system's
capacity rating (resulting in hourly
capacity factors) and then averaging
each hour's capacity factor with the
capacity factors of all hours higher in
the sort-order (i.e. all hours in which
utility load was higher). The resulting
curve illustrates the FV system's
average capacity factor for the highest
n load hours, where n is read off the
ordinate.
By plotting this curve on the same axes
as the normalized LDC, one can
determine for each point on the LDC,
the average PV system capacity factor
for all hours up to and including that
hour. For example, the CACF curve
in Figure ES-6 indicates the PV
system's average capacity factor
during the utility's ten highest load
hours was about 40 percent. CACF
curves were calculated using both
measured PV system performance data
and the performance data generated by
the dispatchable storage simulation.
Charts displaying the utility LDC and
the PV system's CACF curves (both
with and without storage) were created
Third Quarter 1993
Figure ES-6. Example utility
load duration and cumulative
average capacity factor curves.
for each calendar quarter during the
study period. An additional chart
showing the same data for the 100
highest load hours encountered during
the study period was also created.
These charts provide a measure of each
PV system's peak shaving capacity.
Load Reduction Without Storage
Not surprisingly, load matching for PV
systems installed in northern states is
greatest in the spring and summer
months, with the capacity factor during
the highest load hours typically
averaging above 40 percent. Several of
these sites achieved capacity factors
well in excess of 60 percent of their
SOC rating during the highest load
hours in these months. The northern
systems invariably generated little or no
power during winter peak hours, most
or all of which occurred at night.
Utility peak loads in the southern and
western parts of the country invariably
occurred during the summer months
when the solar resource is greatest,
although these peaks consistently
occurred in the mid- to late-afternoon.
Most of the systems installed in these
regions operated at capacity factors in
excess of 40 percent during the highest
load hours in the summer months.
Some systems consistently operated at
capacity factors above 60 percent
during these hours. The one exception
to this is the system in Flagstaff, AZ.
which operated at only about 30
percent capacity factor during the peak
load hour. This low result is most
likely explained by the fact that the
load and weather patterns in Flagstaff
are quite different from those in
Phoenix which is about one mile lower
in elevation and 140 miles south.
Loads in the Phoenix area probably
dominate the Arizona Public Service
system load.
As did their counterparts in the
Midwest and Northeast, systems in the
southern and western states typically
operated at a lower level during winter
peak hours. With the exception of the
systems in southern California, systems
in the West operated at or near zero
percent capacity factor during peak
hours in the first quarter of the year.
Load Reduction With Storage
Except where the power output was
limited by a system outage, results from
the storage simulation indicate that
storage can provide system operation at
the full inverter rating during the peak
load hours in the summer months at all
sites. In regions (such as the Northeast
Utilities service area) where peak utility
loads are highly correlated to the solar
resource, the addition of a dispatchable
storage system may do little to improve
the PV system's load matching
capability, since it will already be quite
good. Systems in northern states are
much less able to provide power during
peak load hours in winter due to the
limited solar resource and snow cover.
Even with storage, some of these
systems were unable to provide power
at more than a few percent of inverter
rating during winter peak load hours.
However, daytime generation at other
northern sites was sufficient to allow
inverter operation well in excess of 50
percent of inverter rating during winter
peak hours.
Unlike many of the systems installed in
northern climates, the addition of
dispatchable storage to systems
installed in the southern and western
states would allow them to operate at
high capacity' factors during winter
ES-6
-------
peak hours. The results of the
simulation indicate that most of the
systems installed in this part of the
country would operate at or near 100
percent of inverter rating during the
highest winter load hours.
it is important to recognize that these
results are substantially determined by
the storage charging/dispatch
algorithm. An algorithm which stores
generation from one or more days and
dispatches only when load exceeds a
predetermined threshold (as opposed to
dispatching during the peak hours of
each day), might substantially improve
the load matching characteristics of all
systems.
Conclusions
This project has provided an initial
demonstration of the effectiveness of
grid-connected photovoltaic energy
systems in reducing the pollutant
emissions of electric utilities. The
broad range of emission offsets
achieved by these systems reflects
differences in both the available solar
resource at each site and differences in
emission rates among utilities. The
results demonstrate that the latter factor
is far more important in determining
the pollution mitigating potential of a
PV system than is the former. Given
current and projected costs of PV
systems, it is unlikely that this
technology will be employed solely for
its pollution mitigating potential.
While there is certainly substantial
value in this potential, PV's
environmental benefits must be
considered in conjunction with the
other benefits provided by the
technology for grid-connected
applications to be considered cost-
effective. These benefits include
conventional energy and power benefits
as well as more subtle and less well-
recognized advantages over central-
station generators.
This report documents case studies of
the peak load reduction benefits, for
utilities and for individual customers at
sites across the country. While PV will
not provide substantial power during
peak load periods at every location,
there are many at which it will, with or
without storage. If a PV system is
interconnected on the customer side of
the meter, this translates into energy-
and demand-charge savings. On the
utility side of the meter, distributed
generating resources such as PV which
provide power during peak load hours
can defer costly and under-utilized
additions to generation and
transmission capacity. In addition,
every kWh generated by a PV system
reduces utility fuel and variable O&M
costs.
As the electric utility industry enters the
world of retail competition, the high
cost of providing power during peak
hours is likely to be much more clearly
reflected in the prices paid by
consumers. The value provided by
resources such as PV that generate
power during such times is therefore
likely to increase substantially for
customers that cannot alter their
consumption patterns, and for utilities
hoping to retain such customers.
Retail competition at the generation
level will also bring the costs of
maintaining the transmission and
distribution (T&D) system under closer
scrutiny. Already, several studies have
demonstrated that such costs are not
homogeneous across a service area, but
are typically highly differentiated.
Communities in which load growth
necessitates an increase in the power
delivery capacity of local distribution
resources may have T&D costs many
times the average for the utility service
area. In such areas distributed
generating resources such as PV might
defer or eliminate the need for T&D
capacity additions, to the degree that
they are able to provide power at the
time when the existing distribution
system is stressed.
In addition to its environmental
energy, and capacity benefits,
photovoltaic technology possesses a
variety of characteristics which,
although less easily quantifiable,
contribute additional, real value.
Among these are (1) its reliance on a
limitless, indigenous resource, which
could reduce growing dependence on
imported oil; (2) its modularity and
speed of installation, allowing
generating capacity to be added as
needed rather than tying up large
amounts of capital in conventional
power plants, the need for which may
not materialize; (3) the relative ease of
siting PV power plants, as opposed to
the permitting hurdles and public
opposition that utilities typically
encounter in attempting to site
conventional power plants and
transmission lines; and (4) its ability to
fulfill consumers' desire for non-
polluting, renewable resources, which
may have strategic value to utilities in
addition to the environmental benefits
themselves.
Taken collectively, the benefits of grid-
connected photovoltaic power may
already outweigh its costs in some
applications. As PV costs continue to
decline, the range of such applications
is certain to grow, but much work
remains in the effort to fully quantify
ES-7
-------
the benefits of the technology. Projects
such as the one this report documents
are an essential component of that
effort.
ES-8
-------
Chapter 1
Introduction
Photovoltaic conversion of sunlight to electricity has become substantially less costly and more efficient
in recent years. Since its first application in the space program in the 1950's, the cost of PV modules has
fallen approximately 70 percent per decade, and module manufacturers continue to make progress in
reducing costs further. Although technological innovation has been responsible for much of the decline
in costs, an international market for remote, off-grid power, growing at the rate of 20 to 30 percent
annually, has resulted in expansion of module production capacity. This, in turn, has led to production
economies which have driven module prices down still further.
Despite these cost reductions, modules remain the dominant factor in the cost of a grid-tied PV power
system accounting for approximately 70 of the total. The power converter (inverter), necessary for
transforming the direct current (DC) power output from a PV array to grid-synchronous alternating
current (AC) power, is another significant component of system cost, accounting for about 15 percent of
total cost. Because the market for grid-tied AC power from PV systems has been relatively small, there
has been little progress in reducing the cost of the inverter. However this project, and other similar
projects are increasing the demand for inverters, and will likely result in technological improvement and
cost reduction. The remaining cost components of PV systems are the array mounting structure, wiring,
and switchgear, collectively referred to as the balance of system (BOS).
Although electricity generated by photovoltaics remains too expensive to compete with conventional
power sources in most grid-connected applications, there is a growing niche of cost-effective
applications (most of them remote from the power grid) which will expand as the cost of PV power falls.
A 50 percent drop in module prices is expected within the decade which has the potential to greatly
expand the grid-connected market. Against this background of falling costs is a heightened public
awareness of the threats to environmental quality posed by the by-products of electricity production. The
most notable concern today is the possibility that emission of carbon dioxide resulting from the
combustion of fossil fuels may lead to climatic changes on a global scale. As a result of this heightened
concern regarding environmental quality, many consumers have shifted their consumption patterns, and
some are willing to pay premiums for products that have lower environmental impacts. Several recent
surveys suggest that about half of electric utility customers would be willing to pay a $10 monthly
premium for electricity generated by renewable resources (Farhar. 1996; Oppenheim. 1995). Given this
context, it is very likely that the domestic market for grid-tied photovoltaic power systems will expand
substantially within the next decade, and continue to grow rapidly.
The potential environmental benefits from PV power generation are quite large. If PV systems were
installed where possible on the roof-tops of the U.S. inventory of residential, commercial, and industrial
1-1
-------
buildings, they could eventually produce roughly 20% of the Nation's electricity1. Currently, fossil fuels used
for electric power generation in the U.S. account for approximately 34% of the carbon dioxide (CO,), 67%
of the sulfur dioxide (SO,), and 37% of the nitrogen oxide (N£)) emissions into the atmosphere from
controllable sources within the U.S. (U.S. Department of Commerce, 1992).
In September 1991 the EPA issued a solicitation for the installation of grid-tied PV systems with the goal of
measuring their environmental and demand-side impacts. Ascension Technology developed a proposal in
response to this solicitation, seeking the support and participation of utilities across the nation. Eleven utilities
supporting the proposal to EPA were (1) "New England Electric System (NEES) with service areas in Rhode
Island, Massachusetts, and New Hampshire; (2) New York State Electric and Gas (NYSEG) in upstate New
York; (3) Northeast Utilities (NU) with service areas in Connecticut, Massachusetts, and New Hampshire; (4)
Atlantic City Electric (ACE) in southern New Jersey; (5) New York Power Authority (NYPA) with customers
throughout New York State; (6) Arizona Public Service (APS) in central and northern Arizona: (7) Wisconsin
Public Service (WPS) in eastern Wisconsin; (8) Northern States Power (NSP) with service areas in Minnesota,
Wisconsin, Michigan, and the Dakotas; (9) Pacific Gas and Electric (PG&E), serving most of northern
California; (10) the City of Austin Municipal Utility (COA); and (11) Southern California Edison (SCE)
serving much of southern California. In addition to the geographic diversity of the service areas represented
by these utilities, their pollutant emission characteristics also proved to be quite divergent. Ascension
Technology's partners from the PV industry were Siemens Solar Industries, which provided PV modules, and
Omnion Power Engineering Corporation, which provided the inverters.
EPA awarded the contract for this project to Ascension Technology in the third quarter of 1992. The final
system design effort began shortly thereafter, and the first system was installed ana operating in April 1993.
Ten of the systems were operating by the end of August 1993, and the last was completed by mid-January
1994. Monitoring of each system began concurrently with initial system operation, although the "official data
start date" was delayed at sites where there were initial technical problems with either instrumentation or PV
system hardware. At each site, 15-minute average values of solar irradiance, ambient temperature, PV system
power output, and building load were recorded and stored for subsequent retrieval by modem. Monitoring of
each site (for the purposes of this study) continued through September of 1994.
Emission rate and load data provided by each of the participating utilities were used in conjunction with the
data collected from each system to conduct analyses of (1) the emission offsets resulting from operation of the
PV systems; (2) the ability of each PV system to reduce the peak power demand of the building on which it
was installed; and (3) the chronological correlation of each PV system's power output to the respective utility's
peak loads. In addition, a model was developed to simulate the operation of each system in conjunction with
dispatchable battery storage. This simulation shifted each system's daily generation to the utility's daily peak
load hour(s), thus increasing the peak load correlation and reducing emission offsets (due to battery charging
and discharging losses).
Chapter 2 of this report describes the design, installation, and cost of each system. Chapter 3 describes the data
acquisition system and presents data collection and review procedures. System performance history is
described generally in Chapter 4 (details are provided in appendix D). Chapter 5 discusses the model used to
simulate the behavior of the PV systems with dispatchable battery storage. Chapter 6 discusses each system's
'This estimate is derived from estimates of floorspace in residential, commercial, and industrial buildings (U.S. Department of Commerce, 1992),
assumptions about the ratio of roof space to floor space in each type of building, and the assumption that 25 percent 35 percent, and 45 percent
respectively of the available roof area of residential, commercial, and industrial buildings would be usable for PV installations.
1-2
-------
impact on the load of the building it is installed on, and utility-level load matching results are presented in
Chapter 7. The marginal emission rate models developed for each of the participating utilities are described
in Chapter 8, as are the site-by-site emission offset estimates. Conclusions from this project are summarized
in Chapter 9.
References
1. Farhar, Barbara. What Customers Say and What They Do: The Importance of Quality Market
Research, DOE/EPRI Green Pricing Workshop, Golden, CO, April 1996.
2. Oppenheim, Jerrold. A Program to Demonstrate that Consumers Place Value on Environmentally
Benign Electricity: Residential Rooftop PV, 13th European Photovoltaic Solar Energy Conference,
Nice, France, October 1995
3. U.S. Department of Commerce. Bureau of the Census. Statistical Abstract of the United States: 1992
(112th edition) Washington, D.C., 1992. Table 354, p.213.
1-3
-------
Chapter 2
PV Systems
2.1 System Design
Designs for 4-kW "building block" PV systems were developed for the EPA Project building upon
project staffs experience with roof-mounted PV arrays in prior projects. The majority of the project's
sites use either one system (4kW) or a group of three systems (12 kW total). Both pitched- and flat-roof
installations utilize Ascension Technology RoofJack PV array supports, which have been used to install
more than 1 megawatt of PV systems, PV arrays are held in place by ballast on flat-roofs: this approach
requires no roof penetrations for hold-dow n of the PV arrays. System design details were developed in
close cooperation with Siemens Solar Industries of Camarillo, CA, the PV module supplier. Siemens
fully supported the project and provided custom mechanical assemblies of PV modules in shipping crates
designed to hold the 12 PV panel assemblies for one 4-kW PV system,
2.1.1 Module Selection
Siemens Solar Industries was selected as the supplier of PV modules through a competitive bidding
process during development of the Ascension Technology proposal. Siemens was low bidder and highly
supportive of Ascension Technology's panel assembly design and the project's requirements to ship
modules to many sites. The PV modules in these systems are Siemens model M55j. Characteristics of
the module and panel assembly are shown in Table 2-1.
2.1.2 PV Array Configuration
PV arrays were configured using 12 PV panel assemblies. Each PV panel assembly contains seven
modules, electrically wired in series, with characteristics shown in the Table 2-1. A PV source circuit is
formed with four PV panel assemblies, wired in series. A 4-kW PV array comprises three PV source
circuits. Characteristics of the PV arrays are shown in Table 2-2 and illustrated in Figure 2-1.
2.1.3 PV Array Wiring
To expedite field wiring, the PV panel assemblies were prepared with single-pole quick-connectors.
These Pulse-Lok™ connectors are manufactured by Alden Products (Brockton, MA). Connector cables
were fabricated by Alden in specified lengths, using 10 AWG type USE-2 cable. The connector cable
assemblies were shipped to Siemens for factory rewiring of PV panel assemblies. When shipped, each
PV panel assembly had two connector cables — a plug (overall positive) and a receptacle (overall
negative).
In the field, connecting PV panel assemblies in series simply required mating the connectors from
adjacent panels. To protect the connectors from mechanical damage and to keep them out of the
elements, they were tucked inside a galvanized steel pipe nipple, which bridges between the RoofJack
upright legs.
2-1
-------
ASCEX3I0X TECHNOLOGY, INC.
FO 3OX 5M. LINCOLN CEM'KK. MA 0:V73 |
rHA-.»:\G 3* _ani 95 Bf ?.CR
NOiiS' ,.j—:-vS L'-'^ .'S_ S NCli. -•¦.-•A5E
3vST£VS, *-S S-C-.V\ -ERE
Figure 2-1. Schematic of a 4-kW PV System.
Source circuits are "bi-polar", with balanced voltages above and below a neutral or center-tap conductor.
The Omnion power conditioner requires this three-wire configuration at approximately ± 250 Volts DC.
The wiring in each bi-polar PV array source circuit terminates at a Source Circuit Protector. A 4kW
array contains three source circuit protectors row of four PV panel assemblies. The source circuit
protector performs important electrical safety functions and makes array wiring convenient. The source
circuit protector contains two connector cables for interconnection to the array groups that form the bi-
polar source circuit, blocking diodes to prevent reverse current flow and a surge suppressor to shunt
lightning-induced surges to ground. The Source Circuit Protector was designed and developed
specifically for this project and proved to be a convenient wiring interface.
Field wiring within the array terminates at the Source Circuit Protectors. To finish the array wiring, each
host utility contracted an electrician to provide and install conduit and run conductors between the
Source Circuit Protectors, This wiring connects the three source circuits in parallel to form the PV array
output. PV array output wiring runs in conduit to the "power panel" inside the host building, where
disconnect switches, the Omnion inverter and metering equipment are located. PV panel assembly
frames. RoofJacks and ballast trays are bonded together by the equipment ground wiring, which also
terminates at the PV Source Circuit Protector.
-------
2.1.4 Mounting Hardware
During the design stage of the project, one objective was to simplify the mechanical installation of the
PV arrays. Building upon earlier roof-mounting experiences, the RoofJack mounting system was
refined. Ascension Technology's principals developed RoofJacks in the early 1980s, for use on pitched
residential roofs. This mounting system effectively reduces materials and field labor. RoofJacks for a
pitched roof are aluminum "L" and "U" shaped brackets, which support PV panel assemblies at four
points, keeping the PV panels parallel-planar to the existing roof. An air gap of several inches, between
the panels and roof, helps promote array cooling (to the benefit of conversion efficiency).
Table 2-1. PV MODULE AND PANEL ASSEMBLY CHARACTERISTICS
FV MODULE MECHANICAL
Manufacturer, Model Number
Siemens Solar Industries, M55j
Length x Width x Frame Depth
(inches)
50.9 x 13 x 1.4
Weight
(lbs)
12.6
PV MODULE ELECTRICAL
Irradiance: Standard Test Conditions
(W/mz)
1,000
Cell Temperature: Standard Test Conditions
25
Maximum Rated Power @ STC
(Watts DC)
53
Max-Power Voltage @ STC
(Volts DC)
17.4
Max-Power Current @ STC
(Amps DC)
3.05
Open-Circuit Voltage @ STC
(Volts DC)
21.7
Short-Circuit Current @ STC
(Amps DC)
3.4
PV PANEL ASSEMBLY
Number of PV Modules, Configuration
7 Modules in series
Length x Width x Frame Depth
(inches)*
91 x 50.9 x 3.4
Weight
(lbs)b
~110
Maximum Rated Power @ STC
(Watts DC)
371
Max-Power Voltage @ STC
(Volts DC)
122
Max-Power Current (2> STC
(Amps DC)
3.05
Open-Circuit Voltage @ STC
(Volts DC)
152
Short-Circuit Current @ STC
(Amps DC)
3.4
al inch = 2.54 centimeters
bl pound = 0.454 kilogram
2-3
-------
Table 2-2. 4kW PV ARRAY CHARACTERISTICS
4 kW PV ARRAY
Number of PV Panel Assemblies
12 (84 PV modules total)
PV Array Tilt Angle
fixed, 15 degrees from horizontal
Overall Electrical Configuration
28 modules series x 3 parallel
Maximum Rated Power @ STC
(Watts, DC)
4,452
Max-Power Voltage @ STC
(Volts, DC)
= 244
Max-Power Current @ STC
(Amps, DC)
9.15
Open-Circuit Voltage @ STC
(Volts, DC)
±304
Short-Circuit Current @ STC
(Amps, DC)
10.2
RoofTacks were supplied to each project, complete with a butyl rubber sealant on the bottom surface.
The RooOack is placed directly on the roof shingles and anchored in place using self-drilling deck
fasteners. A domed sealing washer under the head of the deck fastener provides additional weather-
sealing assurance. These features provide a very effective weather seal for pitched-roof array
installations.
The key feature of the RoofJacks is an "L"-shaped slot, which mates with a pin assembly attached to the
PV panels. The pin assemblies comprise a bolt, a sleeve and two large diameter washers. Four pins are
installed on each PV panel assembly, two at each end. With the RoofJacks in place, a panel assembly is
brought into position and slid down and into the slots where it is passively held. No tools are required to
install the PV panels on the RoofJacks, reducing the risk of module damage. Unfortunately, the pin
assemblies were attached in the field, not in the factory, where this work could have been completed
more easily.
Flat-roof installations use an adaptation of the RoofJacks, designed to work with ballast trays. The flat-
roof RoofJacks provide a tilt angle of 15 degrees. Galvanized steel trays, 94'/i" by 46" in size, were
designed to be placed end-to-end in rows. Prior to placement, a RoofJack was bolted to the tray. When
laid out, the spacing of the RoofJacks was precisely as required to match the pin spacing of the PV
panels. The overall design allows reasonable tolerances to accommodate variations in all dimensions.
Crews shoveled gravel or placed other ballast in the trays to prevent them from moving. With the
ballast in place, the PV panel assemblies were installed. Preparations for flat roofs varied and depended
upon a roofs specific composition and type. In all cases roofers were consulted regarding flat-roof array
installation procedures.
2.1.5 Power Conditioning/Utility Interconnection
Omnion Power Engineering was selected as the supplier of power conditioners. The PV systems were
designed to accommodate the specifications of the 4kW-rated Omnion Series 2200 unit. Specifications
of this inverter are shown in Table 2-3.
2-4
-------
Electricians mounted the inverter(s) on a %-inch plywood attached to a convenient wall space near the
point of interconnection to the building electrical service. Output from the inverter passes through a
kilowatt-hour meter, equipped with a pulse-initiator tied into the datalogger for the site, The AC output
of the PV system ties into the building electrical service through a 40 amp 120 V circuit breaker, in a
convenient distribution panel.
At sites with three systems, a three-phase system is formed. The separate 120 Vac outputs from the three
FV systems feed into a single 120/208 three-phase circuit breaker, providing a balanced interconnection.
2.1.6 Monitoring Instrumentation
Standardized instrumentation was included with each of the 17 PV system sites, to measure
meteorological and PV system performance variables. A Campbell Scientific CR10 datalogger lies at
the heart of the instrumentation systems. The baseline instrumentation is described in Table 2-4.
Ascension Technology's Rotating Shadowband Pyranometer (RSP) (see Appendix C) simplifies the
measurement of the components of sunlight: direct beam irradiance (coming from the solar disk),
diffuse horizontal (coming from the rest of the sky, excluding the solar disk) and global horizontal (all
irradiance falling on a horizontal surface). With the components of sunlight, it is possible to estimate the
irradiance on a surface of any orientation, fixed or tracking.
Wisconsin Public Service Company and Southern California Edison Company purchased additional
instrumentation, to monitor their PV arrays. At these sites, transducers were added to measure PV array
DC current and DC voltage. This extra instrumentation allows calculation of PV array and inverter
efficiencies. More importantly, it has been useful for developing PV system models and performing PV
system troubleshooting.
Table 2-3. POWER CONDITIONER CHARACTERISTICS
POWER CONDITIONER
Manufacturer, Model Number
Omnion Power Engineering Corp.
Series 2200, 4kW
Rated AC Output
33.4 Amps @ 120 Volts AC
single phase output
DC Input Voltage Tracking Range
±200-250 Volts dc
Power Factor
>0.98 from 10% to full power
Efficiency
approximately 94% at full power
Harmonic Current Distortion
<5% RMS from 30% to full power
Frequency
60 Hz, 11 Hz
Weight, Mounting
55 lbs, wall mount4
Dimensions
17" high x 15" wide x 10" deepb
*1 pound = 0.454 kilogram
hl inch = 2.54 centimeters
2-5
-------
Table 2-4. INSTRUMENTATION OF PV SYSTEMS
Variable
Definition
Sensor
F
pv
AC kilowatt-hours produced by the PV system
pulse-initiating kWh meter
(supplied by utility)
E,
AC kilowatt-hours used in the host building
pulse-initiating kWh meter
(supplied by utility)
I
1poa
Plane-of-array irradiance
LiCor 200 SZ pyranometer
mounted in plane of array
Idn
Direct normal irradiance
Ascension Technology Rotating
Shadowband Pyranometer
Igh
Global horizontal irradiance
Ascension Technology Rotating
Shadowband Pyranometer
Ihd
Horizontal diffuse irradiance
Ascension Technology Rotating
Shadowband Pyranometer
T
1 a
Ambient air temperature
Thermistor housed in a radiation shield
2.2 System Installation
2.2.1 Hardware Costs
The installations in this project are based on a nominal 4-kW (3895 WDC at PVLSA Test Conditions
(PTC) of 1,000 W/m2 irradiance and 20° C ambient temperature) system consisting of twelve PV panels,
one inverter and balance-of-system components including RoofJaeks, ballast trays (flat roofs only),
panel-to-panel wiring, row junction boxes, and DC disconnect. The 12-kW systems are made using three
identical 4-kW rated systems. The major part of the total cost consists of the hardware costs presented in
Table 2-5, which are constant for all installations.
Table 2-5. HARDWARE COMPONENT COSTS
4-kW flat roof
12-kW flat roof
4-kW pitched roof
PV Panels
S22.698
$68,094
$22,698
Power Conditioner
$3,456
$10,368
$3,456
BOS (includes trays/roof
jacks)
$2,200
S6.600
$820
TOTAL HARDWARE
COST
$28,354
$85,062*
S26.974
Substitution of roof clips arid braces for ballast trays at Southern California Edison's 12-kW site resulted in a BOS cost
of $3,189 and total hardware cost of 581,651.
2-6
-------
In addition each system has installation costs which are specific to the individual site. These costs
include permits, transport of the materials to the roof, preparation of the roof to receive the trays, buffer
materials used between the trays and the roof, installation of the trays, panels, electrical connections,
wiring and meters, miscellaneous electrical materials and shipping of the PV panels, the power
conditioners, the ballast trays, RoofJacks, and PV Source Circuit Protectors.
To determine total cost, the hardware costs described above are combined with the site specific costs.
The total costs for each system are reported in the following sections and summarized in Table 2-6. The
cost of the data acquisition equipment, site specific engineering and testing fees and any untypical
construction work associated with the installation (for example, removal of a skylight, addition of a roof
hatch) are not included in the total system cost. Salaries for utility personnel time attributed to project
administration are not included. However, when utility electricians and other personnel directly
participated in the installation the cost of their services is included in the total cost.
2.2.2 New York State Electric and Gas: 12-kWSystem
New York State Electric and Gas Corporation (NYSEG) is an investor-owned utility serving 784,000
electric customers and 224,000 natural gas customers in suburban and rural upstate New York. In 1992
NYSEG generated almost 18 billion kWh of electricity at seven coal-fired generating stations, several
small hydroelectric generating stations and one nuclear generating station. 1992 peak demand was 2,597
MW during the summer months and 2,259 MW in the winter. The generation mix for this service area is
predominately coal with coal accounting for 79% of the 1992 generation mix, the remainder coming
from nuclear (5%), hydro (2%) and purchases (14%). Forecasts for the year 2000 show a generation mix
of 64% coal, 29% purchases (22% natural gas, 7% other sources), 5% nuclear and 2% hydro. The Clean
Air Act Amendments of 1990 will result in significant future expenditures for emission reduction at
several of NYSEG's coal-fired generating stations.
NYSEG chose to install a 12-kW system on their Customer Service Center in Pittsburgh, NY.
Constructed in 1980, this one story building contains approximately 45,000 square feet with a roof area
large enough to accept an additional 20 to 24 kW of PV. The flat roof is a protected system with 2-3
inches of gravel ballast on top of the rigid insulation. Since the warantee on the roof had already
expired, no measures were required to preserve it. Annual electrical use for this facility during 1992
totaled 1,484,100 kWh; this ranged from a high of 224,100 kWh in January to a low of 59,400 kWh in
August, This high winter load is caused by the electric heating system which includes a water tank and
heat pump in the basement to utilize off-peak power. 1992 peak demand for this building was 702 kW.
Installation began on March 31, 1993 when the roof trays, data acquisition system and one PV panel
assembly were installed. Since the roof already had 2-3 inches of gravel, no additional ballast was
required. Ballast was temporarily removed from the area designated to receive the trays, and sheets of
60 mil EPDM were placed on the roof under each array to extend six inches beyond the edges of each
tray. A roof hatch and a ladder were also installed to provide easier access to the system. Work
continued on April 1 until midday when snow, ultimately ten inches, stopped the work. Independent
contractors completed the work as weather permitted and project staff returned to put the system on-line
on April 16. Since this is a utility-owned structure, preliminary negotiations were simplified as no
utility-owner contract was required. The installation did have a permit and was inspected by the local
electrical inspector as well as the Fire Bureau of Underwriters, The installation was documented both in
2-7
-------
still photos and in video, and the resultant material shared with other utilities as they prepared for their
own installations.
Installation costs included $5,849 for the crane and general contractor and miscellaneous materials
including the EPDM sheets and $9,502 for the electrical contractor and miscellaneous electric supplies.
Combined with hardware costs of $85,062 and shipping costs of approximately $1,500, this results in a
total cost of $101,913 or $9.69 per Watt (ac PTC). These costs do not include the costs associated with
the roof hatch and ladder.
2.2.3 Northeast Utilities Service Company: 4-kWSystem
Northeast Utilities is the parent company of the NU System (collectively referred to as NU) which serves
S.66 million customers in Connecticut, New Hampshire and western Massachusetts. With 1992 sales
totaling 29,300,000 MWh, NU is among the 20 largest electric utilities in the country and the largest in
New England, System peak demand in 1992 was 4,999.8 MW; this represents a 15.6% decrease from the
previous year. The 1992 generation mix for this utility was 48% nuclear, 25% oil, 11% coal, 8%
cogeneration. 5% hydro and 3% gas. Nuclear energy accounted for 4% more of the mix in 1992 than in
1991 while oil decreased by 3%; other sources changed by 1% or less. NU estimates that by the year
2005 energy conservation and load management will save the company the capacity equivalent of a
generating unit larger than the 1.149-MW Millstone 3 nuclear plant.
NU chose to install a 4-kW system on their Berlin East Headquarters in Berlin, CT. This building,
constructed during 1990-1991, contains a total of 250,000 square feet in three levels with a building roof
area of approximately 80,000 square feet. The usable roof area could accomodate approximately 300
kWp of additional PV array. The roof is a Carlisle EPDM membrane with 14 psf of ballast; it is designed
for 25 psf of snow with a 10 psf margin. The building consists of three cores, each serving approximately
one third of the building; each core is metered separately. The north core was chosen to receive the array
and have its electrical demand monitored and partially met by the PV power produced. This building is
being studied by.NU to document demand-side management features such as efficient lighting (60% of
typical energy use), daylight sensing perimeter fixtures, occupancy sensors in offices and conference
rooms, high energy efficiency rating, staged air conditioner compressors, variable speed drive HVAC
fans, heat pump water heating using waste in electrical closets and low emissivity window glazing. This
building provides an excellent opportunity to test PV generation alongside other DSM features. Electrical
metering for the facility began July 1, 1992, Electrical use for the north core from July 1992 through
April of 1993 totaled 958,463 kWh; this ranged from a high of 116,519 kWh in March to a low of 80,413
kWh in September. Peak demand for the north core during this period was 450 kW which occurred both
in December 1992 and February 1993. Peak demand for the entire building was 1,370 kW in February
1993.
The majority of the installation was done on April 5, 1993 when the roof trays, PV panels and data
acquisition equipment were installed. Since the roof already had 14 psf of ballast, no additional ballast
was required. The local roofer chose not to accept Carlisle's recommendation that a 60 mil EPDM sheet
be placed under the trays and instead elected to use Carlisle Walkway, an EPDM material 120 mil thick.
Electrical wiring was completed by NU electricians in the following few days and project staff returned
to review the completed installation on April 13 before putting it on-line.
2-8
-------
Site specific costs for this installation included $1,015 for the Carlisle Walkway, $491 for electrical
supplies, $881 for vehicles including the crane, $95 for miscellaneous expenses and $4,631 for labor for
a total of $7,113. Adding the site specific costs to hardware costs of $28,354 and shipping costs of
approximately $500, gives a total cost of $35,967 or $10.26 per Watt (ac PTC).
2.2.4 Atlantic City Electric Company: 4-kWand 12-kWSystems
Atlantic City Electric (ACE) serves more than 455,000 customers in 2,700 square miles of southern New
Jersey. 1992 sales totaled 7.7 billion kilowatt-hours, a decline from previous years. ACE expects sales
to increase slightly in the future as improved airport facilities and a new convention center in Atlantic
City are expected to bring growth to the area. From 1988 through 1992 coal and nuclear fuel sources
provided on average more than 70% of ACE's total energy requirements. Coal usage has declined from
54% in 1988 to approximately 37% in 1990-1992. As coal usage has declined, purchased power has
increased.
ACE chose the Marine Mammals Stranding Center in Brigantine, NJ as the site for a 4-kW system. With
over 50,000 visitors a year, this location gives this project good public visibility. Constructed in 1972,
the single story building contains approximately 1,200 square feet. Its asphalt shingle roof has a pitch
of approximately 4:12 with an orientation which gives the panels southern exposure. Annual electrical
use for this facility during 1992 totaled 49,335 kWh, ranging from a high of 6,217 kWh in October to a
low of 2.830 kWh in November. Because there is no demand meter for this installation, peak load is
unavailable.
Prior to installation of the PV system, an existing skylight was removed to allow placement of the 4-kW
system which requires most of the south facing roof area, and an outdoor enclosure was built to
accommodate the power conditioner since there was no existing interior space available for this purpose.
On June 3, 1993 project staff along with a local electrical contractor installed and connected the panels.
Wiring of PV panels on a pitched roof is difficult because of the restricted space beneath the panels. To
facilitate panel-to-panel wiring, the panels were propped up about 10 inches, then lowered to their
normal position after wiring was completed. The system wiring was completed and the system went on-
line this same day. This installation went through the permitting process required by the Town of
Brigantine and was inspected by both the electrical and building inspectors.
Installation costs for this system totaled S3,804. Combined with hardware costs of $26,974 and shipping
costs of approximately $320, the total cost was $31,098 or $8.87 per Watt (ac PTC). These costs do not
include the costs associated with removing the skylight or constructing the enclosure for the power
conditioner. In comparing cost per Watt, it should be noted that cost per Watt for a pitched roof, with its
minimal hardware and preparation costs, is significantly below the cost per Watt for a similarly sired
system installed with ballast trays on a flat roof.
ACE chose to install a 12-kW system on the roof of its Headquarters in Pleasantville, NJ. This single
story building, constructed in 1970, contains a total of 50,000 square feet, enough roof area to install an
additional 250 kW of PV. The roof is a modified bitumen system (Garland Stress-Ply) installed in 1991
over existing roofing. Despite this additional roof w eight, structural calculations showed the roof
capable of supporting the additional weight of the array and ballast. Since the ballast tray system was
originally designed for EPDM roofs, ACE explored the possibility of installing the trays on a modified
bitumen system. The ACE Supervisor of Facilities Operations reviewed the tray installation with a
2-9
-------
Garland Company technical representative. At Garland's recommendation, ACE opted to use an APP
(Attatic Polypropylene) membrane instead of an EPDM membrane between the tray and the roof. The
APP membrane was selected because the modifier is plastic which is the same modifier found in the
existing roof; it is harder than rubber and provides the necessary protection for the roofing system of this
building. Garland also provided detailed specifications for the preparation of the roof to receive the trays
including instructions to spud (the process of removing the roofing aggregate and most of the bituminous
top coating by scraping and chipping) all existing gravel from the roof mat wherever a tray is to be
installed. It was noted that this spudding is best done in cooler weather when the asphalt will be less
sticky.
Project staff visited the Atlantic Electric Headquarters on April 6, 1993. Although the utility was not
ready to begin actual installation, project staff demonstrated the installation procedure for the electrical
contractor by setting up a row of trays and panels on the ground. Later in the month, the electrical
contractor installed and connected the arrays. On April 27, project staff returned to Pleasantville to
install the data acquisition system, review the installation and check out the system. Array wiring was
routed and secured and the system was turned on. One power conditioner failed to start. The failed
power conditioner was replaced with the one slated for the 4-kW Brigantine site resulting in a successful
system start. Subsequently, Ornnion repaired the failed power conditioner and returned it to the
Brigantine site. An Egg Harbor Township construction permit was issued for this work, and the
installation was subsequently inspected by both the electrical and building inspectors.
The cost for electrical work and miscellaneous electrical materials totaled $8,855; cost to prepare the
roof to insure maintenance of the warranty totaled $9,900. Combined with hardware costs of $85,062 and
shipping costs of approximately $1,180, this results in a total cost of $104,977 or $9,985 per Watt (ac
PTC).
2.2.5 New York Power Authority: 4-kW System
A nonprofit, public-benefit energy corporation, The New York Power Authority's (NYPA) mission is to
furnish the people of New York State with low cost electricity. NYPA finances generation and
transmission projects through bond sales to private investors and repays the bondholders with proceeds
from operations. It sells energy to private utilities for resale without a profit to their customers, to
authorized public agencies and publicly-owned utilities and, as an incentive for more firms to locate or
expand in the state, to about 150 companies. The Power Authority sold 34.2 billion kWh of electricity in
1992 which supplied 22% of New York State's needs. Of this total, 5,775,547 kWh were sold to the
White Plains School District. This power came primarily from the Indian Point 3 Nuclear Power Plant
and the Poletti Power Project which burns natural gas or oil depending on the season. 28.6 billion kWh
of NYPA's total 1992 sales were generated by Power Authority facilities; the remainder was purchased
from other sources. The 1992 generation mix for this utility was 62.6% hydro, 14% nuclear, 7% natural
gas and oil and 16.4% purchases.
The 4-kW system was placed on a building in the White Plains (New York) High School campus. This
building is part of a seven building complex which includes a pool, field house, and auditorium-music
unit. Constructed in 1960, the buildings collectively contain 315,000 square feet. The flat roofing system,
which was installed in 1992, is Firestone protected membrane; the insulation is on top of the membrane
and is covered by 8 psf of gravel ballast. Although it would appear that the 24,000 square foot roof could
accommodate a much larger array, structural limitations dictated the location of the 4-kW array and gave
2-10
-------
little option for an additional array. While the majority of the building has long structural spans which
could not accept the additional weight of the array, a structural review determined that a corridor around
a central courty ard with smaller spans could accept the additional weight. Given orientation
requirements and shading from mechanical penthouses, only one of the four structurally suitable
locations was in fact acceptable from a solar standpoint. Annual electrical use for this total complex
during 1992 totaled 1,929,600 kWh. This ranged from a high of 214,800 kWh in December to a low of
100,800 kWh in August. Peak demand in 1992 was 648 kW.
This system was installed on May 3, 1993; with the exception of some electrical connections and wiring
all work was completed on this day. Roof preparation and tray installation was done by an independent
contractor, the roofing contractor for the original roof. The tray installation, done according to AT details
using a 60 mil EPDM membrane, was approved by Firestone keeping the existing roof warranty in
effect. Electrical work was done by an independent electrical contractor. Project staff assisted on May 3
and then returned on May 7 after the electrical work was completed to put the system on-line. A permit
for this work was issued by the White Plains City School District.
The installation costs included $3,300 for the roofing contractor who transported all materials to the roof
(using internal stairs rather than a crane) and installed the trays, and S2.980 for the electrical contractor.
Combined with hardware costs of $28,354 and shipping costs of $560, this results in a total cost of
$35,194 or $10.04 per Watt (ac PTC).
2.2.6 Arizona Public Service: Two 4-kWand One 8-kWSystem
Arizona Public Service (APS) is an investor-owned utility serv ing more than 600,000 customers spread
over approximately 48,000 square miles in Arizona. The company's 1992 electric sales totaled
20,562,903 kWh with a summer peak load of approximately 3,700 MW. APS's 1992 generation mix
was 62.2% coal, 28.7% nuclear, 3,0% gas and oil, 0,2% hydroelectric and 5.9% purchases. APS has
many years of experience in both solar energy and demand side management and has made a
commitment to be one of the top five utilities in the country in terms of environmental performance by
1995,
The private residence chosen to receive a 4-kW system is located in Peoria, Arizona, approximately 20
miles northwest of downtown Phoenix. Like many homes in this area, this residence is a single story
ranch with a dual-pitched asphalt shingle roof. Also, like most homes in this area air-conditioning
equipment is mounted on the roof. Annual electrical use for this residence during 1992 totaled 14,528
kWh; this ranged from a high of 2,152 kWh in August to a low of 611 kWh in April. Peak demand for
this period was 7.8 kW in June 1992, Installation of the PV array took place on May 5. 1993. Layout of
the Roo/Jacks proceeded quickly and without difficulty using a special template provided by AT. The
PV array was placed on the lower-half of the roof, clear of the air-conditioning equipment mounted near
the top of the roof. Two rows of panels were mounted on the low pitch section of roof, while the top row
was mounted on the steeper pitched section. Wiring the PV array was more difficult on the pitched roof
than on the flat roof installations because of the restricted access beneath the PV panel assemblies. This
installation, which had a permit from the City of Peoria, was subsequently inspected and approved by a
City of Peoria inspector as well as the APS System Protection Department. The system went on-line in
June 1993.
2-11
-------
Installation costs included $909 for miscellaneous electrical materials and $3160 for labor. Combined
with hardware costs of $26,974 and shipping costs of $301, this results in a total cost of S31,344 or $8.94
per Watt (ac PTC).
APS conducted an extensive search to locate a building which would accept a 12-kW array. Roofs in the
APS service area are typically not ballasted, and it was difficult to find a flat roof capable of accepting
the weight of the array and ballast, an additional 15 psf. Scottsdale Memorial Hospital Outpatient
Recovery Center was finally identified as a building which could accept this added weight. Since only 8-
kW of PV would fit on this roof, it was agreed that an 8-kW system would be installed at this site and
APS would locate another site to accept the remaining 4-kW array. Part of a large hospital complex, The
Outpatient Recovery Center was chosen to receive the array and have its electrical demand monitored
separately and partially met by the PV power produced. The annual electrical use for this building
during 1992 totaled 677,500 kWh; this ranged from a high of 70,080 kWh in May to a low of 37,500
kWh in November. Peak demand for this period was 216 kW in February 1992.
This installation began on May 6, 1993 using a local contractor with assistance from project staff. No
buffering material was used between the tray and the lightweight concrete roof surface. Concrete blocks
were set in the trays in lieu of gravel ballast. The installation was inspected and approved by the APS
System Protection Department and went on line in June 1993.
Installation costs included $645 for the crane, $785 for miscellaneous electrical materials and $4,747 for
labor. Combined with hardware costs of $56,708 and shipping costs of $1,302, this results in a total cost
of S64,187 or 59.16 per Watt (ac PTC).
The remaining 4-kW array was placed on the roof of the one-story Conconino High School in Flagstaff.
Arizona. The roof system is a ballasted Carlisle EPDM rubber membrane. Although the roof is
approximately 283,200 square feet, no areas other than where the 4-kW array was installed had the
loading capacity to accommodate an additional PV array. Annual electrical use for this facility during
1992 totaled 2,332,300 kWh; this ranged from a high of 272,600 kWh in April to a low of 118,600 kWh
in August. Peak demand for this period was 596 kW in December 1992.
This system was installed the week of October 4 by an independent roofing contractor and an
independent electrical contractor under the supervision of Arizona Public Service staff. This system was
permitted and inspected by a municipal inspector. Difficulty in scheduling a required inspection by the
APS System Protection Department delayed the system start-up to January 1994. Due to a problem with
the phone line, performance data was not successfully transmitted until February 1994.
Installation costs were S639 for roof preparation, $129 for miscellaneous electrical supplies, and $2,758
for electrical labor for a total of $3,526. Combined with hardware costs of $28,354 and shipping costs of
$318. this results in a total cost of $ 32,198 or $9.19 per Watt (ac PTC).
2.2.7 Wisconsin Public Service; 4-kW and 12-kW Systems
Wisconsin Public Service Corporation (WPS) is an investor-owned electric and gas utility with a 10,000
square mile service area in northeastern Wisconsin and an adjacent part of Upper Michigan. In 1992
340,141 customers purchased almost 9.75 billion kWh of electricity representing about a 2% increase in
both sales and customers. Overall, however, through conservation and energy management programs, the
2-12
-------
company plans to reduce its peak demand while still anticipating a 1,5% per year increase in electrical
energy sales. The generation mix of fuels used to operate WPS power plants in 1992 compared to 1991
was 65,3% coal, down from 65.9%; nuclear, 15.6% up from 14.9%; hydro, 3.1% down from 3.2%;
combined natural gas and fuel oil, 0.3% down from 0.4%; and purchased power, 15.7% up from 15.6%.
WPS chose to put a 4-kW system on a Denmark, WI residence owned by one of its employees. This
single story dwelling has an asphalt shingle roof and contains approximately 1,500 square feet. Since this
is a newly constructed building, there are no energy use records. A local contractor, with PV and
solar thermal experience, was hired to install the PV array. In mid-April 1993 project staff met with the
installer to review the hardware and pitched-roof installation procedures. The actual installation
occurred on June 29 while the building was still under construction and electric service had not been
completed. Electric service began on July 26 at which time the system went on-line. The installation
was videotaped by Wisconsin Public Television and a Green Bay television station. Unlike the majority
of residential PV installations whose south facing roofs are too small for or mostly filled by a 4-kW
array, this residence could have accommodated an additional 4-kW of PV using its garage roof.
Installation costs totaled $4,001. Combined with hardware costs of $26,974 and shipping costs of $257,
this results in a total cost of S31,232 or $8.91 per Watt (ac PTC).
WPS chose to install a 12-kW system on the roof of a WPS service center repair garage in
Ashwaubenon, WI. This structure has a rubber membrane roof with ballast on top of the membrane.
Although the panels are located on the repair garage, the building whose electrical demand is monitored
and partially met by the PV power produced is an adjacent WPS office building. Fn 1992 annual
electrical use for this single floor building of approximately 10,000 square feet was 1,022,720 kWh; this
ranged from a high of 104,560 kWh in June to a low of 63,040 kWh in December. Peak load for this
period was 263 kW.
In mid-April 1993, project staff visited the site. Although WPS was unable to begin actual installation at
that time, AT presented slides and a video of previous installations to help the contractor understand the
installation process for a flat roof system. Subsequently WPS using an independent roofer and their own
electrician installed the system. An EPDM rubber membrane sheet was used between the tray and the
roof. This system went on line on May 26. This roof potentially (provided vents and other obstructions
could be accommodated) could receive an additional 4 to 6 kW of PV. On June 17, WPS held a press
briefing at this site. Unfortunately it was a rainy day and press attendance was limited to two television
stations, two radio stations and several newspapers. Ninety information packets and invitations had been
sent to the media in the area.
Installation costs included $2,980 for the crane, roofing contractor and EPDM rubber membrane and $
11,636 for the time spent by WPS electricians and miscellaneous electrical supplies. Combined with
hardware costs of $85,062 and shipping costs of $1,359, the total cost was $101,037 or $9.61 per Watt
(ac PTC).
2.2.8 City of Austin Electric Department; 12-kW System
The City of Austin (COA) operates an Electric Light and Power System which provides electricity for
298,000 customers in the City, adjoining areas of Travis County' and certain adjacent areas of
Williamson County. The City jointly participates with other utilities in the ownership of coal-fired
2-13
-------
electric generation facilities and a nuclear powered generation facility. Additionally, the City owns three
gas/oil-fired electric generation facilities, two of which are available to meet system demand. In 1992
CO A generated 6.6 billion kWh of electricity. Peak demand during this period was 1,498 MW during
the summer months and 1,098 MW during the winter. Tn 1992 the generation mix for this utility was
59.91% gas, 23.55% gas. 16.53% nuclear and .01% photovoltaic.
COA elected to place a 12-kW system on the roof of their Electric Utility Offices at Town Lake Center.
Electric use for this building during 1992 totaled 2,881,227 kWh; this ranged from a high of 306,305
kWh in July to a low of 193,784 kWh in February. Peak demand for this period was 871 kW in
September. Constructed in 1986, this building is five stories and contains 126,123 square feet. The
warrantee on the built-up roof has expired. Concern with trays interfering with drainage paths was
resolved by rotating the trays 90 degrees and redrilling the mounting holes. Evaluations of the strength
of the structure and its ability to receive an additional 15 psf, and of the roof insulation to receive this
additional weight without compression which would create ponding were conducted with satisfactory
conclusions. After considering various interface materials to use between the trays and the roof, COA
used a roof preparation specification similar to the one prepared for Atlantic City Electric's 12-kW
system.
Project staff visited Austin in late March 1993 to help plan for the installation. Resolving roof drainage
issues, and selecting and negotiating with a roofing contractor took longer than COA had anticipated.
Roof preparation finally began in mid-August when the loose gravel was swept away in the 42 places
where the ballast trays were to be placed. In these areas the imbedded gravel was scraped away and a
modified bitumen layer was then torched into place. The roof was inspected by a roofing expert for
compliance with specifications. During the next two weeks, Austin Electric personnel placed the trays.
Trays were taken by freight elevator to the top floor then carried up the final flight of stairs to the roof.
On August 28 a crane lifted the PV panels and the gravel to the roof. Austin Electric personnel spread
the gravel and installed and wired the PV panels. Project staff assisted in this installation. COA
electricians completed the wiring of the system and installation of the power conditioners, and the
system went on-line on September 22. However, because of COA's difficulty in obtaining and installing
a high resolution meter to measure small increments of PV generation for data collection purposes,
monitoring did not begin until March 14, 1994.
Installation costs included S 1,168 for the crane and for salaries of COA personnel who installed roof
trays, gravel, and put panels in place; $3,091 for electrical work; $4,750 for roof work and $1,881 for
miscellaneous materials for a total of $10,890. Combined with hardware costs of $85,062 and shipping
costs of $ 1,566, this results in a total cost of $97,518 or $9.27 per Watt (ac PTC).
2.2.9 Northern States Power Company 4-kWSystem
"Northern States Power (NSP) is an investor-owned utility' which distributes electricity- to about 1.3
million customers and natural gas to more than 380,000 customers in Minnesota, Wisconsin, North
Dakota, South Dakota and Michigan's Upper Peninsula. In 1992 the utility sold 37 billion kWh of
electricity. The peak load for this period was 6,128 MW. NSP customer demand for electricity is
projected to grow 2.2% per year between 1993 and 2010. The 1992 generation mix was 47% (low
sulphur) coal, 28% nuclear, 21% purchased power and interchange (including 11% hydro electricity from
Canada), 3% NSP hydro, and 1% from other sources, including refuse-derived fuel and wood. NSP is
2-14
-------
interested in investigating methods of applying PV to closely match their late summer afternoon peak
load (2:00 to 5:00 p.m. DST) in order to prove PV's value as a DSM tool.
NSP mounted their 4-kW system on the roof of Compar Corporation in Minnetonka, MN. This two-
story building, constructed in 1979-1980, has a footprint of 5,740 square feet and a total area of 11,480
square feet. The sloped site allows both levels grade access. The roof is modified bitumen with a gravel
surface. Recently purchased by Compar Corporation, the building has been totally refurbished. With
new systems and new occupants in the spring of 1993, historic electric use data for this building in not
relevant.
Installation of this system occurred on June 14, 1993. A permit for this work had previously been
obtained from the City of Minnetonka. An independent roofer covered sections of the roof on which
roof trays were to be placed with a 2" layer of rigid foam installation, followed by a layer of 60 mil
EPDM. The trays were then placed on this material and Filled with pea-sized roofing gravel which
spilled over the tray edges a few inches to cover the insulation and EPDM. The system went on line on
June 18, 1993.
Installation costs included $4,322 for the crane and general contractor and miscellaneous materials
including the insulation and EPDM and $4,685 for the electrical contractor and miscellaneous electric
supplies. Combined with hardware costs of $28,354 and shipping costs of $600, this results in a total
cost of $37,961 or $ 10.831 per Watt (ac PTC). This does not include the cost of the walkway ($2,161)
added to protect the roof from traffic to the array, the ship's ladder ($1,870), or the cost of adding a roof
hatch ($2,498).
2.2.10 Pacific Gas and Electric: 12-kW System
Pacific Gas and Electric (PG&E) is an investor-owned utility which supplies electricity, gas and water to
a population of 7,757,000 over an area of 94,000 square miles. In 1992 electricity sales totaled
75,434,046 MWh with a winter peak of 11,980 MW and a summer peak of 14,345 MW. During 1992
PG&E generated 57,467,998 MWh of electricity using a fuel mix that was 28.86% nuclear, 45.97%
steam, 12.43% hydro-conventional, .6% hydro-pumped storage and 12.14% from other power sources.
In addition PG&E purchased 27,264,421 MWh of electricity.
PG&E made an extensive effort to install a 12-kW system on one of their customer's buildings.
However, a number of problems arose relating to permitting, code compliance, roof area, solar access
and the customer's unfarriiliarity with PV. A shopping mall location proved not to have sufficient
unshaded, unobstructed roof area; a Kaiser Warehouse Distribution Center in Livermore, California
proved unfeasible in the project time-frame because of the local building inspector's requirement that the
power conditioner have an Underwriters Laboratories (UL) or Edison Testing Laboratories (ETL) listing.
Project staff contracted Inchcape Testing Serv ices, an ETL Laboratory in Oakdale, MN to field test the
Omnion power conditioner. As a result of this testing ETL recommended a few changes be made to the
printed circuit board in order to make the power conditioner comply with the applicable requirements of
the National Electric Code and UL 508C, While Omnion was prepared to modify their circuit board
design, they wanted to also include changes anticipated from a pending UL review. The process of
2-15
-------
addressing all recommendations simultaneously, making the required design revisions, and submitting
the power conditioner for retesting was not compatible with the project time frame.
Since a building owned by the utility is exempt from municipal electrical inspection, a system on one of
the utility's own buildings would not be required to meet UL listing requirements. A lesson learned
during this project is that having all of the equipment in the system (panels, power conditioners, PV
Source Circuit Protectors) listed by UL would simplify the permitting process and make it easier for
utilities to install systems on customers' buildings. .
Prior to this permitting problem, however, research had been done to determine if the roof, an Owens
Corning mineral cap system over lightweight concrete fill, would accept the PV array. Approval was
obtained from Owens Corning to use EPDM rubber membrane between the trays and the roof. This is
another lesson learned during this project; although the ballast tray system was designed for use on
EPDM ballasted roofs, roofing manufacturers of several other types of roofs have approved this system
for use on their roofs.
Finally, unable to obtain UL or ETL listed power conditioners, PG&E chose to install their system on the
roof of one of their own buildings, the Technical and Ecological Services Main Building in San Ramon,
California. Installation occurred in December 1993. On the first morning of installation, the roof layout
started at 7:00 a.m. and the crane was set up by 8:30 a.m.. With a six-man crew all of the equipment was
staged on the roof by noon. By 3:30 p.m. all of the trays and ballast were in place and two rows of
eleven panels each were installed and wired. The fourteen remaining panels were mounted and wired the
following morning by a two-man crew. An electrician was hired to finish the installation which went on-
line briefly on December 31. Tt was then shut down so that an AC disconnect switch, a PG&E
requirement, could be installed. The system was put back on-line on January 7, 1994. To comply with
Uniform Building Code seismic requirements for this region, this installation included a "tethering
system" designed to keep the array on the roof in an earthquake. This system consists of ten 4 inch by 4
inch square steel stanchions welded to the building frame at the roof edges at each end of the five rows of
panels. A 3/8 inch wire rope is attached to the stanchion at the beginning of each row, strung under the
panels and over the pans through all Roof Jacks in the row and attached to the stanchion at the opposite
end.
Installation costs included $850 for the EPDM sheets, $200 for permits. $11.000 for mechanical
installation and $5,750 for the electrical contractor and miscellaneous electric supplies. Combined with
hardware costs of $85,062 and shipping costs of $2,024, this results in a total cost of $104,886 or $9.98
per Watt (ac PTC). These costs reflect the materials and labor necessary- to install the tether system.
PG&E estimates that without the tethering system, the installation costs would have been reduced from
517,800 to between $10,800 and $11,800 with a resultant per Watt (ac PTC) cost between $9.31 and
$9.40. None of these costs reflect ETL testing costs ($3,000) for the Omnion power conditioner or the
site specific engineering fees ($3,000).
2.2.11 Southern California Edison: 4-kW(2 Residential) & 12-kWSystems
SCEcorp is the parent corporation of Southern California Edison Company and the Mission Companies.
Based on the number of customers, Southern California Edison (SEC) is the nation's second-largest
electric utility. The 106-year-old investor-owned utility serves 4.1 million customers in central and
southern California. Its 50,000-square-mile service territory has a population of nearly 11 million. 1992
2-16
-------
sales totaled 74,186 million kWh, a 4.3% increase from 1991, with apeak demand of 18,413 MW, a
10.2% increase from the previous year. The 1992 generation mix for this utility was 32% purchases from
non-utility producers, 24% gas, 22% nuclear, 14% coal, 5% purchases from other utilities and 3% hydro.
Southern California Edison's first 4-kW system was installed on the pitched roof of a residence at
Edwards Air Force Base. This system was installed on October 21 and 22, 1993. Project staff spent two
days on site installing the PV array and data acquisition system and instructing the electrician on the
wiring still to be completed. The electrician completed the system and instrumentation wiring the
following week. Project staff inspected the system and turned it on during a return visit on November 8,
1993.
Installation costs totaled $4,100. Combined with hardware costs of $26,974 and shipping costs of $92
this results in a total cost of $31,166 or $8,89 per Watt (ac PTC).
Southern California Edison's second 4-kW system was installed on the pitched roof of an officer's
residence on the Marine Corps Logistics Base in Barstow, California. Of note in this installation was the
difficulty in finding a roof on the base of sufficient size and proper orientation to receive this system.
Consequently, a roof modification was required to accommodate the top row of PV panels. Base welders
installed galvanized steel support brackets to extend the east-west dimensions of the pitched roof a few-
feet, so that the PV panels could be properly supported on the roof. It became apparent during this
project is that 4-kW PV arrays are often too large for residential roofs.
The system was installed on November 8 and 9, 1993. The power conditioner and instrumentation
enclosure were mounted outdoors on the north side of the house under a large overhang and awning
which was erected by the MCLB welders to protect the equipment from exposure to rainfall. A local
electrician completed the wiring and put the system on-line on November 17.
Installation costs totaled S3,788. Combined with hardware costs of $26,974, and shipping costs of $102,
this results in a total cost of S30.864 or $8.81 per Watt (ac PTC).
The building Southern California Edison chose for a 12-kW system is the Foundation for the Retarded of
the Desert located in Palm Desert, California, This single story, ten-year-old Butler-system building
houses both offices for the Foundation and light-manufacture/craft/assembly rooms for its clients. The
ribbed roof is a standardized metal design that is used as part of the Butler prefabricated system which is
installed worldwide. Since there are currently 800 million square feet of this type of roof (Butler
"Landmark Series" R-24) in place throughout the world, there is large potential for future PV
installations using a standard fastener to attach the panels on this type of roof. This system also provides
for mechanical attachment of the RoofJacks thus meeting the Uniform Building Code seismic
requirements for this area.
The contractor for this installation, Holmgren, Incorporated, reviewed the proposed installation with
Butler to insure that an attachment directly to the ribs of the metal roof, in lieu of a ballast tray mounting,
was feasible and verified that the roof could support the overall loading presented by the PV array.
Holmgren worked with project staff to develop a design which utilized Butler mounting clips, with
additional support angles, to attach the RoofJacks to the ribs of the metal roof. The contractor also
obtained approval for the installation from the Planning Department of the City of Palm Desert. This
2-17
-------
approval was given without any concerns expressed. Since the Board is typically concerned with any
projects which have negative visual impact, the panels' low profile and setback from the edge of the roof
probably helped this project obtain approval. With the Planning Board approval in hand, the Contractor
was granted a permit from the City of Palm Desert.
Actual installation occurred on November 11 and 12, 1993. Installation of the clamps, Roof Jacks and PV
panels proceeded quickly and was completed in less than one day by a two- person crew. A large part of
the electrical work was completed during these two days, but remaining work, including installation of
two of the three power conditioners, was completed later in the month. The power conditioners were
mounted on an outdoor wall. A screen was fabricated to protect the equipment from exposure to sunlight
and rain, while still providing ventilation for cooling. The screen also hides the equipment and blends
with the architecture of the building. The system, which went on-line on January 13, 1994 was inspected
by the City electrical inspector and by local power company representatives before the link with the grid
was allowed. With no set back from the roof edge, this building could potentially accommodate an
additional 36 kW of PV; if set backs, similar to the ones for the 12-kW system are maintained, an
additional 12 kW could be accommodated.
Installation costs totaled $14,825, hardware costs totaled $81,651 (this allows for a savings of $3,411
resulting from using Butler clips and angles in lieu of galvanized ballast trays) and shipping costs totaled
$220 (shipping costs were substantially reduced in this case given the site's proximity to the
manufacturing plant for the PV panels (Camarillo, CA) and the deletion of the ballast trays). This
resulted in a total cost of $96,696 or $9.20 a Watt (ac PTC). Engineering fees and the cost of the outdoor
screened enclosure are not included in the total cost.
2.3 Cost Summary
In comparing cost per Watt for the various installations, it is important to note that there will be
differences based on the individual utilities' methods of contracting the installation and cost accounting.
Variations in cost per Watt may be attributed in part to the size of the system, type of roof (type of
RoofJacks required, ballast required, preparation required), and geographic location which affects both
shipping and labor costs. It is interesting that these trends are clearly found in the summary in Table 2-6.
For example, the five pitched roof installations, ranging from $8.81 to $8.94 per Watt, have the lowest
cost per Watt, The highest cost per Watt, $10.40 to 10.83, with the exception of the APS 4-kW system,
was for 4-kW ballasted flat roof systems. The 12-kW ballasted flat roof systems ranged from $9.28 to
$9.98 per Watt illustrating the economy of scale of a larger system.
2-18
-------
Table 2-6. SYSTEM COST SUMMARY
Project
Roof Type
Hardware
Cost
Installation
Cost
Shipping
Cost
TOTAL
COST
COST/
WATT*
NYSEG
12-kW
ballasted flat,
protected
membrane
$85,062
$15,351
$1,500
$101,913
S9.69
NU
4-kW
ballasted flat,
EPDM
$28,354
$7,113
S500
S35.967
S10.26
ACE
Pleasantville
12-kW
flat, modified
bitumen
$85,063
$18,755
SI,180
SI 04,997
S9.98
ACE
Brigantine
4-kW
pitched asphalt
shingle
526,974
$3,804
$320
$31,098
$8.87
NYPA
4-kW
ballasted flat,
protected
membrane
$28,354
$6,280
$560
$35,194
$10,04
APS Seottsdale
8-kW
flat, lightweight
concrete
$56,708
$6,177
$1,302
$64,187
$9.16
APS Peoria
4-kW
pitched asphalt
shingle
$26,974
$4,069
$301
$31,344
$8.94
APS Flagstaff
4-kW
ballasted
flat, EPDM
$28,354
$3,526
$318
$32,198
$9.19
WPS
Ashwaubenon
12-kW
ballasted flat,
EPDM
$85,062
$14,616
$1,359
$101,037
$9.61
WPS Denmark
4-kW
pitched asphalt
shinale
$26,974
$4,001
S257
S31,232
S8.91
2-19
-------
Table 2-6. Continued
Project
Roof Type
Hardware
Cost
Installation
Cost
Shipping
Cost
TOTAL
COST
COST/
AC WATT*
NSP
4-kW
flat, modified
bitumen
$28,354
59,007
S600
537,961
S10.83
PG&E
12-kW
PV mech-
anically
attached
flat, lightweight
concrete with
mineral cap
585,062
517,800
($10,800-
S11,800
without
tether)
52.024
SI 04,886
59.98
(S9.31-S9 40
without
tether)
COA
12-kW
flat, modified
bitumen
$85,062
$10,890
SI.566
597,518
S9.27
SCE—Barstow
4-kW
pitched asphalt
shingle
$26,974
$3,788
SI 02
S30.864
S8.81
SC-Edwards
4-kW
pitched asphalt
shingle
$26,974
$4,100
S92
531,166
S8.89
SCE-Palm
Desert
12-kW
flat, Butler
system, no
ballast travs
$81,651
$14,825
S220
S96.696
S9.20
* COST/AC WATT is based on each system's AC
is calculated as follows:
plane of array (POA) irradiance:
ambient temperature (Ta):
cell temperature (Tc):
efficiency degradation coefficient:
inverter full load efficiency:
4 kW (nominal) DC array rating @ SOC:
4 kW (nominal) system AC rating @ SOC:
ing at standard operating conditions (SOC), which
1000 W/m2
20° C
50° C
0.4%/°C
90%
4,452 x (50 - 20) x 0.4 = 3,895 Wdc
3,895 x0.90 = 3,505 Wac
2-20
-------
Chapter 3
Data Collection, Storage, Retrieval, and Review
3.1 Data Collection
Data collection, storage retrieval and review were conducted in accordance with the plan set forth in the
Quality Assurance Project Plan approved for this project, which has been included as Appendix A. The
Quality Control Evaluation Report, included as Appendix B, describes the quality' and completeness of
the collected data sets.
Primary data sets consisting of ac power generated by the PV system ("Ppv") and power consumption by
the host building ("Pbl") were collected at all sites. Solar irradiance and ambient temperature data were
also collected at all sites, to allow simulation of PV system output (needed to verify proper operation of
each PV system). All measurements were recorded by a Campbell Scientific model CR10 datalogger,
equipped with a modem. The CR10 recorded 15 minute averages of each input data field for subsequent
retrieval and review. In two cases, the physical distance between the datalogger and the building load
meter prohibited load monitoring by the CR10, In these cases, the participating utility provided building
load data separately.
3. 1.1 Solar Irradiance and Ambient Temperature Data
Solar irradiance and ambient temperature data were measured and recorded on a continuous basis
throughout the study period (with minor exceptions) as a means of verifying proper PV system operation.
Four irradiance quantities were measured at the site of each PV system: 1) global horizontal ("GH")
irradiance; 2) direct normal ("DN") irradiance; 3) diffuse horizontal ("DH") irradiance; 4) and plane-of-
array ("POA") irradiance. The first three of these were monitored by a standard Ascension Technology
Rotating Shadowband Pvranometer (RSP), which uses a single LI200SZ pyranometer (LiCor
Corporation, Lincoln, NE) and shunt resistor to produce a voltage signal proportional to solar irradiance.
Details on the operation of the Ascension Technology RSP may be found in Appendix C. POA
irradiance was monitored by a separate LI200SZ pyranometer mounted to the PV array, at the array's tilt
angle. A Campbell Scientific model 107 temperature probe was used to monitor ambient air temperature
at the PV array.
Measurements of GH and POA were made once per second and averaged once per minute. As its
measurement is dependent on shadowband rotation, GH measurements were made at the frequency of
rotation, i.e., once per minute. DN irradiance was calculated every second, based on the current
measurement of GH and the DH measurement from the previous minute.
3.1,2 PV System Generation and Building Load Data
General Electric model IW-70-1S or VW-64S kilowatt-hour meters with type D72 pulse initiators (or
comparable Sangamo or Westinghouse meters) were installed by each participating utility to allow for
3-1
-------
continuous monitoring of PV system output and host building load. Once each minute, the CR10
datalogger measured pulse frequency (in Hz) from each meter, multiplied by 3,600 to determine the
current pulse rate per hour, and then multiplied by the pulse constant (kWh/pulse) associated with each
meter. This calculation provided an instantaneous measurement of power production by the PV system
or power consumption by the host building. These minute-by-minute calculations were averaged and
recorded every 15 minutes,
As noted above, building load meters at two of the sites, White Plains, NY (EPA05) and Austin, TX
(EPA 12). were too distant from the datalogger to permit collection of building load data. In each case,
these data were collected by the participating utility and forwarded to Ascension Technology's
headquarters in Waltham, MA.
3.1.3 Additional Instrumentation
In addition to the irradiance, temperature, and power measurements, two of the participating utilities
chose to expand the measurement capabilities of the data acquisition system by having project staff
install additional instrumentation at a total of four sites. At the request of Wisconsin Public Service,
staff installed sensors in the Ashwaubenon, WI system which monitor ac power in each phase of the
three-phase system, dc currents and voltages on the positive and negative legs of each bipolar array, and
cell temperatures in each array. At each of the Southern California Edison systems (EPA14, EPA15, and
EPA 16), project staff installed sensors which monitor reactive power from the inverter, cell
temperatures, and dc voltages and currents on the positive and negative legs of each array. Because the
system at Palm Desert has arrays facing east, south and west, each array has its own POA sensor.
3,2 Daily Data Retrieval and Review
3.2.1 Data Retrieval
All data collected by the CR10 at each site was collected via modem each night by a PC at the Ascension
Technology office using Campbell Scientific PC208 software. Calls to each CR10 were placed shortly
after midnight, local time. In the event of a communications failure, the software was set to retry a call
to a site up to four times per night. Although infrequent, there were occasions on which data could not
be collected due to, for example, a faulty phone line or malfunctioning RSP power supply. In these
cases, project staff would attempt to ascertain the cause of the problem and attempt to rectify it the
following morning. The CR10 has sufficient memory capacity to store approximately 32 days of data
records on board, making data loss due to telecommunications problems alone essentially impossible.
3.2.2 Data Quality Checking
Data retrieval from the CRIOs was followed by automated data processing, using software developed by
Ascension Technology. This software performed a variety of functions, including checking the
completeness of each data set, checking that each data point was within a specified range, calculating
several data quality indicators, and producing output files summarizing the daily irradiance and PV
system performance data.
Each workday morning, the daily summary (ies) of each site for the previous day was reviewed by
project staff. If the summary indicated that any data fields were missing or out of range, or that either
the RSP or PV system were not operating properly, an investigation would commence into the cause of
the problem. Usually, this investigation would begin with a graphical review of the data fields in
question. Many problems (or "events") were easily diagnosed and corrected from the Ascension
3-2
-------
Technology office. Others entailed hardware failures, requiring a site visit by Ascension Technology or
utility personnel to perforin maintenance. In all cases, detailed records were kept regarding the nature of
the event and actions taken for its resolution. Events affecting PV system performance are discussed
generally in Section IV, and a site-by-site description of events is included in Appendix D.
3.3 Monthly Data Processing and Reporting
Following the end of each month, the datafile for each site would be truncated so that it contained only
data from the preceding month. This truncated data file was processed by the software discussed above,
producing summaries of irradiance and system performance for the entire month. After a final review of
these summaries, they and the corresponding data files would be copied onto computer diskettes, and
sent to each of the participating utilities, along with a letter describing any data "events" for the month.
A monthly technical progress report describing the performance and operating status of all sites was
prepared and forwarded to the EPA Air Pollution Prevention and Control Division and the EPA
contracting officer for this project. Finally, all data files were copied onto computer tape cartridges for
archival storage.
3-3
-------
Chapter 4
PV System Performance History
Of the sixteen PV systems installed by this project, all but two suffered events during the study period
which temporarily limited system output or prevented generation altogether. These events, and the
resulting losses in generation are presented in detail in Appendix D, and summarized in Table 4-1,
Table 4-1, SUMMARY OF EVENTS AFFECTING SYSTEM GENERATION
Inverter Outages
Snow Cover Outages
Other Outages
Site
Number
Percent of
Annual
Generation
Number
Percent of
Annual
Generation
Number
Percent of
Annual
Generation
Pittsburgh, NY
1
7
5
12
3
5
Berlin, CT
0
0
3
6
0
0
Pleasantville, NJ
l
3
3
2
"1
£.
5
Brigantine, NJ
2
4
1
0
1
1
White Plains, NY
0
0
3
5
2
8
Scottsdale, AZ
1
1
0
0
3
2
Peoria, AZ
0
0
0
0
0
0
Ashwaubenon, WI
0
0
4
10
4
3
Denmark, WI
1
J
3
4
1
1
Minnetonka. MN
1
2
3
6
1
1
San Ramon, CA
2
2
0
0
5
3
Austin, TX
6
6
0
0
4
3
Flagstaff, AZ
0
0
4
5
1
6
Barstow, CA
0
0
0
0
0
26
Edwards AFB, CA
0
0
0
0
1
2
Palm Desert, CA
0
0
0
0
0
0
4_1
-------
Inverter-related problems were the most vexing of the generation-limiting events. In all, the 27 inverter-
related events detailed in Appendix D resulted in a generation loss of 12,740 kWh, approximately nine
percent of the combined generation of these systems over the relevant time periods.1 Inverter-related
outages ranged from 0 to 7 percent of gross annual generation (measured generation plus estimated
generation lost due to events during the relevant 12-month period). The systems in Pittsburgh, NY and
Austin, TX fared far worse than the others in this respect. In Pittsburgh, the system was turned off for
over three months because the inverters were the suspected source of electromagnetic noise that
interfered with computers in the host building. In Austin, two of the three inverters were found to be
operating intermittently once the PV meter was installed, and nearly two months elapsed before they
were shipped back to the manufacturer for repair. This site also suffered frequent inverter outages due to
DC injection (the inverter sensed a DC component in its output and shut itself down).
As a result of the inverter-related outage experience in this project, the inverter manufacturer made
several design changes and increased product testing across their full line of inverters. In addition, they
extended the product warrantee for the EPA project installations.
As Table 4-1 indicates, snow cover was also a frequent cause of PV system outages for those systems
located in northern locations or at high altitude (i.e., Flagstaff, AZ at 6,910 feet). Of those systems that
were vulnerable to snow cover, the estimated energy loss as a result of snow cover ranged from less than
one percent (Brigantine. NJ) to 12 percent of gross annual generation (Pittsburgh, NY). It is interesting
to note that while snow cover resulted in a generation loss of about 10 percent of measured generation
for the system in Ashwaubenon, WI, the loss was only about four percent in Denmark, WI, just a few
miles away. The difference between losses at these two sites is accounted for by their differing tilt
angles. The Ashwaubenon system is at a 15 degree tilt, whereas the Denmark system is at a 25 degree
tilt. The effect of tilt angle on losses due to snow cover can also be seen by comparing the systems in
Pleasantville and Brigantine, NJ. The Pleasantville system has a tilt angle of 15 degrees and lost two
percent of its generation to snow cover, while the Brigantine system, at a tilt of 25 degrees lost well
under one percent to snow.2
A variety of other outages occurred during the study period, not all of which have identified causes. Of
those "other" outages for which a cause was identified, the most frequent was, by far, a fuse failure in the
DC disconnect switch. Such failures occurred a total of 17 times, at 11 of the sites. It was determined
that the original fuses in the DC disconnect switches did not have the proper surge rating. As they failed,
they were replaced by "slow-blow" fuses which were rated for 600V DC. None of the replacement fuses
has failed to date.
'Note that because the systems did not all commence operation on the same date, performance and outage data cover a different time period for
each site. The figures reported here reflect at least one year of data for each site.
;PV systems designed by Ascension Technology now typically use a tilt angle of 25 degrees, but on flat roofs snow accumulation at the bottom
of the array can still lead to performance degradation.
4-2
-------
Chapter 5
Dispatchable Battery Storage Model
5.1 Modeling Approach
Where peak loads do not coincide with peaks in the available solar resource, the value provided by a PV
system can sometimes be greatly enhanced by the addition of dispatchable battery storage. Although
some energy is lost both in charging and discharging a battery array, the ability to dispatch energy
generated by the PV system during utility peak loads (or for that matter the peak loads of a transmission
line or distribution feeder) allows the PV generation to be used to reduce generation by a utility's highest
operating cost units, which are typically used only during peak periods. Furthermore, by providing better
alignment between a utility's peak load and the availability of PV generation, the need for additional
peaking capacity (or an upgrade to a transmission line) may be deferred. Dispatchable storage may
provide analogous benefits in the transmission or distribution network, to reduce peak loads of
transmission lines, substations, or feeders.
To investigate the degree to which dispatchable battery storage would improve the ability of each PV
system to offset load during utility peak load hours, a model for battery charging, discharging, and
dispatch control was developed. The approach taken was to maximize the contribution of each PV
system during the highest utility load hours of each day, by simulating the daily operation of each
system's inverter(s) at its (their) peak capacity for as long as possible. The duration of operation each
day was determined by the amount of energy actually generated each day, as described further below.
It is important to state at the outset that this simulation is intended solely as a heuristic illustration, and in
fact could not be implemented. This is the case because 1) the simulation dispatches all of the energy
generated over the course of a day by the PV system during the peak hour(s) of that same day, regardless
of when the peak hour(s) occur; and 2) the simulation assumes perfect prior knowledge both of when the
daily peak load will occur3 and the total amount of solar energy available each day. There are
innumerable storage charging and dispatch strategies which could have been investigated. The situation
is complicated still further by the fact that addition of storage allows one to consider resizing the inverter
to optimize power delivery during hours of dispatch rather than optimizing the direct, instantaneous
conversion of solar energy to AC power. A comprehensive review of this diversity of charging and
dispatch algorithms, and their implications for the use of PV as a demand-side management or emission
control strategy was beyond the scope of this study. The algorithm used in this study was selected for its
simplicity, and because it is believed that this algorithm provides an upper bound to the peak load
reduction capability of each PV system as installed, operating under a daily dispatch. As the use of PV
grows, it is reasonable to expect that an optimal algorithms will be developed to size, charge and
discharge PV systems which integrate battery storage.
¦'This assumption is not completely unrealistic. Utilities routinely predict hourly load levels hours, days, and weeks in advance. The accuracy
of these predictions will, of course, vary tremendously depending on the quality of the information available to a utility for load simulations.
5-1
-------
The addition of dispatchable storage to a PV system will clearly alter its hourly output to the utility
system. Because a utility's pollutant emission rates may vary substantially from hour to hour, storage
will also alter emission offsets due to PV generation. Discussion of how storage changes the DSM and
environmental characteristics of the PV systems is included with the discussion of these characteristics
of the systems as they actually operated in Chapters 7 and 8 respectively. Discussion of the PV systems'
ability to offset load at the customer level does not, however, include battery storage. The storage model
was designed to dispatch to maximize PV contribution during each utility's peak load hours, and
inherently assumes that storage would be added for the benefit of the utility rather than the individual
customer.4 Using this dispatch algorithm, the output of the PV/storage system at the time of customer
peak loads would be highly dependent upon the correlation of customer load to utility load, and would
not, therefore, be generalizable to other customers.
5.2 Description of Model
As discussed above, the dispatchable storage model employed for this study concentrates all energy
converted by the PV systems on a daily basis into the highest utility load hours encountered on a daily
basis. This was accomplished in a four-step process. First, total PV generation was determined for each
day, and multiplied by 0.75 to account for losses in charging and discharging the battery. Second, the
de-rated daily generation of each system was divided by the applicable nameplate AC rating of the
inverter in each system to determine the number of hours the inverter could be operated at its full rated
power (4, 8 or 12 kW). Third, the model identified the peak load hour for each day. Where the peak
load occurred during two or more hours in a day, the model selected the first occurrence as the peak
hour. Finally, the model determined which hours to dispatch the system. Energy generated throughout
the course of the day was distributed as follows: if there was sufficient energy (after de-rating) to operate
the inverter at full power for at least one full hour, the model simulated full power at the utility daily
peak. If less than one hour's operation was stored in the battery, the inverter would be operated so as to
drain the battery during the peak hour. If enough energy was generated during the course of a day for
more than one hour of operation at full power, this was equally distributed around the peak hour. For
example, if a PV system with a 4 kWac inverter generated 15.5 kWh on a day on which the utility's
peak load occurred between 7:00 and 8:00 p.m., the model assumes that (0.75*15.5=) 11.6 kWh would
be available after losses, and would dispatch the inverter to operate at 4 kW for two hours starting at 6:00
p.m. followed by one hour of operation at 3.6 kW.
¦•The benefits of more optimal utility operations would presumably be passed on to customers in the form of reduced rates.
5-2
-------
Chapter 6
Host Building Load Impacts
One of the primary goals of this project was to determine the ability of PV systems to reduce building
peak loads. As many of the buildings on which the PV systems were installed have a demand charge
component in their electric rates (i.e., a portion of their bill is based on the highest demand for power
during the month), the power generation by the PV systems during the highest building load hours is of
primary interest. PV system operation during peak load hours was studied in two ways: 1) by identifying
PV generation during the peak building load hour in each month and 2) by developing load duration
curves ("LDC'"s) from net and gross building load data and observing the change in each building's LDC
due to the PV system.
The degree to which a PV system can reduce peak building loads is determined by the correlation of the
building load and the available solar resource. Buildings that have large or dominant loads driven by
solar gain (e.g. air conditioning) are therefore good candidates for PV demand-side-management.
Alternatively, buildings which have peak loads at night or during the winter will obtain relatively little
peak-load reduction from PV. The group of buildings participating in this project contained examples of
both very good and very poor candidates for the use of PV as a measure to reduce customer peak load, as
the remainder of this section will describe.
6.1 Building Load and PV System Data
As described in Chapter 3. average building load and PV system generation were recorded every 15
minutes throughout the study period. Data were further aggregated to one hour averages to correspond
with the hourly system load and emission data collected from the participating utilities. Because
generation by the PV system served host building loads directly, the building load meter recorded
building load net of PV generation. Gross building load was therefore determined by adding hourly PV
output to net building load.
6.2 Data Analysis
The monthly peak impact investigation was straightforward: for each of the 15 months in the study
period, the date and time of the peak gross load for each of the host buildings was determined. The
average power generation by the PV system during that hour was determined, both in kW and as a
percentage of each system's capability under standard operating conditions (SOC) of 1,000 W/nr plane-
of-array (POA) insolation and 20°C ambient air temperature. The results of this analysis are illustrated
by bar charts such as the example in Figure 6-1.
6-1
-------
It is important to note that the amount of power delivered by
a PV system during a building's monthly peak load hour will
not, in most cases, provide an accurate indication of the
amount by which the monthly peak load is reduced. The
reason for this is that the PV system will not be operating at
the same power level during all of the highest load hours in a
month. If there are hours for which the gross load level is
close to the monthly peak, but during which the PV output is
less than that during the peak hour, net load for these hours
may exceed net load during the hour at which the gross load
attains its monthly peak level. In this case, the reduction of
peak load will not be the system's output during the peak
gross-load hour, but the difference between the peak gross
load and the peak net load for the month. This difference
will usually be less than the PV system's output during the
peak gross-load hour. The data collected for this project
provide several examples.
A more comprehensive and accurate understanding of the
effect of the PV systems on peak building loads can therefore be gained by observing the effect of the
systems on building load duration curves. LDCs are constructed by sorting load values for the period of
interest in descending order. The sorted hourly load values are then plotted with the load level on the
ordinate and the rank order of each load value on the abscissa. The chronological continuity of the data
is lost in the sorting process, but by plotting load data in this way, one can easily focus on how PV
system operation affects building load during the highest building load hours.
As is illustrated by Figure 6-2, the PV system's effect on a building's LDC can be determined by
comparing an LDC produced from gross load values to one produced from net load values, or in other
words, comparing the building LDCs with and without the PV system. Where a difference exists
between a LDC based on gross load values and one based on net load values, the PV system can be said
to have reduced building load for the period in question. Unlike Figure 6-1 which indicates the amount
by which gross building load is reduced only at the peak load hour of each month, Figure 6-2 indicates
the impact of PV system operation at all hours (although only the highest load hours are shown). It is
important to note that any point along the net LDC will not necessarily represent the same hour as the
point immediately above it on the gross LDC. The reason for this is that gross load values must be
sorted independently of net load values in the construction of these curves. Again, the goal is to compare
each building's LDC writh and without the PV system. The hours at which peak gross loads occur may
well be very different from the hours at which peak net loads occur.
6.3 Results
The demand-side management effects of the PV systems are illustrated at the end of this section by three
types of charts. An example of the first type of chart is shown in Figure 6-2. This figure compares an
LDC produced from the host building's gross or total load data, represented in the chart by the solid line,
to an LDC produced from building load data net of PV generation, represented by the shaded area. The
difference between these two LDCs therefore indicates the effect of the PV system on the building's
Figure 6-1. Example of PV Capacity
Factor at Peak Building Load
6-2
-------
load at all hours (although to make this difference visible only
the 25 highest load hours have been included). The absolute
magnitude of the difference between the gross and net LDCs
can be read off of the scale on the left side of the chart. The
diamond-shaped data points represent the difference between
the LDCs normalized by the PV system's rating under
standard operating conditions. These data points therefore
represent the PV system's effect on the building's LDC per
kW50E of installed PV capacity.
For each site there is a set of six building LDC charts: one
each for the five calendar quarters during the study period and
a sixth illustrating the highest 25 load hours recorded during
the entire study period. Note that building load is scaled
independently in each chart, so that the PV system's effect,
which is often a small fraction of gross building load, is most
visible.
752
;732
CD 712
692
0 S 10 15 20 25
Highest Building Load Hours
Figure 6-2. Example of Building Load
Duration Curves.
Figure 6-1 is an example of the second type of chart used in this section. The shaded region in the
background indicates monthly peak of the gross building load normalized by the peak load recorded
during the entire study period. The vertical bars in this chart indicate generation by the PV system
during the hour in which the monthly peak load occurred. As in the last chart, PV generation has been
normalized by the system's rating under SOC. The bars therefore represent the PV system's capacity
factor during the peak hour for each month.
The final type of chart used to illustrate the
correspondence of PV system generation to building load
is that shown in Figure 6-3. In this figure, the building's
monthly load factor (monthly energy consumption divided
by the product of the highest peak load recorded for the
building and the number of hours in the month) is shown
in the background, with average monthly PV capacity-
factors (the ratio of measured generation to what the
system would have generated had it operated at rated
capacity 24 hours a day throughout the month) in the
foreground. Again, each PV system's rating at SOC has
been used to derive the capacity factors.
Each set of figures is followed by a brief description of
any events that would have affected system operation in
each quarter.
1.0
0.8
a
T3
S 06
a
0.4
0.2
00
3rd Qtr '93 1st Qtr'94 3rdQtr'94
4th Qtr '93 aid Qtr '94
Figure 6-3. Average Building Load
Factor and PV Capacity Factor.
6.3.1 Pittsburgh, NY
Figure 6-4 illustrates the PV system's effect on the host
building's load duration curve. These charts show that the PV system had little impact in any season,
and almost none at all during the top 25 hours of the study period. By comparing peak building loads
6-3
-------
across quarters, one can see that this building's peak loads were nearly a factor of three higher during
winter months (when the solar resource is at a minimum) than during summer months (when the solar
resource is greatest). Note that the PV system was shut down during the five highest load hours in the
third quarter of 1993: and that one of three inverters was malfunctioning during the following 20 hours in
the LDC. PV system output was also limited by an inverter failure during most or all of the peak hours
in the second quarter of 1994.
The pronounced seasonality of this building's load is clearly illustrated in Figure 6-5, as is the seasonal
variation in PV system generation. Figure 6-6 presents monthly building load factors and monthly
capacity factors for the PV system. Aside from August and September of 1993 when the system was
shut down, the PV system's minimum monthly capacity factor was just over 1 percent in January of
1994, and its maximum capacity factor was 19 percent in May and July of 1994. The annual average
capacity factor for the PV system (based on the 12 months starting October 1, 1993) is 10.2 percent.
6.3.2 Berlin, CT
The building LDCs for the Berlin PV system are similar to those for Plattsburg, in that the highest
building loads occurred in winter when the PV system's output was at its minimum. As illustrated in
Figure 6-7, the PV system contributed little at building peak during the fourth quarter of 1993 and first
quarter of 1994. During most of the highest load hours of the second and third quarters, however, the
system operated at a relatively high fraction of its rated output.
The PV system's output during the peak hour in each month is shown in Figure 6-8. This figure
indicates peak reduction close to 90 percent of system capacity (at SOC) for several months, but almost
no peak reduction during the winter months when the highest building loads occurred. The monthly
capacity factor for this system varied by a factor of 7 over the course of the study period, from a low of
about 3 percent in January 1994 to a maximum of over 21 percent in May of that year. The average
capacity factor for the 12 months beginning in October 1993 was just over 15 percent.
6.3.3 Pleasantville, NJ
Figure 6-10 indicates that peak loads for the host building in Pleasantville occur in the third quarter. All
of the highest 25 load hours occurred in the third quarter of 1993. as can be seen by comparing charts (a)
and (f) in the figure. Chart (a) shows a relatively constant effective PV capacity of about 0.2 kW/k\VS0C
for this quarter. Note, however that the PV system was disabled by inverter failures for much of the
quarter and that effective capacity is higher by approximately a factor of three in the third quarter of
1994.1 This suggests that the effective PV capacity of about 0.6 kW/kW80C shown in chart (e) of Figure
6-10 may be a better indication of the PV system's effect during the highest load hours.
Figure 6-11 shows PV capacity factors during the monthly peak building load hours, and indicates that
the system provided power at peak load for all months other than December 1993, January 1994 and
February 1994. For the 12 months beginning October 1993, the average PV capacity factor at building
peak load was 46%. Average monthly capacity factors range from 6.6% to 22% as shown in Figure 6-
12, The average annual PV capacity factor for the 12 months beginning October 1, 1993 was 14.6%.
'Inverter failures also limited system output during the third quarter of 1994. However, all inverters were functioning properly during the peak
load hours of this quarter.
6-4
-------
6.3.4 Brigantine, NJ
In Brigantine, the building load duration curve (presented in Figure 6-13) was reduced by approximately
0,5 kW/kWS6C for most of the highest load hours in the third quarter of 1993 and the second and third
quarters of 1994. The highest recorded loads for this site all occurred in the third quarter of 1994, thus
the correspondence between charts (e) and (f) in the figure. The PV system had an insignificant effect on
the building LDC during the highest load hours the fourth quarter of 1993 and the first quarter of 1994.
However, this does not mean that the PV system is ineffective as a DSM measure for all of these months.
Figure 6-14 indicates that the system did provide significant peak shaving for the months of October and
December 1993 and February 1994, The fact that the top end of the LDCs for the fourth quarter of 1993
and the first quarter of 1994 were unaffected by the PV system indicates that the PV system provided
little or no generation during the peak hours in November of 1993 and March of 1994 (these loads
occurred after sunset), which is also reflected in Figure 6-14.
Figure 6-15 illustrates the seasonal variation in the PV system's average monthly capacity factors and the
building's monthly load factors. Note that a meter malfunction prevented load measurement from
4/15/94 through 6/18/94, Monthly capacity factors range between 7.4% in December to 24% in June.
The annual average capacity factor is 17%.
6.3.5 White Plains, NY
As demonstrated by the data in Figure 6-16, the PV system's effect on this site's building load shows
little of the consistency exhibited in Brigantine. Effective PV capacity varies considerably across the top
25 hours in each quarter, as well as the top hours of the study period. The effect on the building's LDC is
smallest in the first quarter of 1994 due to the minimal solar resource available during that period.
Surprisingly though, the effect on the building's LDC in the months of October through December 1993
is similar in magnitude to that found in the LDCs for the second and third quarters.
The hours at which the monthly peak loads occurred for this building are remarkably consistent.
Fourteen of the fifteen monthly peaks occurred during the hour preceding 11:00 a.m., and the fifteenth
occurred in the following hour. This suggests that a large portion (and perhaps all) of the highest loads
occurred during the late morning or early afternoon—a time at which the PV system could be expected to
be generating at a substantial fraction of its rated output. The fact that there are numerous high load
hours for which this was not the case cannot be explained by system outages. As can be seen in Figure
6-17, the highest building loads in the third quarter of 1993 occurred in September, after an inverter
failure had been repaired. The only other outage recorded for this system occurred in June of 1994, well
after the highest load hours for that quarter had passed. One possible explanation for the variability of
effective PV capacity during building peak load hours is that these loads were not driven by solar gain in
the building2, and that cloudy or overcast skies occasionally prevented PV generation during high load
hours.
Figure 6-18 shows that the building's monthly load factor was relatively constant throughout the study
period, with the exception of the months of July and August in both years, when school was presumably
out of session. The monthly capacity factor for the PV system ranged from less than 2% in January 1994
;This is supported by the nearly uniform timing of the monthly building peaks. Peak loads may well have been driven by the regular daily use
of a particular piece of equipment, such as an electric oven used to prepare the school's lunches.
6-5
-------
to over 23% in May 1994, The average capacity factor for the 12 months beginning October 1993 was
13.1%.
6.3.6 Scottsdaie, AZ
Figure 6-19 demonstrates that the PV system in Scottsdaie consistently provided power during the
building's highest load hours, often in excess of 60 percent of its SOC rating, As the figure shows, this
system had the greatest effect on the building LDC in the third quarter of 1993 and the first and third
quarters of 1994. Reduction of peak load was quite reliable in the other quarters as well, though at a
lower level, Figure 6-20 shows the seasonal variation in this building's peak loads, and demonstrates
again that this building's load is well correlated to the solar resource. The PV system generated at over
55 percent of its rating in 12 of the 15 months in the study period.
Figure 6-21 shows the consistently high capacity factors this system achieved, ranging from a low of
16.3 percent in December 1993 to 26.5 percent in May 1994. This system's average annual capacity
factor of 21 percent was the highest measured in the project.
6.3.7 Peoria, AZ
The PV system in Peoria, AZ also had a remarkably consistent and reliable effect on the building's LDC
in all quarters, as illustrated by Figure 6-22. This figure indicates that the highest load hours in each
quarter all occurred during the day, and suggests that solar gain is a substantial contributor to building
load. The magnitude of the effective PV capacity in each quarter mirrors the seasonal variation in the
solar resource, which is clearly illustrated in the monthly PV capacity factors in Figure 6-24. During the
highest load hours of the study period, most of which occurred in the third quarter of 1994, the PV
system reduced the building's LDC by very close to half of its rated capacity at standard operating
conditions.
The power output of the PV system during the monthly peak load hour was often greater than the LDCs
suggest, as is illustrated in Figure 6-23. Here it can be seen that the PV system's contribution at monthly
peak was well in excess of half of its rated capacity for most of the months in the study period, rising as
high as 85% in March of 1994. The reason that the LDCs do not reflect the system's actual contribution
at the monthly peaks has to do with the fact that gross load and net load are sorted independently in the
LDCs. Although the PV system may be generating at or near its rated output during the highest load
hours on the gross load duration curve, the system's output may be very low or even zero during the
highest load hours on the net LDC. The result may be a smaller change to the building's load duration
curve than would be expected, if only the data in Figure 6-23 were considered. This has important
implications for demand charges, and illustrates the importance of comparing LDCs of gross and net
load.
The monthly average capacity factors for this PV system range from 12.7% in December 1993 to 24.0%
in May 1994. The annual average was 18.9% for the 12 months beginning October 1994.
6.3.8 Ashwaubenon, WI
As indicated by the data in Figure 6-25, the PV system in Ashwaubenon reduced the gross load of its
host building by approximately 40 percent of its rating during most of the highest load hours after it
6-6
-------
began operating in the fourth quarter of 1993.3 The system had very little effect on the building's LDC
in the first quarter of 1994, but during the second quarter (when the highest loads were recorded) gross
load was reduced by more than 60 percent of the system's rating in most of the highest load hours. For
the 25 highest load hours in the third quarter of 1994, the PV system reduced the LDC by a minimum of
just under 40 percent and a maximum of about 70 percent of its SOC rating.
Figure 6-26 shows that this system operated at or above 40 percent of its rating during the peak hour in
all but two of the 12 months for which building load was recorded, and at more than 80 percent of its
rating for four of those months. The average monthly capacity factors in Figure 6-27 indicate
pronounced seasonal variation in PV generation. Capacity factors vary from less than 1 percent in
January' 1994 to over 27 percent in May of 1994. The average annual capacity factor for this system for
the 12 months beginning October 1994 is 14,3 percent.
6.3.9 Denmark, WI
fn Denmark, WI, the peak output of the PV system was poorly correlated to peak building loads, as is
illustrated in Figure 6-28. Effective PV capacity rarely rises above 20% of SOC rating in these charts,
and is particularly low during the fourth quarter of 1993 and first quarter of 1994, when most of the
highest recorded building loads occurred. It should be noted, however, that this system was installed on
a residence, which is not subject to demand charges. For this reason, correlation of PV generation to
building load may be immaterial. Figure 6-30, which presents month-by-month PV capacity factors and
building load factors may be more informative regarding the customer bill impacts of this system, since
residential customer bills are driven primarily by energy consumption rather than peak load. Note also
that there are no load duration curves for the third quarter of 1993 because building load monitoring at
this site began in November of that year.
PV generation during the monthly peak load hour is shown in Figure 6-29. This figure provides
additional evidence that this site was not a good application for PV as a demand reduction measure. For
many months, PV generation at building peak is quite small or zero. For others, the PV system generated
at a high fraction of its rated capacity during the peak load hour. However, the fact that the high capacity
factors for the peak load hours in these months do not appear in the corresponding LDCs suggests that
there are hours in these months for which the building load is near the monthly peak, but during which
the PV system generated at a much lower level.
The monthly PV capacity factors shown in Figure 6-30 range from 8.4% in January 1994 to 23% in July
1994, The annual average capacity factor was just over 16% for the 12 months beginning October 1994.
6.3.10 Minnetonka, MN
Whereas the results for the Denmark, WT system indicate that PV was a poor DSM measure for that site,
the system in Minnetonka, MN is an example of a building in the northern U.S. for which PV is a good
match. The LDCs in Figure 6-31 indicate reductions of the building's highest loads by about 60 percent
of the PV system's SOC rating during the third quarter of 1993 and the second and third quarters of 1994.
For the highest 25 load hours of the entire study period, the minimum effect on the building's LDC was
40 percent of the system's SOC rating. Although the PV system had little impact on the building LDC
'Note that ] 5 of the 25 highest load hours in this quarter occurred while one of the three inverters in this system was inoperative.
6-7
-------
during the fourth quarter of 1993 and the first of 1994, peak loads during these quarters were less that
half of those in other quarters.
Figure 6-32 presents PV generation during the monthly peak load hour. This figure illustrates the
seasonality of this building's monthly peak loads and further demonstrates the ability of the PV system to
provide power when the building's demand is greatest.
The average monthly capacity factors shown in Figure 6-33 range from 6% in November 1993 to 27% in
May 1994. The average annual capacity factor for this system for the 12 months starting October 1993
was 15.3%.
6.3.11 San Ramon, CA
The LDCs for the San Ramon system in Figure 6-34 indicate a significant contribution from the PV
system during the top load hours in all four quarters (note that because load monitoring began on March
7, 1994 the LDC for the first quarter of 1994 contains less than one month of load data). The system's
effect on the LDC for the highest 25 load hours is relatively stable at about 70% of the system's SOC
rating. The relative stability of the effective PV datapoints in these charts suggests that peak and near-
peak loads for this building tended to occur at the same time of day, and may well have been driven by
solar gains (i.e. air conditioning).
Figure 6-35 shows that the PV system's generation during the monthly peak load hours was also quite
high, averaging 60% for the months between March and December 1994 (note that load measurement
began in March of that year). Average monthly capacity factors for the system are shown in Figure 6-36.
These range from 11 percent in January 1994 to over 30% in June 1994, The annual average capacity
factor for calendar year 1994 was 20.9 percent.
6.3.12 Austin, TX
The load duration curves in Figure 6-37 indicate that the PV system in Austin had little effect on peak
loads during the fourth quarter of 1993 (when the highest loads were recorded), but was effective in
reducing peak loads during the first three quarters of 1994. Although there are a few hours during the
fourth quarter of 1993 for which the PV system had a substantial effect on the LDC, for most hours the
LDC is reduced by well below 20% of the system's SOC rating. In the following quarter, loads at the top
of the LDC are typically reduced by about 40% of the system's rating, although the reduction rises to as
high as 80%. In the second and third quarters of 1994, load reductions in excess of 60% of the system's
rating are typical during the highest load hours, despite outages caused by inverter and fuse failures.
Since the 25 highest loads recorded during the study period all occurred during the fourth quarter of
1993, chart (e) in the figure duplicates chart (a).
The PV system's output during the peak load hour in each month is shown in Figure 6-38. These results
are consistent with the LDCs, indicating substantial load matching capability in all months other than
November, December, and January. Monthly average capacity factors for the PV system ranged from
9% in May 1994 (due to inverter and fuse failures) to 24% in July 1994, as shown in Figure 6-39.4 The
annual average capacity factor for the 12 months beginning November 1993 was 15.1 percent.
4Nlote that two-thirds of the system was shut down for two weeks in May due to an inverter failure.
6-8
-------
6.3.13 Flagstaff, AZ
After the PV system in Flagstaff began operating in February 1994, its output matched building loads
better than any of the other systems installed in this project. Figure 6-40 shows that generation by the PV
system reduced the host building's LDC by over SO percent of its rating in all but two of the 25 highest
load hours in the first quarter of 1994, and generation exceeded 80 percent for nine of those hours. In the
second quarter of 1994, the PV system operated at 100 percent of its capacity rating during the single
highest load hour, and again, the LDC was reduced by at least 50 percent of system rating in all but two
of the highest 25 load hours. The figure shows a similar impact on the LDC for the third quarter of 1994,
with a load reduction in excess of 60 percent of system capacity- in most of the highest load hours. Many
of the highest load hours in the fourth quarter of 1994 occurred at the end of November and beginning of
December, at a time when the system had been manually shut down for unknown reasons. This fact is
reflected in chart (d) of the figure, which shows that the PV system had little effect on the LDC for many
of the highest load hours in that quarter.
Figure 6-41 shows the PV system's generation during the peak hour in each of the months in the study-
period. The system generated at over 80 percent of its capacity during the peak hour in four of the
months, reaching 100 percent in April 1994. Note that an outage in late November and early December
1994 resulted in zero PV system output during the monthly peak load hours. Aside from the partial data
set for February 1994, average monthly capacity factors ranged from 5.9 percent in January 1995 to 27.4
percent in May 1994, as illustrated in Figure 6-42. For the 12 month period beginning with February
1994, the average annual capacity factor for this system was 19.5 percent.
6.3.14 Barstow, CA
The PV system in Barstow, CA did not begin operation until June 9, 1994. The time period of the
analysis for this system therefore runs through June 1995. The data in Figure 6-43 indicate that the PV
system at this site did little in any season to alter the building's load duration curve. This is due in large
part to the fact that this building's highest net load hours tended to occur close to or alter sunset. It
should also be noted that much of the PV array was shaded for most of the day5. In the summer months,
shading reduced this system's output to less than 15 percent of simulated generation during the afternoon
hours. It should be noted that this building is a residence, which accounts for the fact that its peak loads
do not occur during daylight hours. Because residential rate structures rarely include demand charges
(precisely for the reason that residential peak loads are not typically coincident with utility peak loads)
the fact that this building's LDC remained relatively unchanged by the PV system would have little
bearing on its economic viability.
Although this PV system did little to modify the building's LDC, Figure 6-44 shows that the system did
generate at a substantial fraction of its rated capacity during the peak gross load hours of most months.
Although it would appear so at first, the information in this figure is riot at odds with the charts in Figure
6-43. Figure 6-44 indicates only how the system performed during the hour of highest gross load in each
month. The fact that the building's LDCs are reduced by only a small fraction of the system's rating
reflects the fact there were many nighttime hours during which the building's load was nearly as high as
the quarterly peak load.
5 A number of other candidate homes w ith better solar apertures were proposed for the installation of this system. but were rejected by the
Marine Corps Logistics Base.
6-9
-------
Monthly capacity factors ranged from 12 percent in July 1994 (note that the system was shut down for 9
days that month) to 21.4 in April 1995, as illustrated in Figure 6-45. For the 12 month period beginning
June 1994, the annual average capacity factor for this system was 16.7 percent. Again, these capacity
factors were reduced due to significant shading at this site.
6.3.15 Edwards AFB, CA
The load duration curves in Figure 6-46 demonstrate that the PV system at this site provided a consistent
reduction in the building's highest loads in all four quarters of 1994. The system reduced the LDC by
between 35 and 45 percent of its SOC rating for most of the highest load hours in the first three quarters.
In the fourth quarter, the LDC was reduced by about 20 percent of the system's SOC rating.
Figure 6-47 indicates PV generation between approximately 60 and 80 percent of the system's SOC
rating in the peak hour of each month, with the exception of April and December. The fact that this
figure shows a peak load reduction in excess of 80 percent of SOC for the highest load hour in the first
quarter of 1994 would seem to be at odds with the first quarter 1994 LDC in Figure 6-46. The different
conclusions one would draw from these two charts once again illustrates the point that it can be
deceiving to consider only PV system performance during the peak hour in each month. If there are
hours in any given month for which the gross load level is below, but close to that of the peak load hour,
then these hours may very well become the peak hours of the net load duration curve. This can happen
when the PV system generates at a higher level during the peak hour than it does during hours when
gross building load is slightly less than the peak load. This effect is particularly strong when the capacity
of the PV system is a substantial portion of the overall gross building load, as it is in this case.
Figure 6-48 shows exceptionally high capacity factors for the months of February through October 1994,
The annual average capacity factor for the 12 months beginning February 1994 is 20%.
6.3.16 Palm Desert, CA
The LDC charts for Palm Desert are quite similar to those for Edwards AFB in that they indicate a
consistent reduction in the building's highest load hours in all quarters, although for this system the
reduction in the LDC is a larger fraction of its SOC rating. Figure 6-49 shows that for all but a few of
the highest load hours in each quarter, the load duration curve was reduced by at least 50 percent of the
PV system's rating. As at Edwards AFB, the reduction in the LDC is not as great as that indicated for the
peak monthly loads in Figure 6-50. Again, this is due to the PV system operating at a lower level during
hours at which the building load was slightly below the peak load for the quarter, Because the PV
system doesn't offset as much of the gross load in such hours, they become the highest load hours in the
net LDC.
Average monthly capacity factors are quite high for this site as well, ranging from 10% in December
1994 to 25% in May of that year. The annual average for the 12 months beginning February 1994, is
18.7%.
6.4 Conclusions
Two general conclusions may be drawn from the above discussion. The first is the relatively self-evident
conclusion that if reduction of customer peak loads is the primary motivation for the installation of a PV
system, it is critical to investigate the correlation of building peak loads to solar irradiance. The set of
host buildings participating in this project included some with loads which were very well matched to the
6-10
-------
solar resource as well as some for which the match was very poor. The systems in Ashwaubenon, WI
and Scottsdale, AZ are examples of systems which reduced host building LDCs by a substantial fraction
of their SOC rating. The highest loads in these buildings occurred during the midday hours, when the
solar resource peaks. The systems in Barstow, CA and Denmark, WI, on the other hand had very little
effect on the host building's LDC, despite ample solar resource. Many of the highest building loads at
these sites occurred near or after sunset.
The second general conclusion to be drawn from the data is that the generation of a PV system during an
individual building's peak load hour provides little information regarding that system's ability to reduce
the building's peak monthly load, or to reduce demand charges. As discussed above, even if the system
generates at full power during the monthly peak, there may be hours during which building load is just
below the monthly peak and during which the PV system operates at a much lower level. In cases such
as this, there may be very little change in the building's net LDC and correspondingly small changes in
demand charges. The monthly peak load will have simply been shifted to another hour.6
'Of course, this may be precisely the intent, Many commercial and industrial electricity consumers face rate structures which strongly discount
charges for peak loads that do not occur during utility-defined peak periods. Usually, these peak periods are during daytime hours when the PV
system would be capable of reducing building load,
6-11
-------
(a ) T hird Q ua rte r 1 993
223 , , 1
2
a
183
-! % s
5 10 15 20 25
Highest Building Load Hours
(c) First Quarter 1994
654
634
ra 614
594
574
~~
»~««
5 10 15 20
Highest Building Load Hours
225
at 205
fli.-iipi'
(e } Third Quarter 1994
245 1
5 10 15 20
Highest Building Load Hours
(b ) Fourth Quarter 1993
633 —, 1
5 10 15 20
Highest Building Lead Hours
(d) Second Quarter1994
459
359
309
5 10 15 20 25
Highest Building Load Hours
(f)Top 25 Hours in Study Period
659 , , 1
cn 619
599
579
- 0.4
5 10 15 20
Highest Building Load Hours
Total Building Load 13 Net Building Load ~ Effective PV Capacity ¦
Figure 6-4 Building Load Duration Curves with and without PV
for Pittsburgh, NY
6-12
-------
PV Capacity Factor
atBuiiding Peak Load
3rd Q»r '93 istQtr'94 3rd Gtr'94
4Ul Q tf "93 2nd Q tr "94
Figure 6-5 PV Capacity Factor at Monthly
Peak Load for Plattsburgh, NY.
¦ PV Generation [j£jj Building Load
Monthly Building Load Factor
and PV Capacity Factor
O
o
a
u.
>.
'5
ea
ex
a
Q
D
13
CD
Li.
13
CO
o
3rd Oif '93 1sr0lr'94 3rd Qtr "94
4lh Qtr 93 2nd Qtr *94
Figure 6-6 Average Building Load Factor and
PV Capacity Factor for
Plattsburgh, NY.
Notes on System Operation;
3rd Q '93:
lnvener failure limits system to 2/3 power 7/1/93 - 7/16/93. System shut down completely 7/23/93 - 9/30/93.
4th Q '93:
Data loss 10/1/93 - 10/5/93 due to datalogger short-circuit.
1st Q *94;
System fully operational
2nd Q '94:
DC injection limited system to 2/3 power 4/1/94 - 4/25/94. DC fuse failures limited system to 2/3 power 6/5/94 -
6/15/94 and again 6/18/94 - 6/30/94.
3rd Q '94:
System fully operational
6-13
-------
(a ) Third Quarter1993
5 10 15 20
Highest Building Load Hours
{c) First Qua rte r 1994
696
85S
336
795
0.3 a
5 10 15 20
Highest Building Load Hours
(e ) T hird Qua rte r 1 994
751
731
711
U
>
Q.
5 10 15 20
Highest Building Load Hours
(b) Fourth Quarter 1993
5 10 15 20
Highest Building Load Hours
u
>
CL
(d) Second Q u a rte r 1 994
7?6
756
736
Q 5 10 15 20 25
Highest Building load Hours
(f) T op 25 Hours in Study Period
905
0 5 10 15 20 25
Highest Building Load Hours
Total Building Load
Net Building Load ~ Effective PV Capacity
Figure 6-7 Building Load Duration Curves with and without PV for
Berlin, CT
6-14
-------
PV Capacity Factor
atBuilding Peak Load
3rd 2tr '93 lsEQtr'94 3rdQ?r'94
4 Ui Q Jr '93 2nd Qtr'S4
Figure 6-8 PV Capacity Factor at Monthly
Peak Load for Berlin, CT.
| PV Generation Jjjj} Building Load
Monthly Building Load Factor
and PV Capacity Factor
O
o
f3
L_
>¦
O
ro
Q.
ra
U
o
a
t?
m
14.
X>
o
3rd G tr '93 1stOtr"94 3rdQtr*94
4th QU '93 2nd Q Er '94
Figure 6-9 Average Building Load Factor and
PV Capacity Factor for Berlin, CT.
Notes on System Operation;
3rd Q '93:
Inverter failure prevented system operation 7/17/93 - 8/6/93.
4th Q '93:
System fully operational
1st Q *94:
System fully operational
2nd Q 94:
System fully operational
3rd Q '94:
System fully operational
100% p-
80% -
60% -
6-15
-------
(a ) T h Ird Quarter 19 9 3
5 10 IS 20 25
H ighest Build in g Load H ours
(c) First Q ua rte r 1994
406 . 1
346
AsJwStt-j,*' *
tM/>\
Mm
5 10 15 20
Highest Building Load Hours
0.6 a.
0.£
25
(e) T hird Gua rte r 1 994
392
* *
I 4
tm* .
0.8 Q-
W'"r 0.6
>
Q.
5 10 15 20 25
Highest Building Load Hours
{b ) Fourth Quarter 1993
342
5 10 15 20
H ighesi Building Load Hours
(d) Second Quarter 1994
cq 386
376
5 10 15 20
Highest Building Load Hours
(f) T o p 2 5 Hours in S tu d y P e rio d
440 , 1
e
O
o.a
Q.
>
"w
ra
0.6
a.
n
O
>
0.4
CL
m
>
a
0-2
2=
yj
5 10 15 20
Highest Building Load Hours
Total Building Load 0 Net Building Load ~ Effective PV Capacity
Figure 6-10 Building Load Duration Curves with and without PV for
Pleasantville, NJ
6-16
-------
PV Capacity Factor
alBuildmg Peak Load
3rd Gtr'93 1siG?r'94 3rd Qtr'94
4in Qtr '93 2nd Qtr '94
Figure 6-11 PV Capacity Factor at Monthly
Peak Load for Pleasaritville, NJ,
f PV Generation UJ] Building Load
Figure 6-12 Average Building Load Factor and
PV Capacity Factor for
Pleasantville. NJ.
Notes on System Operation:
3rd Q "93:
Inverter failure limited system to 1/3 power for 43 days, 2/3 power for 5 days.
4th Q '93:
System fully operational
1st Q '94:
Invsner failure limited system to 2/3 power 1/18/94 - 3/11/94 and 3/18/94 - 3/31/94.
2nd 0 '94:
Inverter failure limited system to 2/3 power 4/1/94 - 4/22/94.
3rd Q '94:
DC disconnect fuse failure limited system to 2/3 power from 7/23/94 to 8/2/94.
6-17
-------
(a } T hird O ua rte r 1 993
5 10 15 20
Highest Building Load H ours
>
LL
(c) First Q ua rie r 1994
0,8
0 .6
I
0 5 10 15 20
Highest Buildrng Load Hours
(e ) T hird Qua rte r 1 994
5 10 15 20 26
Highest Building Load Hours
(b ) F o u rth Qua rte r 1 9 9 3
11 i — 1
=
1
0 5 10 15 20 25
Highest Building Load Hours
(d) Second Quarter 1994
5 10 15 20
Highest Building Load Hours
(f) Top 25 Hours in Study Period
18
17
16
.£
2 15
13
k
s.
M
m
5 1G 15 2D
Highest Building Load Hours
0.8 Q-
!
0.6 |
o
0.4
°-2 £
Total Building Load Q Net Building Load ~ Effective PV Capacity
Figure 6-13 Building Load Duration Curves with and without PV for
Brigantine. NJ,
6-18
-------
PV Capacity Factor
at Build ing Peak Load
100%
3rd QJf '93 1stQSr"94 3rd Qlr'94
4rh G If '9 3 2nd Qlr'94
Figure 6-14 PV Capacity Factor at Monthly
Peak Load for Brigantine, NJ.
PV Generation
Building Load
Monthly Building Load Factor
and PV Capacity Factor
o 100%
8C%
B0%
40%
3-d Qv '93 IslQtr 94 3rd Qlr'94
4 th Qtr *93 2ndQtr'94
Figure 6-15 Average Building Load Factor and
PV Capacity Factor for Brigantine,
NJ.
Notes on System Operation:
3rd Q '93:
DC disconnect fuse failure prevented generation 8/18/93 - 8/20/94 and again 9/15/94 - 9/20/94.
4th Q '93:
Inverter failure prevented generation 12/6/93 - 12/9/93.
1st Q '94:
System shut down 2/25/94 - 2/28/94. DC injection error prevents generation 3/5/94 - 3/17/94.
2nd Q '94: Building load meter recorded zero load 4/15/94 - 6/18/94.
3rd Q '94:
System fully operational
6-19
-------
(a ) Third Quarter 1993
5 604
0.6 «
o
5 10 15 20
H igies; Budding Load Hours
25
(c) F irs t Q ua rte r 1994
647
627
567
5 10 15 20
Highest Building Load Hours
(e)Third Q ua rte r 1994
s 625
xt
S SOS
05
C
5 5S6
"3
c
565
545
v A,;'
• I-
u
0.B a
0.4
5 10 15 20
Highest Building Lcaa Hours
25
(b) Fourth Quarter 1993
0 5 10 15 20 25
Highest Building Load Hours
(d ) Second Quarter 1994
(f) T o p 25 Hours in Study Period
Highest Building Load Hours
Total Building Load 0 Net Building Load ~ Effective PV Capacity
Figure 6-16 Building Load Duration Curves with and without PV for
White Plains, NY.
6-20
-------
PV Capacity Factor
at B u i id irig Peak Load
3rd Qlf '93 1slQtr'B4 3rd Otr'Sd
4th Qlr'93 2nd Qtr'94
Figure 6-17 PV Capacity Factor at Monthly
Peak Load for White Plains, NY.
| PVGemation |j EUIdingLoad j
Monthly Building Load Factor
and PV Capacity Factor
20%
3rd Qtr'0 3 tstQir'94 3rd Qlf'94
4th Ctr 93 2nd Qlr 94
Figure 6-18 Average Building Load Factor and
PV Capacity Factor for White
Plains, NY.
Notes on System Operation:
3rd Q '93:
Inverter failure prevented generation 7/22/94 - 8/13/94.
4th Q '93:
System fully operational
1st Q '94:
System fully operational
2nd Q '94:
DC disconnect fuse failure prevented generation 6/11/94 -6/30/94,
3rd Q '94:
System fully operational
6-21
-------
ta > Third Qua rte r 1993
179 —, 1
5 10 15 20
Highest Building Load Hours
(c) First Qua rte r 1 994
1S6
>
EL
5 10 15 20
Highest Building Load Hours
(e ) T hird Q ua rte r 1 994
193
O
>
5 10 15 20 25
Highest Building Load Hours
(b) Fourth Quarter 1993
155
c
'
3
»_
3)
a.
0-8
<-»
Q.
0 fi
a
O
>
CL
0-4
>
y
09
0.2
U
5 10 15 20
Highest Building Load Hours
(d) Second Quarter 1994
192
5 1G 15 20
Highest Building Load Hours
(f)Top 25 Hours in Study Period
5 10 15 20
Highest Building Load Hours
Total Building Load
Net Building Load ~ Effective PV Capacity
"1
Figure 6-19 Building Load Duration Curves with and without PV for
Seottsdale, AZ.
6-22
-------
PV Capacity Factor
atBuilding Peak Load
100%
3rd Q Jr '93 'stCir'94 3rd Qlr 94
4th Q Ir *93 2nd Qlr'94
Figure 6-20 PV Capacity Factor at Monthly
Peak Load for Scottsdale, AZ.
PV Generation
Building Load
Monthly Building Load Factor
and PV Capacity Factor
o 100%
60%
40%
20%
0%
3rd Q tr "93 1 si Qtr 94 3rd Qtf '94
4 111 Qtr'93 2nd Qtr'94
Figure 6-21 Average Building Load Factor and
PV Capacity Factor for Scottsdale,
AZ.
Notes on System Operation:
3rd Q "93:
System fully operational
4th Q '93:
System fully operational
1st Q '94:
One of two inverters shut down 3/4/94 - 3/6/94. Cause unknown.
2nd Q '94:
System fully operational
3rd Q '94:
System at 1/2 power 7/18/94 - 7/20/94 due to DC disconnect fuse failure.
6-23
-------
>
Q.
5 10 15 20
Highesl Building Load Hours
(c ) First Qua rte r 19 94
O
_J
S7i 7
— —
\
\ #
-
f 1 >».
V—
, 2*"1'
—
v-;,;k ••k- ^
»4RMf>«Ni«fe
re
06 S-
(3
o
>
0.4
a
Z
o
a
0 2 ^
5 10 15 20
Highest Building Load Hours
25
(©) Third Quarter 1994
2
3
m
0.6
0.4
5 10 15 20
Highest Building Load Hours
(b ) Fourth Quarter 1993
S 10 15 20
H ighest B irldirtg Load Hours
(d) Second Qua rte r 1 994
5 10 15 20
Highesl Building Load Hours
(f) Top 25 Hours in Study Period
11
„ ^x®]iiSPFIv
' ' j-i, -
i-ailS;yiai£E^^ifcf!a
C.8 °-
O
>
Q.
£ 10 15 20
Highest Building Load Hours
25
Total Building Load B Net Building Load ~ Effective PV Capacity
Figure 6-22 Building Load Duration Curves with and without PV for
Peoria, AZ.
6-24
-------
PV Capacity Factor
atBuilding Peak Load
o
<3
ts
IL
>s
o
w
a
ra
U
>
a
3rd Qtr'93 1stQEr'S4 3rd Qtr '94
4ih Q tr '8 3 2nd Qtr '94
Figure 6-23 PV Capacity Factor at Monthly
Peak Load for Peoria, AZ.
B PV Generation O Building Load
Monthly Building Load Factor
and PV Capacity Factor
O
O
-------
(a) Third Qua rte r 1 9 9 3
-
_
"
-
5 10 15 20
Highest Building Load Hours
(c) First Q ua rte r 19 94
212
202
5 10 15 20
Highest Building Load Hours
(e)Third Q ua rte r 19 94
274
254
0,6 n
O
25
i i j »
0 S 10 15 20 25
Highest Building Load Hours
(b ) Fourth Qua rte r 1993
237 r , 1
5 232
222
217
O
>
5 10 15 20 25
Highest Buildrng Load Hours
(d) Second Quarter 1994
290 — , 1
270
260
5 10 15 20
Highest Building Load Hours
(f)Top 25 Hours in Study Period
293 ( , 1
263
~ ~~ %
a
>
a
0 5 10 15 20 25
Highest Building Load Hours
Total Building Load 3 Net Building Load ~ Effective PV Capacity
Figure 6-25 Building Load Duration Curves with and without PV for
Ashwaubenon. WI.
6-26
-------
PV Capacity Factor
at Building Peak Load
100%
0% I 1—
3rd 0 lr '93
1 St Qtr '94 3rd Qf 94
41H Qf'93 2nd Qtr "94
Figure 6-26 PV Capacity Factor at Monthly
Peak Load for Ashvvaubenon, WI.
PV Generation
Building Load
Monthly Building Load Factor
and PV Capacity Factor
o 100%
80%
° 80%
v 40%
20%
3rd Q tr '33 1sfcOtr'S4 3rdQtr'9£
4th Gtr 93 2nd Qtr '94
Figure 6-27 Average Building Load Factor and
PV Capacity Factor for
Ashwaubenon, WI.
Notes on System Operation:
3rd Q '93;
Building load not monitored until 10/L/93. System at 2/3 power 8/9,'93 - 9/9/93,
4th Q '93:
System at 2/3 power 10/1/93 - 10/11/93 due to inverter failure.
1st Q '94:
System fully operational
2nd Q '94;
System fully operational
3rd Q '94:
System operated at 2/3 power 9/12/94 - 9/15/94 due to DC disconnect fuse failure.
6-27
-------
(a)Third Quarter 1993
? 2
5 10 15 20
Highest Building Load Hours
0 .6 «i
O
>
a.
+4-#++ r ~ 1 ,
0 5 10 15 20 25
Highest Building Load Hours
(e ) Third Quarter 1994
o 10
5 10 15 20
Highest Building Load Hours
(b) Fourth Quarter 1993
(d) Second Quarter 1994
Highest Building Load Hours
(f) T o p 25 Hours in Study Period
0 5 10 15 20 25
Highest Building Load Hours
Total Building Load H Net Building Load ~ Effective PV Capacity
Figure 6-28 Building Load Duration Curves with and without PV for
Denmark. WI.
6-28
-------
Figure 6-29 PV Capacity Factor at Monthly
Peak Load for Denmark, WI.
H PV Generation [H Building Load
Monthly Building Load Factor
arid PV Capacity Factor
2 100%
80%
20%
3rd Of'93 1slQtr'S4 araQlr'94
4<«Qtr*93 2nd Qtr'94
Figure 6-30 Average Building Load Factor and
PV Capacity Factor for Denmark.
WI.
Notes on System Operation:
3rd Q '93:
Svstem shut down due to inverter failure 7/12/93 - 7/26/93, 9/7/93 • 9/8/93. Building load not monitored until
11/4/93.
4th Q '93;
Building load monitoring begins 11/4/93.
1st Q '94:
System fully operational
2nd Q '94:
Inverter shut down 5/16/94 - 5/20/94 due to DC injection.
3rd Q '94:
System fully operational
6-29
-------
0 5 10 15 20 35
H ighest Building Load Hours
(c) FirstOuarter 1994
19 1
...
! ¦ "'i!
tj4 J S
* 11 .-<¦««
-yt* ~.
0 5 10 15 20 25
H ighesl BuiJding Load Hours
(e ) T h ird Quarter 1994
> ~ %*~
kV '"*%
>
Q.
0 5 1Q 15 20
Highest Building Load Hours
2S
(b ) Fourth Quarter 1993
0 5 10 15 20 25
Highest Building Load Hours
(d) Second Quarter 1994
Highest Bui'ding Load Hours
(f)Top 25 Hours in Study Period
Highest Building Load Hours
Total Building Load SI Net Building Load ~ Effective PV Capacity '
Figure 6-31 Building Load Duration Curves with and without PV for
Minnetonka. MN.
6-30
-------
PV Capacity Factor
atBuilding Peak Load
100%
3rd Glr 'S3 IslQtr S4 3rd Qtr 34
4tti Q tr '93 2nd Q'i 94
Figure 6-32 PV Capacity Factor at Monthly
Peak Load for Minnetonka. MN.
PV Generation Hj Building Load
Monthly Building Load Factor
and PV Capacity Factor
2 100%
80%
U 60% -
O 40%
20%
3rd G lr '93 tsIQtr'94 3rd Qir'94
4th Qt 'S3 2nd Gir '94
Figure 6-33 Average Building Load Factor and
PV Capacity Factor for
Minnetonka, MN.
Notes on System Operation;
3rd Q '93:
System fully operational
4th Q '93:
System off-line 10/1/93 - 10A2/93 due to inverter failure.
1st Q '94: System fully operational
2nd Q '94:
System fully operational
3rd Q '94:
System fully operational
6-31
-------
(a) First Quarter 1994
356
335
S
CD
326
10 15 20 25
H ighesl Builchng Load Hours
{c) Third Quarter 1994
%
~
-
0 5 10 15 20 25
Highest Building Load Hours
(b) Second Qua rte r 1 9 9 4
^ 373
o 363
>
CL
5 10 15 20
Highest Buildrng Load Hours
(d) Fourth Quarter 1994
343 i , 1
5 ^0 15 20 25
Highest Building Load Hours
|e)Top 25 Hours in Study Period
384
374
en 334
5 10 15 20
Highest Building Load Hours
Total Building Load
Net Buildicg Load ~ Effective PV Capacity
Figure 6-34 Building Load Duration Curves with and without PV for
San Ramon, CA,
6-32
-------
PV Capacity Factor
atfiuilding Peak Load
100%
60%
O 40%
>
a.
20%
1 st Q If '94 3rd Qtr '94
2nd Qtr '94 4th Qtr '94
Figure 6-35. PV Capacity Factor at Monthly
Peak Load for San Ramon, CA.
PV Generation
B Gliding Load
Monthly Building Load Factor
and PV Capacity Factor
o 100%
80%
W 60%
151 Qtr'94 3rd Qtr '94
2naQ»'94 4!nQtr'94
Figure 6-36, Average Building Load Factor
and PV Capacity Factor for San Ramon. CA.
Notes on System Operation:
1st Q "94: PV system dam acquisition commences 1/7/94, Load measurement commences 3/7/94, System shut down 2/11/94 •
2/14/94.
2nd Q '94:
System operated at 2/3 power 5/14/94 - 5/27/94 due to inverter failure.
3rd Q '94:
One string disabled after tasting. System limited to about 80% 7/1/94 - 8/11/94,
4th Q '94:
System operated at 2/3 power 10/19/94 - 10/24/94 due to inverter failure,
6-33
-------
(a) Fourth Quarter 1993
10*8 - i 1
0 5 10 15 20 25
Highest Build mg Load Hours
{c ) Second Q ua rte r 1 994
908 i 1
848
0 5 10 15 20 25
Highest Building Load Hours
{b) First Quarter 1994
o
>
6 10 15 20 25
Highest Building Load Hours
(d) Third Quarter 1994
906 i , 1
5 10 15 20
Highest Building Load Hours
(e) Top 25 Hours in Study Period
Highest Build ing Load H ours
Total Building Load IS Net Building Load ~ Effective PV Capacity-
Figure 6-37 Building Load Duration Curves with and without PV for
Austin, TX,
6-34
-------
PV Capacity Factor
at Building Peak Load
100% ,—
1 1/93 1/94 3,'94 5/94 7/94 9/94
Figure 6-38. PV Capacity Factor at Monthly
Peak Load for Austin, TX.
H PV Generation §2 Building Load
Monthly Building Load Factor
and PV Capacity Factor
g 100%
O
ra
Figure 6-39, Average Building Load Factor
and PV Capacity Factor for Austin, TX.
Notes on System Operation:
4th Q '93:
Properly functioning PV meter not installed until 3/14/94. PV generation prior to that date was simulated using
irradiance data.
1st Q '94:
Properly functioning PV meter not installed until 3/14/94, PV generation prior to that date was simulated using
irradiance data.
2nd Q '94:
System at 1/3 power 5/11/94 - 5/25/94 due to inverter failure, off-line6/ll/94 - 6/14/94 due to fuse failure.
3rd Q "94:
System operated at 2/3 power for total of 12 days due to fuse failure.
6-35
-------
(a) First Quarter 1994
5 10 15 20 25
H tghesi Building Load Hours
(c) Third Quarter 1894
231 , 1 1
O
>
CL
0 5 10 15 20 25
Highest Building Load Hours
(b } Second Quarter 1994
5 10 15 20
Highest Building Load Hours
(d) Fourth Quarter 1994
273
256
5 10 15 20
Highest Bui-ding Load Hours
(e) Top 25 Hours in Study Period
¦" Total Building Load Q Net Building Load ~ Effective PV Capacity
Figure 6-40 Building Load Duration Curves with and without PV for
Flagstaff, AZ.
6-36
-------
PV Capacity Factor
atBuildirig Peak Load
2/94 4/34 6/94 8/94 10/94 12/94
Figure 6-41, PV Capacity Factor at Monthly
Peak Load for Flagstaff, AZ.
| PV Generation §{1 Building Load :
Figure 6-42, Average Building Load Factor
and PV Capacity Factor for Flagstaff, AZ.
Notes on System Operation:
1st Q '94:
System fully operational
2nd Q '94:
System fully operational
3rd Q '94:
System fully operational
4th Q '94:
Inverter manually shut down for four days.
6-37
-------
10 15 20
Highest Building Lead H o u' s
(c) F i rs t Q u a rte r 18 9 5
5 10 15 20 25
Highest Building Load Hours
(b ) Fourth Quarter 1994
o
>
5 10 15 20
H ighesi GyilSing load Hours
(d) Second Quarter 1995
5 10 15 20 25
Highest Building Load Hours
(e ) Top 25 Hours iri Study Period
7 r__ 1
5 10 15 20
Highest Building Load Hours
Total Building Load
Net Building Load ~ Effective PV Capacity
Figure 6-43 Building Load Duration Curves with and without PV for
Barstow, CA.
6-38
-------
Figure 6-44. PV Capacity Factor at Monthly
Peak Load for Barstow, CA.
9 PV Generation j^j Building Load
Figure 6-45. Average Building Load Factor
and PV Capacity Factor for Barstow, CA.
Notes on System Operation:
3rd Q '94:
System shut down 7/22/94 - 8/3/94 due to DC disconnect fuse failure. Shading reduces output by 30 to 40
percent.
4th Q '94:
System fully operational.
1st Q '95:
System fully operational.
2nd Q '95:
Work on local distribution circuit causes inverter to operate intermittently 5/25/95 - 6/10/95.
6-39
-------
(a) First Qua rta r 1994
o
>
5 10 15 20
H ighest BuSiding Load Hours
(c) Third Quarter 1994
5 10 15 20
Highest Building Load Hours
-------
PV Capacity Factor
at Building Peak Load
IrtAW
Figure 6-47. PV Capacity Factor at Monthly
Peak Load for Edwards AFB, CA.
| PV Oeueratian f[| Building Load !
Monthly Building Load Factor
and PV Capacity Factor
100%
>.
2$4 At94 6/94 6/94 10.S4 12/94
Figure 6-48. Average Building Load
Factor and PV Capacity Factor for
Edwards AFB, CA.
Notes on System Operation:
1st Q '94:
Data collection begins 2/1/94.
2nd Q '94:
System fully operational
3rd Q '94:
System fully operational
4th Q '94:
System shut down 12/21/94 - 12/31/94 due to DC disconnect fuse failure.
6-41
-------
(a) First Quarter 1994
5 10 15 20
Highest Bui;ding Load Hours
(c) Third Quarter 1994
0 5 10 15 20 25
Highest Building Load Hours
(b) Second Quarter 1994
5 10 15 20
H ighest Building Load Hours
(d ) Fourth Quarter 1994
82 i 1
5 10 15 20
Highesl Building Load Hours
(e)Top 25 Hours in Study Period
0 5 10 15 20 25
H ighest Buifding Load Hours
I ¦¦¦¦ Total Building Load 0 Net Building Load ~ Effective PV Capacity
Figure 6-49 Building Load Duration Curves with and without PV for
Palm Desert, CA.
6-42
-------
Figure 6-50. PV Capacity Factor at Monthly
Peak Load for Palm Desert, CA.
| PV Generation Building Load
Figure 6-51. Average Building Load Factor
and PV Capacity Factor for Palm Desert, CA.
Notes on System Operation:
1st Q '94; Dan acquisition commences 2/1/94,
2nd Q '94: System fully operational
3rd Q '94: System fully operational
3rd Q '94: System fully operational
6-43
-------
Chapter 7
Utility Coincident Peak Load Reduction
From an electric utility's perspective, much of the potential value of a PV system resides in its ability to
provide power at the time when it is most needed— i.e., during the utility's peak load hours or during
peak loads for particular portions of the transmission and distribution system. This section describes
each PV system's track record of power generation during the highest load hours faced by each
respective utility. Results for each system are presented as a set of graphs similar to the one in Figure 7-
1, which depicts the utility load duration curve for a quarter and the corresponding "cumulative average
capacity factor curve" (CACF curve) for the PV system in each hour. Results are presented both for the
PV system as it actually operated and for the simulated operation of the system with dispatchable
storage, as described in Chapter 5,
Third Quarter 1994
The CACF curve describes, for any point on the load duration curve, the average capacity factor of the
PV system over all hours up to and including that point (i.e. the "running average" of the PV system's
capacity factor up to that point). The CACF curves were constructed by first sorting utility load values
and PV generation values jointly, with utility load determining the sort order. This resulted in a ranking
of utility load values in descending order with a PV generation value corresponding to each utility load
value. For each hour, the PV system's capacity factor was then determined by dividing its hourly output
by its rated capacity. Hourly capacity factors for the systems' actual operation were calculated by
dividing hourly kWh generation by each system's
respective SOC rating (the system's expected hourly
kWh generation under conditions of 1,000 Watts/m2
irradiance and 25° C ambient temperature). For the
dispatchable storage simulation, hourly capacity factors
were calculated using inverter nameplate ratings in the
denominator, since the model simulated inverter
operation at its nameplate rating. Finally, cumulative
average capacity factors were calculated by averaging
each hourly capacity factor with the capacity factors of
all hours for which utility load was at a higher level.
The resulting values (the CACF curve) indicate the
system's average capacity factor for the highest n load
values on the utility LDC. Again, the charts in this
section contain curves of the cumulative average
capacity factor for each hour, rather than the hourly
capacity' factors themselves.
'¦a D.1
3
100
Load-Hours
1,000
10,000
In Figure 7-1, the grey line indicates the utility load
duration curve, which has been normalized by the
Figure 7-1. Example Utility Load Duration
Curve with Cumulative Average Capacity
Factors.
7-1
-------
highest load value encountered during the study period. Any point along the load duration curve
therefore represents per-unit system load. At any given point on the LDC, all points to the left of that
point are at equal or greater load levels; points to the right are at equal or lesser load levels. The black
line in the figure represents the cumulative average capacity factor of the PV system as calculated from
measured performance data, and the shaded area represents cumulative average capacity factor based on
the simulated performance of the system with dispatchable storage. For any load level on the LDC, the
cumulative average capacity factor up to and including that hour is indicated by the corresponding point
on the black line (for actual system operation) or at the top of the shaded area (for the simulated
dispatchable storage scenario). Note that in order to provide greater resolution for the highest load hours
in each quarter, a logarithmic scale has been used on the abscissa.
In the example in Figure 7-1. the tenth-highest load hour in the quarter was about 87 percent of the
highest load recorded during the study period. The average capacity factor of the PV system during the
ten highest load hours for the quarter was about 55 percent without storage, and the storage simulation
resulted in an average capacity factor of almost 70 percent over the ten highest-load hours. Note that for
the lowest load hour of each quarter, the average capacity factor for the battery simulation will always be
about 66 percent of the average capacity factor based on measured system operation. This is due to the
fact that the simulation assumes that 25 percent of the energy converted by the PV system will be lost in
charging and discharging the battery, and because capacity factors for the simulation were calculated
using the inverter's nameplate capacity, which are greater than the SOC system ratings used to calculate
hourly capacity factors for measured system operation/
In addition to describing the operation of each system during utility peak load hours (with and without
storage), each of the sections below provides a brief description of the battery bank necessary to provide
storage capacity sufficient for the greatest daily kWh generation of the PV system. In each case, the size
of the battery bank is determined as follows: The greatest daily energy generation is divided by the
nominal operating voltage of the battery, taken to be 14 V, to determine the equivalent number of amp-
hours required. For this analysis, batteries with a storage capacity of 115 amp-hours and capable of
sustaining an occasional 50 percent discharge are assumed.4 The number of batteries is therefore
determined by dividing the amp-hour requirement by 115 and multiplying by two. As a final step, the
number of batteries is divided by the inverter rating, to determine the size of the barter) bank required
per kW of inverter rating.
PV system outages make the charts below somewhat more difficult to interpret. While the figures
represent actual system operation, many of the outages were permanently resolved once their causes
were understood. For this reason, charts for quarters in which there was a significant outage may not be
representative of system behavior over the long term. A good example of this is the set of system
outages caused by failures of the fuses originally used in the DC disconnect switches. A total of 17 such
'For a system with a 4kW inverter rating, the SOC rating is 3.5 kW. Therefore, taking battery charging and discharging losses into account and
using the larger inverter rating to determine hourly capacity factors, the average capacity factor for a quarter will be lower under the dispatch
model than that measured by a factor of
[(1 - 0.25)*3.5/4,0] = 0.66.
'These battery characteristics are consistent with those of batteries Ascension Technology has used in other PV projects that have storage
capacity. In an application such as that modeled here, other considerations, such as inverter voltage requirements would also play a
determining role in the design of the battery storage system.
7-2
-------
fuse failures were recorded, and they were replaced as they occurred by fuses that were better suited to
the application.
7.1 Pittsburgh, NY (EPA01).
Figure 7-2 illustrates the utility system level results for New York State Electric and Gas (NYSEG). The
figure indicates that although the PV system in Plattsburgh achieved relatively high average capacity
factors in spring and summer months (despite the outages noted below the figure), average capacity
factors were extremely lowr during the fourth quarter of 1993 and first quarter of 1994, when NYSEG's
highest loads occurred. The final chart in the figure shows that the PV system provided no power at all
during the 100 highest load hours encountered during the study period.
The addition of dispatchable storage raises average capacity factors during the highest load hours of all
quarters, particularly the third quarter of 1993 and second quarter of 1994, when the average capacity
factors are near or at 1.0 for the top 3 load hours of each quarter. Even with dispatchable storage though,
the cumulative average capacity factor never rises above 5 percent in the top 10 hours of the fourth
quarter of 1993. The storage system does raise the capacity factor above 45 percent for the highest load
hour in the first quarter of 1994. but the average declines steadily thereafter until hour 10, indicating
several hours with very low levels of generation. The maximum daily kWh generation by this system
was 81 kWh. which is equivalent to about 5,800 amp-hours of battery storage. This amount of storage
could be reliably provided by a bank of 100 batteries, or by 9 batteries per kW of inverter rating.
The figure indicates clearly that because NYSEG is a winter-peaking utility in a region with limited
winter solar resources, PV has little capability to reduce peak system load either with or without energy
storage. PV systems in the NYSEG service area can provide power at about 40 percent of their SOC
rating during summer peak load hours, however.
7.2 Berlin, CT (EPA02).
Although the PV system in Berlin, CT is in relatively close proximity to the Plattsburgh system (less than
200 miles separate the two), the Berlin system performed markedly better during utility peak load hours,
primarily because many of the highest load hours during the study period occurred during the summer
months. Figure 7-3 indicates that even without storage, the average PV system capacity factor remained
close to 90 percent for the highest 20 load hours in the third quarter of 1993. As with Plattsburgh, the
cumulative average capacity factors are dramatically lower in the winter months, at or near zero for all of
the highest load hours during the fourth quarter of 1993 and the first quarter of 1994.
The addition of dispatchable storage to the system improves the CACF curves throughout the study
period, most notably in the fourth quarter of 1993. Without storage, the PV system's capacity factor was
zero for the 10 highest load hours of that quarter, all of which occurred after sunset. Because the
dispatch simulation did serve load during these hours, the average capacity factor was about 55 percent
for the highest ten hours of the quarter. As the figure indicates, the addition of storage raised the average
capacity factor to 1.0 for the highest load hours in the spring and summer months. Although
dispatchable storage does provide a moderate improvement to the average capacity factor curve in the
first quarter of 1994, chart (c) indicates that the PV system did not generate at all on the day(s) on which
the three highest load hours for the quarter occurred.
7-3
-------
Chart (f) indicates that dispatchable storage provides only a moderate improvement in this system's
ability to match utility load during the highest 100 load hours. Although the addition of storage
improves the system's capacity factor from about 90 percent to 100 percent for the highest load hour
during the study period, many of the subsequent hours on the load duration curve occurred during the
first quarter of 1994, when the PV system's output was at a minimum. Because the storage system was
unable to substantially improve the system's performance during peak load hours in that quarter, the
average capacity factor curve for the storage simulation case shows the same initial decline as the CACF
curve for measured system operation.
Taken together, the charts in Figure 7-3 suggest that PV can provide excellent peak load matching for
Northeast Utilities during the summer months, but should not be relied upon to serve peak loads during
winter peak hours. The charts also suggest that, if charged and discharged in a manner similar to the
procedure used for this study, dispatchable storage would provide a rather modest improvement to this
system's ability to generate during peak load hours. It should be noted however, that the storage model
discharged the battery completely every day. If a battery were sized to store more than one day's output
from the PV system and dispatched only when load reached a certain kW threshold, storage might
greatly improve this system's ability to deliver power during utility winter peak load hours.
The highest daily generation for this 4 kW (nominal) system was 33 kWh. This amount of storage could
be provided reliably by an array of 40 batteries, or 10 batteries per kW of inverter rating.
7.3 Pleasantville and Brigantine, NJ (EPA03 and EPA04).
These systems are separated by approximately 10 miles, and are both in the Atlantic City Electric service
area. One would therefore expect that the CACF curves would be essentially identical. A quarter-by-
quarter comparison of the charts in Figure 7-4 and Figure 7-5 shows that the output of these systems
during peak ACE load hours was quite similar in some quarters (e.g. the second and third quarters of
1994), but very different in others (e.g. the third quarter of 1993). The reason for the difference in the
CACF curves is that both systems suffered a variety of outages at different times over the course of the
study period, in addition, because the system in Pleasantville consists of three 3.6 kWsuc subsystems, an
outage of any single inverter would leave two-thirds of the system operating. The system in Brigantine
consists of only one array and one inverter, thus an outage here shut the entire system down, resulting in
hourly capacity factors of zero.
Chart (a) of Figure 7-4 indicates that without storage, the average capacity curve for the system in
Pleasantville stays in the mid to upper teens throughout the third quarter of 1993. The system's output
that quarter was severely hampered by inverter failures for a total of 48 days. Longer-term system
performance in the third quarter may therefore be better indicated by chart (e) in Figure 7-4 or charts (a)
or (e) in Figure 7-5. These charts suggest that for a fully operational system, average capacity factors
during peak load hours in the third quarter would typically be between 40 and 50 percent.
The system in Pleasantville operated at full power throughout the fourth quarter of 1993, and with the
exception of four days, the Brigantine system did as well. Chart (b) in Figure 7-4 and Figure 7-5 are
therefore quite similar, and indicate that without storage, these systems provided no generation during
the highest load hours of this quarter. Results for the simulation indicate that dispatchable storage could
improve the capacity factor of these systems during high load hours, but that typically one could not
expect operation in excess of 20 percent of inverter rating during this quarter.
7-4
-------
The measured results for the first quarter of 1994 are similar to those for the previous quarter. Both
systems had outages which affected their operation during this quarter. Inspection of chart (c) in Figure
7-5, however, indicates that the highest load hours for this quarter occurred after sunset. The fact that the
CACF curve for the storage simulation shows very high average CFs for the highest load hours in this
chart indicates that the system was in fact operating on the days on which the ten highest load hours
occurred. The PV system outages were therefore not responsible for the fact that the PV system's
measured capacity factor was at or near zero for each of the ten highest load hours in the quarter. Rather,
this was due to the fact that these loads occurred at night. The simulated results for this quarter indicate
that battery storage would substantially improve the system's ability to generate during peak utility
loads.
The highest load hours during the second quarter of 1994 occurred during June when both PV systems
were fully operational. The CACF curves for this quarter indicate average capacity factors between 40
and 50 percent for the highest load hours. Energy storage raises the CACF curves substantially for both
systems. In Brigantine. the CACF curve remains at 100 percent for the seven highest load hours of the
quarter. Simulation results for Pleasantville show the CACF curve dropping from 100 percent for the
two highest load hours to an average of 67 percent for the 10 highest load hours in the quarter.
Although an outage in late July 1994 reduced the power output of the Pleasantville system for about two
weeks, the outage occurred during a period of moderate load. The third quarter 1994 charts are therefore
almost identical for these two systems, both with and without storage. Chart (f) in Figure 7-4 and Figure
7-5 present the CACF curves for the 100 highest load hours encountered during the study period, all of
which occurred during the third quarter of 1993. These charts indicate that PV system outages limited
the Pleasantvil le system's performance to an average of about 15 percent of its SOC rating for the
highest load hours, and about twice that for the Brigantine system. For both of these systems, the charts
indicate that energy storage more than doubles their ability to provide power during utility peak load
hours.
The data illustrated in the figures indicate that PV systems interconnected to the Atlantic City Electric
power system can be expected to operate at 40 to 50 percent of their SOC rating during the utility's
highest load hours, which occur during the summer months. With battery storage, power output could be
increased to 100 percent of inverter rating during utility peak load hours. The maximum daily generation
of these systems was 81 kWh and 29 kWh for the Pleasantville and Brigantine systems respectively.
Both systems would require banks of 9 batteries per kW of inverter rating to supply this amount of
storage.
7.4 White Plains, NY (EPA05).
The system in White Plains shows characteristics similar to those for the systems discussed above.
CACF curves based on measured results indicate that this system's capacity factor would typically be
between 30 and 40 percent of its SOC rating during the utility's highest load hours in the summer
months, and at or near zero for the highest load hours in the winter. A fuse failure shut the system down
for 20 days in the second quarter of 1994, (when the quarter's highest loads occurred), resulting in the
very low CACF curves in chart (d) of Figure 7-6. While this chart does reflect actual system operation,
it should be noted that this outage resulted from an initial system design flaw which was subsequently
7-5
-------
rectified. For this reason, chart (d) is probably not very representative of typical system performance in
the second quarter.
Because most of the highest load hours during the study period occurred in the third quarter of 1993,
chart (f) is quite similar to chart (a) (with the exception of the change in abscissa scaling). This chart
indicates that for NYPA's highest load hours, PV systems could typically be expected to provide power
at between 30 and 40 percent of their SOC ratings.
Unlike the system in Berlin, CT, adispatchable storage system would greatly improve this system's
ability to provide power during peak utility load hours. As charts (a), (e) and (f) of Figure 7-6 indicate,
the storage simulation more than doubled the system's average capacity factors during peak hours.
Battery storage is more effective in improving system performance during peak hours for NYPA's
service territory because peak loads there tend to occur in the morning or late afternoon, when the PV
system operates well below its peak rating. Daily peaks on the Northeast Utilities system (which the
Berlin system is interconnected to) tend to occur at mid-day, there is therefore a natural correlation
between system load and the availability of PV power. Because of this correlation, the addition of
dispatchable battery storage can do little to improve the peak shaving capability of the PV system.
The peak daily energy output for this system was 30 kWh. An array of 38 batteries, or 10 batteries per
k W of inverter rating would be necessary to provide this amount of storage reliably.
7.5 Scottsdale, Peoria, and Flagstaff, AZ (EPA06, EPA07, and EPA13)
The ability of these three systems to provide power during Arizona Public Sendee peak load hours is
illustrated by Figure 7-7 through Figure 7-9. Although they are interconnected to the same utility, there
are appreciable differences between the average capacity factor curves for each system. In Scottsdale,
the charts in Figure 7-7 indicate that without storage this system typically operated at a capacity factor in
excess of 50 percent during the highest load hours during the summer months. The system's
performance dropped to an average 40 percent capacity factor for the ten highest load hours in the fourth
quarter of 1993 and dropped precipitously to an average of only about 3 percent for the top ten hours in
the first quarter of 1994. A review of the average monthly capacity factors for this system (illustrated in
Figure 6-21 on page 6-23) shows that the energy generated by this system was relatively stable
throughout the year. This suggests that the seasonal fluctuation in the system's ability to provide power
during peak load hours has more to do with the timing of peak hours than with seasonal variation in the
solar resource. This implication is borne out by the system load data, which during the summer months
tends to peak in the mid-afternoon, and during the winter peaks in the early morning or after sunset.
Because utility peak loads rarely occurred during the middle of the day (when the system was operating
at or near its rated capacity), the addition of dispatchable storage has a pronounced affect on the system's
load matching capability in all seasons. This is most evident in chart (c) of Figure 7-7, which shows that
the simulation operated the inverter essentially at its nameplate rating in each of the 20 or so highest load
hours in the first quarter of 1994, despite the fact that measured system operation during those hours was
close to zero.
Results for the system in Peoria are very similar to those for Scottsdale, as one would expect given their
proximity. The average capacity factor curves for the Peoria system however, are consistently lower
7-6
-------
than those for Scottsdale. This is due to the fact that two-thirds of the Peoria array were installed at a
lower-than-optimal tilt angle. The low tilt angle reduced plane-of-array irradiance, and allowed dirt to
build up on the modules, further reducing their output. Note also that data collection for this system
began on September 1, 1993 so chart (a) in Figure 7-9 contains only one month of data.
The Flagstaff system began operating in February 1994. Figure 7-9 does not, therefore, contain charts for
the last two quarters of 1993. The capacity factor curves for the first two quarters of 1994 are very
similar to those for the Scottsdale and Peoria systems. The measured results for the third quarter of 1994
however, are substantially lower during the highest load hours for the Flagstaff system than for the other
two systems. The minimal PV output during peak load hours for this quarter is a result of frequent
afternoon cloud cover in the mountains surrounding Flagstaff. As it did for the other Arizona systems,
the addition of storage to this system increased its load matching capability dramatically.
The highest daily generation achieved by the Scottsdale system (nominally 8 kW) was 54 kWh. The
corresponding numbers for Peoria and Flagstaff (both nominally 4 kW systems) were 24 kWh and 30
kWh respectively. Banks of 9. 8, and 10 batteries per kW of inverter rating would be necessary to
provide this degree of storage capacity for Scottsdale, Peoria, and Flagstaff respectively. The difference
between Scottsdale and Peoria is due to the lower tilt angle used in Peoria as described above. The
reason for the greater storage requirement for the Flagstaff system is more subtle. Although average
daily generation per kW by the Scottsdale system was higher than that for the Flagstaff system (5.1 kWh
per kW-day versus 4.7 kWh per kVV'-day). peak daily generation was higher in Flagstaff than in
Scottsdale (8.4 kWh per kW-day vs. 7.5 kWh per kW-day). This may be due to the fact that the
Flagstaff system was not exposed to the very high ambient temperatures that reduced the output of the
other two systems.
7.6 Ashwaubenon and Denmark, WI (EPA08 and EPA09).
Both measured and simulated results for these two sites are quite similar, as shown by Figure 7-10 and
Figure 7-11. Charts (a) and (e) of these figures indicate that without storage, these systems can typically
be expected to generate at between 40 and 50 percent of their rated SOC capacities during the third
quarter of the year (when many of the highest load hours occurred), although an inverter outage resulted
in lower average capacity factors in the third quarter of 1993 for the Ashwaubenon system. The power
output of these systems was at or near zero for the highest load hours in the fourth quarter of 1993 and
the first quarter of 1994, when peak load hours almost invariably occurred after sunset. In the second
quarter of 1994, the average capacity factor curves of both PV systems exceed 70 percent for the ten
highest load hours.
The charts in Figure 7-10 and Figure 7-11 show that dispatchable storage raises average system capacity
factors during peak hours in all four calendar quarters, although the effect of storage varies tremendously
from quarter to quarter. The effect of storage is most pronounced during the fourth quarter of 1993 and
the first quarter of 1994, when most peak loads occurred either before or after sunset. For the Denmark
system, the average capacity factor resulting from the storage simulation for the ten highest load hours in
these quarters were approximately 60 percent and 87 percent respectively. In Ashwaubenon,
dispatchable storage had a much smaller affect on the system's load matching capability, averaging just
43 percent and 10 percent respectively. The difference between the two systems is accounted for by the
7-7
-------
fact that the PV array in Denmark was installed at a steeper tilt angle, allowing snow cover to melt and
slide off much more quickly than was the case for the Ashwaubenon array.
In the third quarter of 1993, storage roughly doubles average capacity factors during the highest load
hours. Peak loads in this quarter typically occurred in the mid- to late-afternoon hours, when the PV
system was operating well below its rated capacity. In the second quarter of 1994, peak loads tended to
occur in the early afternoon, thus average capacity factors were higher than during the third quarter, and
the effect of storage less appreciable.
All but four of the 100 highest load hours recorded during the study period occurred in the summer
months. Chart (f) in both Figure 7-10 and Figure 7-11 indicates average measured capacity factors for
both systems in excess of 60 percent of SOC ratings for the ten highest load hours. These charts indicate
that dispatchable storage will improve system performance during the highest 10 or 20 load hours, but
that its effect beyond those hours is minimal or even negative (due to the fact that energy is lost in
charging and discharging the battery). Maximum daily generation for these systems was 92 kWh and 30
kWh for Ashwaubenon and Denmark respectively. Both systems would require 10 batteries per kW of
inverter rating to supply this amount of storage.
7.7 Minnetonka, MN (EPA10)
The ability of this system to supply power during peak load hours is similar to that of the other systems
installed in northern climates. Without storage, peak matching is quite good in the summer months when
Northern States Power annual peak occurs, and poor in the winter. Charts (a), (d), and (e) in Figure 7-12
indicate average capacity factors between 60 and 70 percent of the system's SOC rating for the highest
load hours in the second and third quarters. The system provided no power at all during the highest load
hours of the fourth quarter of 1993 and the first quarter of 1994, because peak loads during these quarters
occurred after sunset.
Dispatchable storage increases average capacity factors to 100 percent of inverter rating for the highest
load hours during the summer months. As one proceeds to the right along the load duration curves in the
figure, however, the value of storage diminishes quickly. Because the dispatch algorithm concentrated
all generation into a few daily peak load hours, little energy was left to serve load during hours with
average load levels below about 90 percent of the peak load hour. Although the average capacity factor
curves for the storage simulation drop off even more rapidly in the fourth quarter of 1993 and the first
quarter of 1994, storage improves load matching capability dramatically over that available from the PV
system alone. Again, this is because storage allows energy to be transferred to the highest load hours,
which occurred after dark during these months.
The 100 highest load hours occurred in the second quarter of 1994 and the third quarters of 1993 and
1994. Chart (f) in Figure 7-12 therefore resembles the charts for these quarters (other than the change in
the abscissa scale). This chart suggests that PV is well suited to the Northern States Power service area
from a system load matching perspective. The PV system's average capacity factor for NSP's ten
highest load hours during the study period was just over 61 percent of its SOC rating. The corresponding
average over the top 100 hours was just over 51 percent. A storage system operated as described for this
study could raise these averages to 100 percent and 62 percent for the top 10 and top 100 load hours
respectively. The peak daily generation during the study period for this system was just over 30 kWh. A
7-8
-------
bank of 10 batteries per kW of inverter rating would be necessary to supply this amount of storage
reliably.
7.8 San Ramon, CA (EPA 11)
Data acquisition for the PV system in San Ramon began in early January 1994. This system's ability- to
provide power during Pacific Gas and Electric peak load hours is illustrated in Figure 7-33. Peak loads
in the first and fourth quarters of that year consistently occurred in the evening after sunset. The capacity
factors for the 100 highest load hours during both quarters were therefore zero. Measured results were
considerably better in the second and third quarters of 1994, with average capacity factors between 51
and 61 percent for the 10 highest load hours in the second quarter, and between 46 and 53 percent for the
10 highest load hours in the third quarter.
Because PG&E is a summer peaking utility, the chart for the 100 highest load hours is quite similar to
those for the second and third quarters of 1994. The average capacity factor curve remains between 44
percent and 56 percent of inverter rating throughout the 100 highest load hours. Results of the storage
simulation indicate that the average capacity factor curve never dips below 80 percent during this period.
Battery storage is very effective in improving this PV system's average capacity factors during peak load
hours. Results of the simulation indicate that because of the ample solar resource in the winter months,
the average capacity factor for the ten highest load hours in the first quarter of 1994 is 97 percent of the
inverter rating. The corresponding figure for the second quarter of 1994 exceeds 84 percent, and for the
third quarter, the average capacity factor is 100 percent for the ten highest load hours. The peak daily
generation by this system was 84 kWh, A bank of nine batteries per kW of inverter rating could provide
this storage capacity.
7.9 Austin, TX (EPA12)
Utility load matching results for the PV system in Austin are illustrated in Figure 7-14. Because meter
malfunctions prevented accurate measurement of PV generation until mid March 1994, generation prior
to that date has been simulated based on irradiance data. These data indicate very poor load matching in
the fourth quarter of 1993 and the first quarter of 1994. Most of the peak load hours in these quarters
occurred either before sunrise or after sunset. This system's ability to provide power during Austin's
spring and summer peak load hours is substantially greater, with the average capacity factor curves
varying between 50 and 60 percent of inverter rating during the ten highest load hours of each quarter.
Because the Austin utility's highest loads occurred during the second and third quarters, the chart for the
top 100 hours is quite similar to charts (c) and (d) in the figure. Over the 100 highest load hours, the
average capacity factor for the PV system was 50 percent.
The results of the battery storage simulation for this system indicate that storage is quite effective in
making PV power available during peak load hours in the spring and summer months, but less effective
the rest of the year. The average simulated capacity factor for the ten highest load hours in the fourth
quarter of 1993 was 44 percent. This figure drops to 27 percent in the first quarter of 1994, but rises to
95 percent and 100 percent for the second and third quarters of 1994 respectively. For the 100 highest
load hours during the study period, the battery storage simulation resulted in an average capacity factor
7-9
-------
of 87 percent. The peak daily generation by this system was 76 kWh, This amount of storage capacity
could be provided by 8 batteries per kW of inverter rating.
7.10 Barstow, Edwards AFB, and Palm Desert, CA (EPA14, EPA15, and EPA16).
The ability of these systems to supply power during Southern California Edison's peak load hours is
illustrated by Figure 7-15 through Figure 7-17. These systems were among the last to be installed in the
project. The charts therefore begin with the first quarter of 1994 for the Edwards AFB and Palm Desert
systems and the third quarter of 1994 for the Barstow system. At Edwards Air Force Base the average
capacity factor curve drops sharply from 75 percent for the highest load hour to an average of 45 percent
for the ten highest load hours in the first quarter of 1994. The first quarter chart for the Palm Desert
system appears quite different from that for Edwards AFB. This is due primarily to a one-day data gap at
Palm Desert that happened to occur on the day with the highest SCE loads in the quarter. For the ten
highest load hours in the quarter, this system's capacity factor averaged 22 percent.
The average capacity factor curves for the third and fourth quarters of 1994 are more consistent from site
to site, differing primarily in magnitude. The system at Edwards AFB exhibited the best peak load
correspondence in the third quarter, with an average capacity factor in excess of 60 percent for the ten
highest load hours. System output was limited at the Barstow site by afternoon shading, resulting in an
average capacity factor just below 40 percent for the top ten hours. The equivalent average for the Palm
Desert system in the third quarter was 47 percent, despite the fact that the array here was configured in
thirds, one-third each facing east, south, and west. In the final quarter of 1994, the average capacity
factor for all three systems was very close to 50 percent for the ten highest load hours.
Charts (c) and (d) in Figure 7-15 indicate that the Barstow system's ability to produce power during peak
utility load hours was reduced considerably in the first part of 1995. In the first quarter of the year, all of
the peak load hours occurred well after sunset. In the second quarter, most of the highest loads occurred
in the mid-afternoon, when shading limited this system's output to less than 20 percent of expected
generation,
The figures illustrate that battery storage could boost the performance of the Edwards AFB and Palm
Desert systems to 100 percent of inverter rating for the highest load hours in each quarter. Although
storage improves the Barstow system's load matching capability in the latter half of 1994, its effect is
substantially reduced at this site for the first two quarters of 1995. Here again, shading at this site
severely limited system generation in the afternoon, and thus limited the amount of stored energy
available for dispatch by the simulation.
The maximum battery capacity necessary to store the daily generation of these systems was 21 kWh. 28
kWh, and 77 kWh for the Barstow, Edwards AFB, and Palm Desert systems respectively. These levels
of storage capacity could be provided reliably by banks of 7, 9, and 8 (again respectively) batteries per
kW of inverter rating. The differences between the required storage capacities are again due to the
shading of the Barstow array and the orientation of the Palm Desert system.
7-10
-------
7.11 Conclusions
7.11.1 Load Reduction Without Storage
Not surprisingly, load matching for PV systems installed in northern States is greatest in the spring and
summer months, with the capacity factor during the highest load hours typically averaging above 40
percent. Several of these sites achieved capacity factors well in excess of 60 percent of their SOC rating
during the highest load hours in these months. The northern systems invariably generated little or no
power during winter peak hours, most or all of which occurred at night.
Utility peak loads in the southern and western parts of the country invariably occurred during the
summer months when the solar resource is greatest, although these peaks consistently occurred in the
mid- to late-afternoon. Most of the systems installed in these regions operated at capacity factors in
excess of 40 percent during the highest load hours in the summer months. Some systems consistently
operated at capacity factors above 60 percent during these hours. The one exception to this is the system
in Flagstaff, which operated at only about 30 percent capacity factor during the peak load hour. This low
result is most likely explained by the fact that the load and weather patterns in Flagstaff are quite
different from those in Phoenix which is about one mile lower in elevation. Loads in the Phoenix area
probably dominate the Arizona Public Service system load.
As did their counterparts in the Midwest and Northeast, systems in the southern and western states
typically operated at a lower level during winter peak hours. With the exception of the systems in
southern California, systems in the West operated at or near zero percent capacity factor during peak
hours in the first quarter of the year.
7.11.2 Load Reduction With Storage
Except where the power output was limited by a system outage, results from the storage simulation
indicate that storage can provide system operation at the full inverter rating during the peak load hours in
the summer months at all sites. In regions (such as the Northeast Utilities service area) where peak
utility loads are highly correlated to the solar resource, the addition of a dispatchable storage system may
do little to improve the PV system's load matching capability, since it will already be quite good.
Systems in northern states are much less able to provide power during peak load hours in winter due to
the limited solar resource and snow cover. Even with storage, some of these systems were unable to
provide power at more than a few percent of inverter rating during winter peak load hours. However,
daytime generation at other northern sites was sufficient to allow inverter operation well in excess of 50
percent of inverter rating during winter peak hours.
Unlike many of the systems installed in northern climates, the addition of dispatchable storage to
systems installed in the southern and western states would allow them to operate at high capacity factors
during winter peak hours. The results of the simulation indicate that most of the systems installed in this
part of the country would operate at or near 100 percent of inverter rating during the highest winter load
hours.
It is important to recognize that these results are substantially determined by the storage
charging/dispatch algorithm. An algorithm which stores generation from one or more days and
dispatches only when load exceeds a predetermined threshold (as opposed to dispatching during the peak
hours of each day), would substantially improve the load matching characteristics of all systems.
7-11
-------
(a) Third Quarter 1993
10 100
Hours
1,000
10,000
(b) Fourth Quarter 1993
k.
u
" 0.9
£
a 06
CO
3 07
^ 0.6
-O
C 0.5
cb
E °4
1 0.3
a
g 0.2
cu
S,0 1
cn
0
10 100 1,000
Hours
10.000
(c) First Quarter 1994
0
CO
CL
(3
£
1
I
T>
SO
O
1
0.9
0.8
0.7
0.8
0.5
0,4
0.3
0.2
0.1
0
\
insx
;;:«C
1
-
1
10 100 1,000
HoufS
10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-2. Utility Load and Cumulative Average PV Capacity
Factor for Pittsburgh, NY.
7-12
-------
(e) Third Quarter 1994
10 100
Hours
1,000 10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-2, (continued) Utility Load and Cumulative Average PV
Capacity Factor for Plattsburgh, NY.
Notes on System Operation:
3rd Q '93: Inverter failure limits system to 2/3 power 7/1/93 - 7/16/93. System shut down completely 7/23/93 - 9/30/93.
4th Q '93: Data loss 10/1/93 - 10/5/93 due to datalogger short-circuit.
1st Q '94: System fully operational
2nd Q '94: DC injection limited system to 2/3 power 4/1/94 - 4/25/94. DC fuse failures limited system to 2/3 power 6/5/94 - 6/15/94 and
6/18/94-6/30/94.
3rd Q '94: System fully operational
7-13
-------
(d) Second Quarter 1994
1 10 100 1,000 10,000
Hours
Utility Load Average PV CF ™ Average PV CF
(per unit) without Battery ™ with Battery
Figure 7-3. Utility Load and Cumulative Average PV Capacity
Factor for Berlin, CT.
7-14
-------
(f) Top 100 Hours
1 00
Hours
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-3, (continued) Utility Load and Cumulative Average PV
Capacity Factor for Berlin, CT.
Notes on System Operation:
3rd Q '93; inverter failure prevented system operation 7/17/93 ¦ 8/6/93.
4th Q *93: System fully operational
lstQ'94: System fully operational
2nd Q '94: System fully operational
3rd Q '94: System ftilly operational
7-15
-------
(a) Third Quarter 1993
1 ,000
10,000
(b) Fourth Quarter 1993
1,000
10,000
(d) Second Quarter 1994
100
Hours
1,000
10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-4. Utility Load and Cumulative Average PV Capacity
Factor for Pleasantville, NJ.
7-16
-------
(e) Third Quarter 1994
1 10 100 1,000 10,000
Hours
Utility Load Average PV CF ™ Average PV CF
(per unit) m"m without Battery ® with Battery
Figure 7-4. (continued) Utility Load and Cumulative Average PV
Capacity Factor for Pleasantville, NJ.
Notes on System Operation:
3rd 0 '93:
Inverter failure limited system to 1/3 power for 43 days, 2/3 power for 5 days.
4th Q '93-
System fully operational
1st 0*94:
Inverter failure limited system to 2/3 power 1/18/94 - 3/11/94 and 3/18/94 - 3/31/94.
2nd 0*94:
Inverter failure limited system to 2/3 power 4/1/94 - 4/22/94,
3rd Q*94:
DC disconnect fuse failure limited system to 2/3 power from 7/23/94 to 8/2/94.
7-17
-------
(a) Third Quarter 1993
1 10 100 1,000 10,000
Hours
(b) Fourth Quarter 1993
O
2 °-9
Houre
(d) Second Quarter 1994
1
O
g 0.9
LL
& 0.8
u
a 0.7
> o®
Q-
T3 0-S
Z0A
3
& 0.3
T3
ra . „
o 0.2
1 10 100 1,000 10,000
Houis
_ Utility Load Average PV CF — Average PV CF
(per unit) mm without Battery with Battery
Figure 7-5. Utility Load and Cumulative Average PV Capacity
Factor for Brigantine, NJ.
7-18
-------
(e) Third Quarter 1994
1 10 100 1,000 10,000
Hours
[ Utility Load Average PV CF ^ Average PV CF
| (per unit) without Battery with Battery
Figure 7-5. (continued) Utility Load and Cumulative Average
PV Capacity Factor for Brigantine, NJ.
Notes oil System Operation:
3rd Q '93; DC disconnect fuse failure prevented generation 8/18/93 - 8/20/94 and again 9/15/94 - 9/20/94.
4th Q '93: Inverter failure prevented generation 12/6/93 - 12/9/93-
Ist Q '94: System shut down 2/25/94 - 2/28/94. DC injection error prevents generation 3/5/94 ¦ 3/17/94.
2nd Q '94: Building load meter recorded zero load 4/15/94 - 6/18/94.
13rd Q '94: System fully operational
7-19
-------
(a) Third Quarter 1993
10 100 1,000
Hours
10,000
(c) First Quarter 1994
100
Hours
1,000
10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-6. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for White Plains, NY.
7-20
-------
Utility Load Average PV CF ^ Average PV CF
(per unit) mm without Battery £3 wjth Battery
Figure 7-6. (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for White Plains. NY.
Notes on System Operation:
3rd Q '93: Inverter failure prevented generation 7/22/94 - 8/13/94.
4th Q'93: System fully operational
ist Q '94: System fully operational
2nd Q '94~ DC disconnect ftise failure prevented generation 6/11/94 - 6/30/94.
3rd Q '94. System fully operational
7-21
-------
(b) Fourth Quarter 1993
1 10 100 1,000 1QS000
Houre
(c) First Quarter 1994
1 10 100 1,000 10,000
Hours
Utility Load Average PV CF ^ Average PV CF |
(per unit) without Battery ™ with Battery
Figure 7-7. Utility Load Duration Curv e and Cumulative Average
PV Capacity Factor for Scottsdale, AZ.
7-22
-------
(e) Third Quarter 1994
1 10 100 1,000 10,000
Hours
; Utility Load Average PV CF ng Average PV CF
i (per unit) without Battery with Battery
Figure 7-7. (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for Scottsdale, AZ.
Notes on System Operation:
3rd Q '93: System fully operational
4th Q '93: System fully operational
I st Q '94: One of two inverters shut down 3/4/94 - 3/6/94 Cause unknown,
2nd Q '94: System fully operational
3rd Q '94: Svstem at 1/2 power 7/18/94 - 7/20/94 due to DC disconnect fuse failure.
7-23
-------
(a) Third Quarter 1993
Hours
(c) First Quarter 1994
1 10 100 1,000 10,000
Hours
Utility Load Average PV CF Average PV CF
(per unit) without Battery ^ with Battery
Figure 7-8. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for Peoria, AZ,
7-24
-------
(e) Third Quarter 1994
Hours
| Utility Load Average PV CF ^ Average PV CF
(per unit) without Battery ^ with Battery
Figure 7-8. (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for Peoria, AZ.
Notes on System Operation:
3rd Q '93: Data collection began 9/1/93.
4th Q '93: System fully operational
1 st Q '94: System fully operational
2nd Q *94: System fully operational
3rd Q *94: System fully operational
7-25
-------
(a) First Quarter 1994
100 1,000 10,000
Hours
(b) Second Quarter 1994
100
Hours
1,000
10,000
(c) Third Quarter 1994
1,000
10,000
(d) Fourth Quarter 1994
1,000
10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-9. Utility Load Duration Curve and Cumulative Average PV
Capacity Factor for Flagstaff, AZ.
7-26
-------
(e) Top 100 Hours
i
m
4is\u
. •
100
Hours
Utility Load Average PV CF r—> Average PV CF
(per unit) without Battery ^ with Battery
Figure 7-9. (continued) Utility Load Duration Curve and
Cumulative Average ?V Capacity Factor for Flagstaff. AZ.
Notes on System Operation:
3rd Q '93:
PV metering begins 2/12/94,
4th 0 '93:
PV metering begins 2/12/94,
1st 0 94:
System fully operational
2nd Q'94:
System fully operational
3rd Q '94:
System fiilly operational
7-27
-------
(b) Fourth Quarter 1993
1,000
10,000
(c) First Quarter 1994
100
Houns
1,000
10,000
(d) Second Quarter 1994
10 100
Hours
1,000
10,000
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-10. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for Ashwaubenon, WI.
7-28
-------
(e) Third Quarter 1994
Hours
Utility Load Average PV CF Average PV CF
| (per unit) without Battery ™ with Battery
Figure 7-10. (continued) Utility Load Duration Curve and Cumulative
Average PV Capacity Factor for Ashwaubenon, WI.
Notes on System Operation:
3rd Q '93: Building load not monitored until 10/1/93. System at 2/3 power 8/9/93 - 9/9/93.
4th Q '93: System at 2/3 power 10/1/93 - 10/11/93 due to inverter failure.
1st Q *94: System fully operational
2nd Q '94: System fully operational
3rd Q *94: System operated at 2/3 power 9/12/94 - 9/15/94 due to DC disconnect fuse failure.
7-29
-------
(a) Third Quarter 1993
O
>
rj
% - ..p.;*;. >:r< 'k¦ i-f-c,„ |
10
100
Hours
1,000
10,000
(b) Fourth Quarter 1993
100
Hours
1,000
10,000
(c) First Quarter 1994
10 100
Hours
1,000 10,000
(d) Second Quarter 1994
100
Hours
1,000
10,000
Utility Load
(per unit)
Average PV CF
without Batteiy
Average PV CF
with Battery
Figure 7-11. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for Denmark, WT.
7-30
-------
Utility Load Average PV CF p. Average PV CF
(per unit) without Battery ^ with Battery
Figure 7-11. (continued) Utility Load Duration Curve and Cumulative
Average PV Capacity Factor for Denmark, Wl.
Notes on System Operation:
3rd Q '93: System shut down due to inverter failure 7/12/93 ¦ 7/26/93, 9/7/93 - 9/8/93. Building load not monitored until 11/4/93.
4th Q '93: Building toad monitoring begins 11/4/93-
1st Q '94: System fully operational
2nd Q '94: Inverter shut down 5/16/94 - 5/20/94 due to DC injection.
3rd Q '94: System fully operational
7-31
-------
(b) Fourth Quarter 1993
<3 0 9-.
u_
1 10 100 1,000 10.00C
Hours
r 1
i Utility Load Average PV CF Average PV CF
| (per unit) without Battery *—3 with Battery
Figure 7-12. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for Minnetonka, MN.
7-32
-------
(e) Third Quarter 1994
1.000
10,00c
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-12. (continued) Utility Load Duration Curve and Cumulative
Average PV Capacity Factor for Minnetonka, MN.
Notes on System Operation:
3rd Q '93: System fully operational
4th Q '93: System off-line 10/1/93 - 10/12/93 due to inverter failure.
1st Q '94: System fully operational
2nd Q '94: System fully operational
3rd Q '94: System fully operational
7-33
-------
(a) First Quarter 1994
100
Hours
10.00C
(b) Second Quarter 1994
to 100
Hours
1,000
10.00C
'3
a
o.
C5J
O
>
Q_
¦o
03
o
(c) Third Quarter 1994
10
100
Hours
1,000
10.00C
(d) Fourth Quarter 1994
10 100 1,000
Hours
10.00C
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-13. Utility Load Duration Curve and Cumulative Average PV
Capacity Factor for San Ramon, CA.
7-34
-------
Utility Load Average PV CF ™ Average PV CF
I (per unit) without Battery » with Battery
Figure 7-13. (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for San Ramon, CA.
Notes on System Operation:
1st 0 '94: PV system data acquisition commences 1/7/94. Load measurement commences 3/7/94. System shut down 2/11/94 -
2/14/94.
2nd Q '94: System operated at 2/3 power 5/14/94 - S/27/94 due to inverter failure.
3rd Q '94: One string disabled after testing. System limited to about 80% 7/1/94 - 8/11/94.
4th Q '94: System operated at 2/3 power 10/19/94 - 10/24/94 due to inverter failure.
7-35
-------
(c) Second Quarter 1994
(d) Third Quarter 1994
1,000
10.00C
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-14. Utility Load Duration Curve and Cumulative Average
PV Capacity Factor for Austin, TX,
7-36
-------
Utility Load Average PV CF ^ Average PV CF
(per unit) without Battery ^ with Battery
Figure 7-14. (continued) Utility Load Duration Curve and Cumulative
Average PV Capacity Factor for Austin, TX.
Notes on System Operation:
4th Q '93: Properly functioning PV meter not installed until 3/14/94. PV generation prior to that date was simulated using irradiance data.
1st Q '94: Properly functioning PV meter not installed until 3/14/94. PV generation prior to that date was simulated using in-adiance data.
2nd Q '94: System at 1/3 power 5/11/94 - S/2S/94 due to inverter failure, off-line6/l 1/94 - 6/14/94 due to fuse failure.
3rd Q '94: System operated at 2/3 power for total of 12 days due to fuse failure.
7-37
-------
(a) Third Quarter 1994
1000
10,00c
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-15, Utility Load Duration Curve and Cumulative Average PV
Capacity Factor for Barstow, CA,
7-38
-------
(e) Top 100 Hours
1 10 100
Hours
¦ Utility Load Average PV CF Average PV CF
! (per unit) without Battery ^ with Battery
Figure 7-15. (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for Barstow, CA.
Notes on System Operation:
3rd Q '94: System shut down 7/22/94 - 8/3/94 due to DC disconnect ftise failure. Shading reduces output by 30 to 40 percent.
4th Q '94: System fully operational.
1 st Q '95: System fully operational.
2nd Q '95: Work on local distribution circuit causes inverter to operate intermittently 5/25/95 - 6/10/95.
7-39
-------
(a) First Quarter 1994
100
Hours
10.00C
o
T5
<3 I
U.
^ ,
O
>
Q_
T3
a'
1'
-j i
I,
3
(c) Third Quarter 1994
^ , '¦tJ'S -„¦>
, '.'xffitoS* , j
. - ¦«u
10
100
Hours
1,000
10,00c
(d) Fourth Quarter 1994
1,000
10.00C
Utility Load
(per unit)
Average PV CF
without Battery
Average PV CF
with Battery
Figure 7-16. Utility Load Duration Curve and Cumulative Average PV
Capacity Factor for Edwards Air Force Base, CA.
7-40
-------
(e) Top 100 Hours
m
Cl
m
U
> i
EL
«• ^ 't t'jji
-fv *
-------
(a) First Quarter 1994
1 10 100 1,000 10.00C
Hours
(b) Second Quarter 1994
1 10 100 1,000 100CX
Hours
(d) Fourth Quarter 1994
o '
¦3
m o.9
§¦ 0.8
rc
g- 0.7
o
> 06
EL
"g 0.5
fC
3 0.4
Q.
-a 0,3
ro
o
—'0 2
5?
ip 0.1
Z3
0
1
•*r\
10 100 1,000
Hours
io,:oc
| Utility Load mm Average PV CF _ Average PV CF
(per unit) without Battery ™ with Battery
Figure 7-17. Utility Load Duration Curve and Cumulative Average PV
Capacity Factor for Palrr. Desert, CA.
7-42
-------
(e) Top 100 Hours
Hours
numi Utility Load Average PV CF r~| Average PV CF
(per unit) without Battery ™ with Battery
Figure 7-17 (continued) Utility Load Duration Curve and
Cumulative Average PV Capacity Factor for Palm Desert. CA,
Notes on System Operation:
1 st Q '94: Data acquisition commences 2/L/94. System fully operational.
2nd Q '94: System fully operational
3rd Q"9A: System fully operational
4th Q '94: System fully operational
7-43
-------
Chapter 8
Calculation of Emission Offsets
8.1 Case Study Approach
One intent of this project was to determine the effectiveness of photovoltaic power generation as a
pollution mitigation strategy. Given this goal, every effort was made to calculate, on a real-time basis,
the emission offsets (i.e., reduction in pollutant emissions) from each utility system as a result of PV
power generation. As each of the participating utilities has a unique combination of generating
resources, fueling options, load characteristics and solar resources, the results of these calculations,
discussed in section 8.3 below, are a series of case studies representing the pollution mitigating effects of
PV on each utility system.
8.1.1 Description of Methodology
The methodology for calculation of emission offsets is straightforward: multiply PV energy generation
in each hour of the study period by the per-kWh emission rate for each of the pollutants in each hour.
Except in cases where PV generation data was missing or inaccurate due to a delay in commencement of
metering or a meter malfunction (see section 3.1 of the Quality Assurance Project Plan), hourly PV
generation (in kWh ) was simply the average of the four 15-minute PV power generation averages leading
up to the hour (e.g., the value for 10:00 would be the average of the four values recorded at 9:15, 9:30,
9:45, and 10:00). When a PV system was operating, but was either unmetered or the meter pulse
constant was incorrect, simulated PV generation (based on measured plane-of-array irradiance) was
substituted for measured generation (note that this method penalizes the PV system for lack of
reliability). Calculation of the second component of emission offsets, pollutant emission rates, was less
straightforward, and is discussed in the next two sections.
8.1.2 Emission Rates of Marginal Generating Units
A fundamental assumption of this study is that the energy generated by the PV systems would not be of a
sufficiently large magnitude to alter the dispatch of the utility's conventional generators. Based on this
assumption, it follows that the energy generated by the PV systems reduced the output required of the
marginal generating unit(s) on each system (i.e.. the power plant providing the last kWh demanded by
consumers—typically the plant with the highest operating costs which is generating at any given time
would be identified as the marginal unit). Accordingly, all participating utilities were asked to provide
emission rates for all units used for automatic generation control and an indication of which unit was on
the margin in each hour. Following a quality assurance audit conducted by the EPA on April 18 and 19,
1994, calibration records of instruments used to measure emissions and fuel consumption were also
requested (see Appendix B).
8-1
-------
8.2 Development of Marginal Emission Models
The methodology employed to determine marginal emission rates for each utility depended upon several
factors, including limitations on the emissions information possessed by the utility for the units
concerned and the purchase of power from other utilities or utility pools. The methodologies can be
grouped into three categories: 1) utilities modeled by a constant rate for all hours; 2) utilities modeled by
hourly weighted averages of units generating at or near the margin; and 3) utilities modeled by the
emission rate of a single unit in each hour. The methodology applied to each site is summarized in Table
8-1 and described in greater detail below. Computer software was developed to accept the emission rate
data provided by each utility and produce hourly emission rates in a consistent format. Additional
software integrated these files with hourly PV generation data to determine hourly emission offsets.
Table 8-1. EMISSION RATE CALCULATION METHODOLOGIES
Methodology Utilities
Constant Rate
NYSEG, NYPA, NSP, SCE
Hourly Weighted Average
NU, NEES, ACE, WPS, PG&E
Single Unit in Each Hour
APS, COA
8.2.1 Sites Modeled with a Constant Marginal Rate
a. New York State Electric and Gas (NYSEG)
NYSEG provided constant hourly emission rates for each season of the study period. For the winter of
1993/1994 and the summer of both years, these rates %vere calculated by determining the type of unit (oil-
fired boiler for the winter, gas-fired boiler for the summers) that would typically be on the margin in
each season, assuming average heat rates for these units, and applying the pollutant emission factors
contained in the EPA publication Compilation of Air Pollutant Emission Factors - Vol. J: Stationary
Point and Area Sources, commonly referred to as the AP-42 report. For the spring and fall seasons,
emission rates were based on stack test and sampling data from actual units identified as operating at or
near the utility's margin,
b. New York Power Authority (NlfPA)
NYPA also derived emission rates from the AP-42 report for a single load-following unit on its system.
The emission rates for this unit were applied to PV generation for all hours throughout the study period.
c. Northern States Power (NSP)
NSP supplied two sets of emission rates: one characterizing average emissions from fossil generating
units on its system, the other characterizing average system wide emission rates including nuclear and
hydro generation. The average system wide emission rates were selected for calculation of emission
offsets because NSP claimed that it uses its three "cleanest" fossil units as well as hydro to follow load.
The emission rates based only on fossil units would therefore have overstated emission offsets by the PV
system. The average system wide emission rates were applied to PV generation for all hours throughout
the study period.
d. Southern California Edison (SCE)
SCE provided a single set of emission rates representative of the natural gas-fired combustion turbines
the company uses to follow load. The emission rates assume selective catalytic reduction of NOx
8-2
-------
emissions on all units. These rates were applied to PV generation for all hours throughout the study
period.
8.2.2 Sites Modeled by a Weighted Average of Units onAGC
a. Northeast Utilities (NU) and New England Electric System (NEES)
Because both of these companies belong to the New England Power Pool (NEPOOL), the same emission
rates were used to determine emission offsets for both companies. The emission rate for each pollutant
in each hour was calculated by weighting the emission rate for each of 45 units on automatic generation
control ("AGC") by each unit's generation for the hour, summing the resulting products across all units,
and then dividing by the aggregate generation of all units on automatic generation control. The resulting
rates therefore represent an hourly weighted average of all units on the NEPOOL margin.
b. Atlantic City Electric (ACE)
The methodology for determining houly emission rates for ACE was identical to that used forNU and
NEES, with the exception of hours during which ACE serves marginal load with power purchased from
the Pennsylvania-New Jersey-Maryland (PJM) power pool. For these hours, constant average rates
representing emissions from the PJM pool were used.
c. Wisconsin Public Service (WPS)
WPS provided emission rates and hourly generation data for each of five plants which are used to follow
load. These data were used to calculate weighted average emission rates for each hour of the study
period.
d. Pacific Gas and Electric (PG&E)
PG&E provided emission rates and hourly generation for ten power plants. Hourly weighted average
emission rates were calculated in the same manner as described above for NU and NEES.
8.2.3 Sites Modeled by the Emission Rate of a single unit in each hour.
a. Arizona Public Servic (APS)
APS supplied emission rate data for each of 21 load-following power plants on its system, and identified
which of these units was on the margin in each hour. The emission rates of the marginal plant in each
hour were used to determine emission offsets for that hour.
b. City of Austin Municipal Utility (COA)
COA provided one set of average emission rates for all coal-fired units on its system and another set for
all gas-fired units. COA also indicated which type of unit was on the margin in each hour. Emission
rates corresponding to the type of unit on the margin were applied to PV generation to determine
emission offsets for each hour.
8.3 Results
The emission offset results for each system are discussed below, and are illustrated in the figures at the
end of this section. For each site, a set of four bar charts similar to the one in Figure 8-1 shows the
monthly offset for each of the pollutants of interest (sulfur dioxide (SQ2), nitrogen oxides (NOx), carbon
dioxide (C02), and particulates). Results are shown both for the system as it actually operated and for a
8-3
-------
simulation of system performance assuming the dispatchable storage system described in chapter 5 had
been installed. Note that all offsets have been normalized by each system's rating at standard operating
conditions ("SOC"), defined as 1000 W/nr irradiance, and 20CC ambient air temperature. The offsets
shown in these charts are therefore in units of kg/kWsm. Offsets have been normalized so that those due
to one system may be compared with those from another, and so that these results may be easily scaled
up to any size of PV system.
The lighter shaded region in the background of each of these charts provides a rough estimate of what an
"average" PV system would offset, were its output displacing
generation by the current mix of generating units on the
national grid. This estimate was developed by calculating
three regional averages of monthly generation (per kW5ac)
for PV systems in the Northeast, the Midwest, and the
Southwest.1 The mean of these regional averages was taken
to be roughly representative of the monthly generation (per
k\VS0C) that could be expected from the "average" PV system
in the U.S. Average monthly offsets for each of the four
pollutants were then calculated by applying average national
emission rates to monthly generation.2 Again, this estimate
of the national average offset is a rough, order of magnitude
approximation, but it does allow one to identify at a glance
sites which are particularly good or particularly poor
candidates for the use of PV as an emission mitigation
measure.
For some of the PV systems, the month-to-month variation in
the level of pollutant offsets closely follows the seasonal variation in the output of the PV system.
However, where a utility's emission rate for a pollutant varies substantially from month to month, due to
a change in its fuel mix for example, the pattern of offsets can diverge from this seasonal pattern. It can
therefore be difficult to discern whether higher offsets in a given month are the result of a particularly
good month for the PV system or particularly high utility emissions in that month. In the discussion that
follows, it will therefore be useful to note that, with very few exceptions, the monthly variation in C02
offsets accurately reflects variation in monthly generation by the PV system (i.e. the month-to-month
change in C02 offset is proportional to the change in generation). The monthly variation in CO, offsets
can therefore be used as a proxy for the pattern of generation, which is useful in interpreting the results
for other pollutants.
Table 8-2 contains the average monthly offset for this hypothetical 1 kWS0C PV system, based on the 12
months beginning October 1993. The table also contains a 12 month total for this average system, as
Sulfur Dioxide
3.5
3.0
^ 25
8
g 2.0
§
1 1"S
Jm.o
2
0.5
0.0
^Without Storage
|W sih Storage
~National Average
3rdQtr'§3 IstQtr '94 3rdQT94
4 th Qtr '93 2nd Gfr '94
Figure 8-1. Example Offset Chart.
'This step was necessary since the 16 PV systems studied in this project are riot distributed equally across the country. A simple average of all
16 systems -would have overestimated average monthly generation, since eight of the 16 systems are in areas where the solar resource is
substantially greater than the rest of the country.
^National emission rates were derived by dividing total utility emissions for 1993 as listed in Table 44 of the Energy Information
Administration publication Electric Power Annual 1993 (12/94) by total kWh generation for that year.
8-4
-------
well as the average offset per MWh (i.e. emission rate). Similar tables are provided to describe the
emission offsets of each of the real PV systems.
Table 8-2. MONTHLY AND ANNUAL OFFSETS FOR THE AVERAGE U.S. PV SYSTEM
SO;
NOs
CO,
Particulates
Average Monthly Offset (kg/kWSK)
0.54
0.22
12
0.017
Total Annual Offset (kg/kWS0C)
6.4
2.6
870
0.21
Average Emission Rate (kg/MWh)
4.5
1.8
610
0.17
8.3.1 Pittsburgh, NY (EPA01)
The data in Figure 8-2 indicate a substantial seasonal variation for all pollutants. This variation is due
both to the normal seasonal fluctuation in the solar resource and to seasonal changes in the units NYSEG
used to follow load. The latter factor is far more influential in determining monthly offsets of S02, NO*
and particulates than is the former. The ratio of highest to lowest emission rate during the study period is
4,800 for SO;. 19 for NO„ and 500 for particulates. Thus a kWh generated by the system would have
profoundly different offsets of these pollutants, depending on the season in which it occurred.
By referring to the average monthly capacity factors in Figure 6-6 on page 6-13, one can see that the
variation in the monthly offset of CO, reflects monthly variation in PV generation far more closely than
do the offsets for SO,, NOx or particulates. As noted above, the monthly pattern of CO. offsets may
therefore be used as a proxy for the monthly pattern of PV generation.
The CO, offsets roughly follow the seasonal pattern of the national average, moderated by the system
outages noted below the figure. Offsets of S02. NO„, and particulates follow this pattern in most months.
However, these offsets are greatly reduced in the months of June, July, and August of 1994. As the CO,
data suggest, there was substantial generation in these months. However, the marginal emission rates for
the summer months were those of a gas-fired boiler with low emissions of SO,, NOx. and particulates,
resulting in the very low offsets shown in the figure.
The figure indicates that offsets of all four pollutants are sharply reduced in the simulation of
dispatchable storage. This is due to the modeling assumption that 25 percent of the energy generated by
Table 8-3. MONTHLY AND ANNUAL OFFSETS FOR EPA01
SO,
NO,
CGj
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWS0<;)
0.58
0.44
0.18
0.13
55
41
0.020
0.015
Total Annual Offset (kg/kW,0C)
7.0
5.2
2.2
1.6
650
490
0.24
0.18
Total Annual Offset (kg)
75
57
23
17
7.100
5.300
2.7
2.0
Average Emission Rate (kg/MWh)
7.7
7.7
2.4
2.4
730
730
0.27
0.27
8-5
-------
the PV system would be lost in charging and discharging the batten, . Because marginal emissions from
the NYSEG system were modeled with a constant rate for all hours (within each season), the difference
between the timing of actual generation by the system and that of the simulated dispatch had no effect on
offsets. Emission offsets under the dispatchable storage scenario are therefore exactly 25 percent below
actual offsets in all months.
Table 8-3 presents average monthly and annual offsets for this system, based on the 12-month period
beginning October 1993. Although NYSEG's marginal emission rates vary widely from month to
month, the data in this table indicate that over the course of a year, the PV system in Pittsburgh reduced
emissions by about the same amount as the "national average" system.
8.3.2 Berlin, CT (EPA02)
The charts of pollutant offsets in Figure 8-3 closely follow the pattern of monthly generation by this PV
system (as reflected by the CO, offsets), and remain close to the national average in all months. This
indicates that NEPOOL's marginal emission rates for the four pollutants of interest remained relatively
stable throughout the study period. It should be noted that marginal emissions from this utility pool were
modeled by calculating the weighted average (by plant loading) of 45 units on automatic generation
control for each hour of the study period. The fuel diversity of these generators may explain why the
emission rates appear to have remained so stable.
As expected, the emission offsets resulting from the dispatchable storage simulation are substantially
below the offsets due to actual PV generation. Again, this is largely due to the modeling assumption that
25 percent of the energy generated by the PV system would be lost in charging and discharging the
battery. Observation of Table 8-4 however, indicates that this energy loss was not the only factor. For
this PV system, the average emission rate of all four pollutants was between 10 and 17 percent lower in
the storage scenario. This may be due to the fact that the battery dispatch algorithm used stored energy
to offset utility generation only during the highest load hours of each day, when the generating units with
the highest variable costs are called into service to carry peak loads. Often, these units bum natural gas,
which is c leaner, but usually more expensive than either oil or coal. If the proportion of load served by
natural gas-fired units on automatic generation control is greatest during NEPOOL's highest load hours,
then emission rates will be lowest during these hours.
Table 8-4. MONTHLY AND ANNUAL OFFSETS FOR EPA02
SO,
NO,
CO,
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWS0C)
0.59
0.40
0.21
0.14
72
50
0.021
0.014
Total Annual Offset (kg/kWJ
7,1
4.8
2.6
1.6
870
600
0.25
0.17
Total Annual Offset (kg)
26
17
9 3
5.9
3,200
2,200
0.89
0.60
Average Emission Rate (kg/MWh)
5.3
4.8
1.9
1.6
660
600
0.19
0.17
8-6
-------
8.3.3 Pleasantville and Brigantine, NJ (EPA 03 and 04)
The emission offsets from these systems, shown in figures 8-4 and 8-5, are substantially higher than
average for all four pollutants in all but the last few months of the study period. The fact that monthly
variation in S02, NO,, and particulates follows that of CO, exactly for the first four quarters of the study
period indicates that this variation is due to the month-to-month changes in generation by the PV
systems, rather than changes in Atlantic City Electric's emission rates. Offsets in the third quarter of
1994 were sharply reduced relative to previous months. However, the monthly system capacity factors
shown in Figure 6-12 on page 6-17 indicate that PV generation in these months was relatively stable at a
level only slightly lower than generation during the previous quarter. This suggests that the cause of
these reduced offsets must be a decline in the emission rates. This is in fact the case. There are a large
number of hours in these months during which Atlantic City Electric served its marginal load with power
purchased from the Pennsylvania-New Jersey-Maryland (PJM) power pool. Average PJM emission rates
(which are lower than ACE marginal emission rates) were used to determine offsets during such hours.
Table 8-5. MONTHL Y AND ANNUAL OFFSETS FOR EPA03
S02
NO,
O
o
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kW^)
1.2
0.81
0.52
0.36
130
97
.030
0.021
Total Annual Offset (kg/kW„)
14
9.7
6.2
4.4
1,600
1,200
0.37
0.26
Total Annual Offset (kg)
150
100
68
47
17,000
13,000
4.0
2.8
Average Emission Rate (kg/MWh)
11
10
4.9
4.5
1,300
1,200
0.28
0.26
Table 8-6. MONTHLY AND ANNUAL OFFSETS FOR EPA04
SO,
NO„
CO,
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWMe)
1.4
0.96
0.60
0.43
150
110
0.037
0.026
Total Annual Offset (kg/kWS0C)
16
11
7.2
5.1
1,800
1,400
.44
0.31
Total Annual Offset (kg)
59
41
26
19
6.700
4,900
1.6
1.1
Average Emission Rate (kg/MWh)
11
10
4,7
4.5
1,200
1,200
0.29
0.27
The data in tables 8-5 and 8-6 show that emission offsets from this site were close to twice the national
average for all four pollutants, suggesting that, relative to other areas of the country, PV would be an
effective pollutant mitigation measure in the Atlantic City Electric service area. Emission offsets under
the dispatchable storage scenario are reduced, as expected, due to battery losses. These tables also
indicate that the average emission rates are slightly lower for the battery simulation than they were for
8-7
-------
the actual system operation. As described in the previous section, the reduced average rates are most
likely attributable to increased reliance on natural gas-fired generation during the highest daily load
hours, when the simulation dispatched all energy stored in the battery.
8,3.4 White Plains, NY (EPA05)
Marginal emissions for the New York Power Authority were modeled by a single gas- and oil-fired
generator for the entire study period. Figure 8-6 shows that because this unit's SO, emission rate was
quite low, the SO, offsets from this system were only a small fraction of the national average.
Particulate offsets, on the other hand, were among the highest observed (on a per-kW basis) in the
project. The particulate emission rate used for this generator was quite high due to the fact that it has no
equipment to reduce emissions of particulates. C02 offsets follow the seasonal trend of the national
average, with the exception of July and August 1993 and June 1994 when outages limited system
generation.
As both Figure 8-6 and Table 8-7 indicate, emissions offsets were considerably smaller under the
dispatehable storage scenario. Again, this is due to the modeling assumption that 25 percent of the
electricity generated by the PV system would be lost in charging and discharging the battery.
Table 8-7. MONTHLY AND ANNUAL OFFSETS FOR EPA05
SO,
NO*
CO,
Particulates
vv/o
storage
with
storage
w/o
storage
with
storage
Wr/O
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWJ1!C)
0.018
0.013
0.12
0.092
59
43
0.050
0.037
Total Annual Offset (kg/kW^.)
0.22
0,16
1.5
1.1
700
520
0.60
0.44
Total Annual Offset (kg)
0.78
0.58
5.4
4.0
2,500
1,900
2.2
1.6
Average Emission Rate (kg/MWh)
0.19
0.19
1.3
1.3
600
600
0.52
0.52
8.3.5 Scottsdale, Peoria, and Flagstaff, AZ (EPA06, 07, and 13)
These three systems were installed for Arizona Public Service. Their emission offsets therefore have
similar characteristics and will be discussed jointly. The charts of C02 offsets in figures 8-7 through 8-9
indicate that generation by all three systems followed the seasonal cycle quite closely, with little
disruption due to outages. COz offsets from these systems are roughly twice the national average,
resulting from a combination of higher than average generation and the fact that some of the units used to
follow load had relatively high heat rates.
Note that the monthly SO, and particulate offsets do not follow the seasonal pattern of the solar resource,
and in fact appear to counter that pattern in many months. This is explained by the fact that although
Arizona Public Service uses gas-fired generators to do most of its load following, these generators are
dual-fueled, and APS does occasionally use oil in the load following units. The month-to-month
variation in SO, and particulate offsets follows the monthly pattern of oil consumption precisely,
suggesting that the emission of these pollutants is due primarily to the use of oil.
8-8
-------
The use of oil-fired generation also appears to play a role in the nearly flat pattern of NO„ offsets shown
in the figures. Oil consumption by the units APS uses to follow load appears to be highest in winter
months and lowest during the summer, as is reflected in the SO, offsets. Because the NO* emission rates
of these units are relatively high when burning oil, NO, offsets per kWh are higher in the winter months,
so that even though the PV system was generating fewer kWh in these months, its NO, offset remained
relatively stable.
The figures illustrate that under the storage simulation, emission offsets would be reduced compared to
offsets due to actual system operation. As discussed above, the primary factor is the assumed energy
losses in charging and discharging the battery. Tables 8-8, 8-9, and 8-10 indicate that another factor
causing reduced SO, offsets is an approximately 12 percent reduction of the SO, emission rate in the
storage scenario. The lower S02 emission rate is due to the fact that during the hours in which the
simulation dispatches storage, generation by gas-fired units was a greater proportion of total generation.
Interestingly, the NQX emission rate was slightly higher for the storage scenario, despite the diminished
role of oil-fired generation during dispatch hours. A likely explanation for this is the increased reliance
on gas turbines during peak hours. The NOx rates of the gas turbines are higher than many of APS' oil-
fired units, and are approximately twice as high as the gas-fired combined-cycle units that provide the
bulk of the APS load-following generation.
The summary statistics in the tables indicate that the annual C02 offset is very close to twice the
average offset. This suggests that the APS service territory is a relatively good one for the use of PV as a
C02 mitigation measure. The tables also demonstrates that PV would be relatively ineffective as an SO,
mitigation measure.
Table 8-8. MONTHLY AND ANNUAL OFFSETS FOR EPA06
SO,
NO,
CO,
Particulates
w/o
storage
with
storage
w/o
Storage
with
storage
W/O
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWS0C)
0.10
0.067
0.29
0.23
160
120
0.016
0.012
Total Annual Offset (kg/kWSK)
1.2
0.80
3.5
2.7
2,000
1500
0.20
0.14
Total Annual Offset (kg)
8.8
5.8
25
20
14.000
11.000
1.4
1.0
Average Emission Rate (kg/MWh)
0.66
0.58
1.9
2.0
1,100
1,100
0.11
0.10
Note that the average monthly and total annual offsets per kWS0C are somewhat lower for the Peoria
system than those listed above. The Peoria system did not perform as well as the Scottsdale system
because 2/3 of the PV array at this site was installed at a very low tilt angle. The low tilt angle reduced
system output during the winter months, and allowed a thick band of clay to cover a portion of the cells
at the bottom of each module, further reducing system output. Note also that this system has half the
rating of the Scottsdale system, thus the correspondingly lower offsets.
8-9
-------
Table 8-9. MONTHLY AND ANNUAL OFFSETS FOR EPA07
SO,
NOx
co2
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/k\VJ0C)
0,092
0.059
0.26
0.21
150
no
0.015
0.011
Total Annual Offset (kg/kWMC)
1.1
0.71
3,2
2.5
1,800
[,300
0.18
0.13
Total Annual Offset (kg)
4.0
2.5
11
8.9
6.400
4,800
0.64
0.46
Average Emission Rate (kg/MWh)
0.66
0.57
1.9
2.0
1,100
1,100
0,11
0.10
Table 8-10. MONTHLY AND ANNUAL OFFSETS FOR EPA 13
SO,
NO,
O
O
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kW!0C)
0.089
0.059
0.27
0.21
151
no
0.015
0,014
Total Annual Offset (kg/kWSM)
1.1
0.71
3.2
2.5
1,800
1,400
0.17
0,17
Total Annual Offset (kg)
3.9
2.6
12
8.9
6.600
4,900
0.63
0.63
Average Emission Rate (kg/MWh)
0.62
0.55
1.9
1.9
1,100
1.100
0.10
0.14
8.3.6 Ashwaubenon, WI and Denmark, WI (EPA 08 and EPA09)
These systems were both installed in the Wisconsin Public Service serv ice area, and are therefore
discussed jointly. Marginal emission rates for the WPS system were determined by calculating a
weighted average rate for the five coal-fired units the company uses to follow load. Since coal is used to
follow load, one would expect to see relatively high C02 offsets for these systems. As figures 8-10 and
8-11 demonstrate, this is in fact the case for most months. However, because system generation in
Ashwaubenon was well below the project average in the winter months (due to lower insolation and
snow cover), this system's offsets of CO. as well as the other pollutants are near or even below the
average in the months November 1993 through February 1994. These charts, and the values in the
following tables suggest that despite limited generation in the winter, this service area is a relatively
good one for the use of PV as a mitigation measure for both CO, and NO,.
The figures illustrate the expected reduction in emission offsets in the storage scenario. Because
marginal emission rates were determined using generation by the same set of coal-fired generators in all
hours, these rates do not show the variability demonstrated by some of the systems discussed above.
Tables 8-11 and 8-12 indicate that the average emission rate was the same for the PV systems as they
actually operated and for the storage simulation scenario.
8-10
-------
Table 8-11. MONTHLY AND ANNUAL OFFSETS FOR EPA08
SO,
NO,
CO,
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kW,0C)
0,59
0.44
0.65
0.49
120
91
0,022
0.016
Total Annual Offset (kg/kWscc;)
7.0
5.3
7.8
5.8
1.500
1,100
0,26
0.19
Total Annual Offset (kg)
76
57
84
63
16,000
12,000
2,8
2.1
Average Emission Rate (kg/MWh)
5,6
5.6
6.2
6.2
1,200
1,200
0.21
0.21
Table 8-12. MONTHLY AND ANNUAL OFFSETS FOR EPA09
SO,
NO,
CO,
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWso<;)
0.67
0.50
0.72
0.54
140
100
0.024
0.018
Total Annual Offset (kg/kWsoc)
8.0
6.0
8.7
6,5
1,600
1,200
0.29
0.22
Total Annual Offset (kg)
29
22
31
24
5,900
4,500
1.1
0.80
Average Emission Rate (kg/MWh)
5.6
5.6
6.1
6.1
1,200
1,200
0.21
0.21
8.3.7 Minnetonka, MN (EPA JO)
The emission offsets due to this system are presented in Figure 8-12. Since emissions for Northern
States Power were modeled with a single system wide average rate for all hours (for each pollutant), the
offsets shown in the figure are simply scaled versions of what a plot of monthly kWh generation would
look like. With the exception of S02, this PV system reduced pollutant emissions from NSP power
plants by approximately the national average in all months. SO, offsets are about half the average level
in most months. The S02 emission rate used for this company was relatively low due to the fact that
NSP serves about 35 percent of its load with nuclear power, and much of the rest is served by units
burning low sulfur coal with 70 percent sulfur removal.
Emission offsets under the dispatchable storage scenario are reduced due to energy losses in charging
and discharging the battery. Because this system was modeled as having constant emission rates in all
hours, there is no difference between the average emission rates offset by the PV system as it actually
operated and the emission rates during the simulation's dispatch hours.
8-11
-------
Table 8-13. MONTHLY AND ANNUAL OFFSETS FOR EPA10
SO,
NOv
O
u
Particulates
w/o
storage
wilh
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWS0C)
0.18
0.14
0.22
0.17
72
54
0.017
0.013
Total Annual Offset (kg/kWS0C)
2.2
1.7
2.7
2.0
870
650
0.20
0.15
Total Annual Offset (kg)
8.0
6.0
9,7
7.3
3,100
2.300
0.74
0.55
Average Emission Rate (kg/MWh)
1.7
1.7
2.0
2.0
640
640
0.15
0.15
8.3.8 San Ramon, CA (EPA11)
Despite the fact that Pacific Gas and Electric uses gas-fired units to serve most of its marginal load, C02
offsets from the PV system in San Ramon were more than twice the average in all months following
system startup in January 1994. The magnitude of these offsets is partially due to this system's
generation level which was consistently well above average, and by the fact that some of the units used
to follow load had relatively high C02 emission rates. Figure 8-13 shows that offsets of the other three
pollutants were quite low, due both to the use of natural gas and emission control equipment.
Table 8-14. MONTHLY AND ANNUAL OFFSETS FOR EPA11
SO,
NO,
f-l
O
u
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kVV^)
0.027
0.021
0 13
0.096
190
140
0.0048
0.0036
Total Annual Offset (kg/kWMC)
0.33
0.25
1.5
1.2
2.300
1,700
0.057
0.043
Total Annual Offset (kg)
3.6
2.7
16
12
25,000
18,000
0.62
0,46
Average Emission Rate (kg/MWh)
0.19
0.19
0.90
0.90
1,300
1,300
0.034
0.034
8.3.9 Austin, TX (EPAJ2)
As discussed above, marginal emissions for the Austin municipal utility were modeled by using the
emission rates for either a coal- or a gas-fired generator in each hour of the study period. Varying
relative proportions of generation by these fuels over the course of the study period resulted in the
atypical pattern of emission offsets shown in Figure 8-14. As one might expect, offsets of all pollutants
closely followed the pattern of coal consumption, although the effect on CO, and NOx is less evident
since gas combustion also produces these pollutants at appreciable rates.
The clearest example of the effect of coal on the PV system's offsets can be seen by comparing
emissions in "November and December of 1993. In November, marginal load was served by a coal-fired
unit for only a very few daylight hours. As a result, offsets of SO, and particulates were very low. In
8-12
-------
December, a coal-fired unit served marginal load in all daylight hours, resulting in some of the highest
monthly offsets of all pollutants despite the fact that generation by the PV system that month was at the
second-lowest level observed during the study period. This is one of the few instances observed where
variation in monthly C02 offsets does not match variation in monthly PV generation.
The simulation of the PV system with dispatchable storage resulted in SO,, NO*. C02, and particulate
offsets that were 48%, 68%, 70%, and 53% respectively of the offsets calculated for the PV system
without storage. These low ratios indicate that the Austin utility is substantially more reliant upon its
cleaner gas-fired load following unit during daily peak load hours. This is reflected in the lower average
emission rates for the dispatchable storage simulation, as shown in the table below.
Table 8-15. MONTHLY AND ANNUAL OFFSETS FOR EPA 12
SO:
NO,
n
O
Particulates
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWM)
0.10
0.050
0.14
0.094
79
56
0.0093
0.0049
Total Annual Offset (kg/kW„c)
1.3
0,60
1.7
1.1
1,000
670
0.11
0.059
Total Annual Offset (kg)
14
6.5
18
12
10,000
7,200
1.2
0.64
Average Emission Rate (kg/MWh)
0.94
0,60
1.3
1.1
720
680
0.084
0.059
8.3.10 Barstow, Edwards A FB, and Palm Desert, CA (EPA 14,15 and 16)
As generation by these three PV systems offset emissions from the Southern California Edison (SCE)
power system, the characteristics of their offsets are similar arid are therefore discussed jointly.
Emission offsets from these systems are illustrated in figures 8-15 through 8-17. Note that these figures
present 12 months of data starting in June 1994 for EPA14 and February 1994 for EPA15 and EPA16.
These time periods reflect the fact that these systems began operation much later than most of the other
systems installed by this project.
As discussed above, marginal emission rates for this system were modeled by a constant rate for all
pollutants. Because SCE uses only gas-fired units to follow load, emission rates of both S02 and
particulates are very low. Also, the NO, emission rate provided by SCE assumed that selective catalytic
reduction is in place in all load-following units and is therefore quite low as well.
The data displayed in the figures reflects the very low emission rates for S02, NOx, and particulates.
Offsets for the system in Barstow are lower than those for the other two systems because much of this
PV array was shaded in the afternoon, substantially reducing generation. As the figures indicate, CO,
offsets for these systems are higher than the estimated national average offset in the winter and spring,
and close to the average in summer months. Because marginal emissions were modeled with a constant
rate in all hours, the fact that the simulation shifted each system's hours of operation had no effect on
emission offsets. The simulated offsets from each system were smaller than the offsets due to actual
system operation, but again this is due to the modeling assumption that 25 percent of energy generated
by the system would be lost in charging and discharging the battery.
8-13
-------
The average emission rates for SO,, NO„, and particulates in tables 8-16 through 8-18 are small fractions
of the corresponding estimates of the national average rates listed in Table 8-2. The CO, emission rate
for SCE is also somewhat lower than the national average CO, rate, but because these systems achieved
higher than average capacity factors, their annual C02 offsets per kWsoc exceed the average annual CO,
offset listed in Table 8-2,
Table 8-16. MONTHLY AND ANNUAL OFFSETS FOR EPA14
SO,
NO*
CO,
Particulates
w/o
storage
with
storage
vv/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kW^)
,00034
.00025
0.0088
0.0066
67
50
0.0014
0.0011
Total Annual Offset (kg/kWJ
0.0040
.0030
0.11
0,079
800
600
0.017
0.013
Total Annual Offset (kg)
0.015
0.011
0.38
0.29
2,900
2,200
0.063
0.047
Average Emission Rate (kg/MWh)
0.0028
0.0028
0.072
0.072
550
550
0.012
0.012
Table 8-17. MONTHLY AND ANNUAL OFFSETS FOR EPA15
SO;
NO,.
O
O
Particulates
vv/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kWsoc)
,00038
.00030
0.011
0,0079
80
59
0.0017
0.0013
Total Annual Offset (kg/kWwc)
0.0046
0,0036
0.13
0.094
960
710
0.021
0.016
Total Annual Offset (kg)
0.017
0.013
0.46
0.34
3,300
2,600
0.075
0,056
Average Emission Rate (kg/MWh)
0.0028
0.0028
0.073
0.073
550
550
0.012
0.012
Table 8-18. MONTHLY AND ANNUAL OFFSETS FOR EPA 16
so.
NO,
O
O
i-j
Particulates
vv/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
w/o
storage
with
storage
Average Monthly Offset (kg/kW^)
.00036
.00028
0.0099
0.0074
75
56
0,0016
0,0012
Total Annual Offset (kg/kWMC)
0.0045
0.0034
0.12
0.089
900
670
0.019
0.015
Total Annual Offset (kg)
0.047
0,037
1,3
0.96
9,700
7,300
0.21
0.16
Average Emission Rate (kg/MWh)
0,0028
0,0028
0.073
0.073
550
550
0.012
0.012
8.4 Conclusions
Two primary factors contribute to determine the emission offsets that a PV system will provide in any
given location; the marginal emission rates of the utility system into which it is connected and the solar
resource available for conversion to electricity. The data collected through this project has indicated that
8-14
-------
these two factors combine to create tremendous variability in the emission offsets resulting from the
operation of these systems.
Because marginal emission rates are dependent upon the pollutant content of the fuel used in load
following generators, the combustion temperature and operating efficiency of those generators, and the
presence or absence of pollution control equipment, these rates exhibited considerably greater variability
between utilities than did the solar resource. The ratio of the highest average SO, emission rate
(determined by dividing the total S02 offset over a 12-month period by a PV system's total generation
over that period) to the lowest such rate was nearly 4.000. Corresponding ratios for NOx, CO,, and
particulates were 85. 2.4, and 44 respectively. The variability of CO, emission is relatively low because
CO, emissions are not controlled, and, unlike SO,, NOx, and particulates, the fuel pollutant (carbon)
content for those fuels used in load following plants (coal, oil, and gas) varies by only about a factor of
two.
The second primary factor determining a PV system's ability to offset emissions can be measured by the
system's average annual capacity factor. The capacity factor (calculated by dividing actual generation
for a 12-month period by the product of a system's SOC capacity and the number of hours in a year)
incorporates all factors influencing a PV system's ability to generate power such as the available solar
resource, unplanned outages, and snow cover. There was considerably less variability in the average
capacity factors of the PV systems in this project than in the emission rates of the participating utilities.
The ratio of the largest to smallest average annual capacity factor was only 2.1:1.0 (Seottsdale, AZ to
Pittsburgh, NY).
The consequence of the variability in these factors is that systems with relatively low average capacity'
factors may produce offsets well in excess of those produced by systems with very high capacity factors.
Conversely, PV systems in regions with excellent solar resources may be relatively ineffective at
mitigating pollution if regional emission rates are low. One can find several examples of this by
reviewing Table 8-19, w hich summarizes the annual offsets of each system, normalized by its respective
SOC capacity, as well as the average annual capacity factor of each system. For example, the system in
Pittsburgh, NY had the lowest annual capacity factor of the systems in the project, yet its annual S02
offset was among the highest calculated. The three Southern California Edison systems had some of the
highest annual capacity factors, but some of the lowest SO,, NOs and particulate offsets.
Analysis of the emission offsets of the simulated dispatchable storage systems indicated that utility
marginal emission rates tend to be lower, and in some cases much lower, during peak load hours than
during off-peak hours. This is due to the fact that many utilities rely on gas-fired combustion turbines
(with high operating costs but relatively low emission rates) to follow load during their highest load
hours. One clear implication of this is that modeling utility emissions with a constant average rate for all
hours (as was done for NYSEG, NYPA, NSP, and SCE) results in inaccuracies in the calculated offsets.
Actual emission rates will vary over the course of a day and over the course of a year as units used to
follow load, and the fuel used to operate those units, change. If on any given day a PV system generates
during a utility's highest load hours, the offsets calculated using constant average rates will likely
overstate the actual offsets. Alternatively, if most of the system's generation occurs during off-peak
hours, offsets will likely be greater than calculated using a constant, average rate.
8-15
-------
Table 8-19. ANNUAL OFFSETS AND CAPACITY FACTORS BY SITE
Site
Annual Offset (kg/kWS0C)
Average
Annual
Capacity
Factor
S02
NO.,
o
u
Particulates
National Average
6.4
2.6
870
0.21
0.16
Platsburgh, NY
7.0
2.1
650
0.24
0.10
Berlin, CT
7.1
2.6
870
0.25
0.15
Pleasantville, NJ
14
6.2
1,600
0.37
0.15
Brigantine. NJ
16
7.2
1,800
0.44
0.17
White Plains, NY
0.22
1.5
700
0.60
0.13
Scottsdale, AZ
1.2
3.5
2,000
0.20
0.21
Peoria, AZ
1.1
3.2
1,800
0.18
0.19
Ashwaubenon, W]
7.0
7.7
1,500
0.26
0.14
Denmark, WI
8.0
8.7
1,600
0.29
0.16
Minnetonka, MN
2.2
2.7
870
0.20
0.15
San Ramon. CA
0.33
1.5
2,300
0.06
0.1S
Austin, TX
1.2
1.7
950
0.11
0.15
Flagstaff, AZ
0.79
2.4
1.400
0.13
0.18
Barstow, CA
0.0040
0.11
800
0.02
0.17
Edwards AFB, CA
0.0048
0,13
950
0.02
0.2C
Palm Desert CA
0.004.1
0,12
900
0 70
0 1<3
A conclusion that can be drawn regarding PV system emission offsets is that using energy storage in
conjunction with a PV system will necessarily entail an appreciable penalty in the level of offsets. This
is due primarily to the energy lost in charging and discharging the battery. The storage dispatch
algorithm itself may reduce emission offsets if, as was the case in this study, the algorithm is designed to
maximize the availability of stored energy during the utility's highest load hours, when its marginal
emission rates are likely to be at a minimum. Conceptually, a loss-free storage system could be used to
maximize offsets by delaying the dispatch of PV-generated electricity until the hours at which marginal
emission rates peak. However, the energy lost in charging and discharging any real storage system
would likely reduce offsets to a degree greater than the increase in offsets that might be obtained through
the use of such a dispatch algorithm.
References
1. U.S. Department of Energy, Energy Information Administration, Electric Power Annual 1993,
Washington, D.C. (1994), Table 44, p. 74.
8-16
-------
Sulfur Dioxide
3rd Qtr '93 1slQtr'S4 3rd Qtr'94
4th Qtr'93 2nd Qlr'94
Carbon Dioxide
400
350
Q
o
w
300
o
<9
§
250
o
200
a.
»
s
1 50
s
TO
o
100
i.
50
0
-
-
r t
-
-
-
""li
1 > 1.
J
Tv'jjiJ
3rd Qtr '93 1st Qtr'94 3rd Qtr'94
4th Qir'93 2nd Qtr'94
Nitrogen Oxides
1.5
u
O
Q
*0
1.0
aj
a
£
ffl 0.5
os
o
0,0
3rd Qtr'93 1st Q[r '94 3rd Qtr'94
4lh Qtr 93 2nd Qtr 94
Particulates
3rd Qtr '93 1st Qtr'94 3rd Qtr'94
41hQtr'93 2nd Qtr'94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-2, Emission Offsets for Pittsburgh, NY (EPAOI),
Notes on System Operation:
3rd Q '93:
Inverter failure limits system to 2/3 power 7/1/93 - 7/16/93
System shut down completely 7/23/93 - 9/30/93,
4th Q '93;
Data loss 10/1/93 - 10/5/93 due to datalogger short-circuit.
1st Q '94:
System fully operational
2nd Q '94:
DC injection limited system to 2/3 power 4/1/94 - 4/25/94.
6/15/94 and again 6/18/94 - 6/30/94.
DC fuse failures limited system to 2/3 power 6/5/94 -
3rd Q '94:
System fully operational.
8-17
-------
Sulfur Dioxide
3.5
3rd Qtr'93 1st Qtr'94 3rd Otr'94
4th Qtr '93 2nd Qtr'94
Carbon Dioxide
o 100
3rd Qtr'33 1st Qtr'94 3id Qlr'94
4th Qtr '93 2nd Qtr'94
Nitrogen Oxides
1.6
O
o
»
§ i-o
a
a
«
£
g 0.5
a
o
0.0
3rd Qtr'93 tit Qtr'84 3rd Qtr '94
4th Qtr'93 2nd Qtr'54
Particulates
0
0,
o
O 0,
w
f 0
KB
5 0,
« 0
CL
1°
n
o
go.
0
0
10
09
08
07
06
05
04
03
02
01
00
3rd Qtr'S3 tit Qtr'94 3rd Qlr'94
4th Qtr '93 2nd Qtr '94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-3. Offsets for Berlin, CT (EPA02).
Notes on System Operation:
3rd Q '93:
Inverter failure prevented system operation 7/17/93 - 8/6/93,
4th Q '93:
System fully operational
1st Q '94:
System fully operational
2nd Q '94: System fully operational
3rd Q '94:
System fully operational
8-18
-------
Sulfur Dioxide
3.5
O.D
3rd Qtr '93 1st Qtr'94 3rd Qtr'94
4th Qtr'93 2nd CHr'94
Carbon Dioxide
3rd Qtr "93 1st Qtr'94 3rd Qtr 94
4th Qtr'93 2ndQtr'94
Nitrogen Oxides
1.5
0.0
3rd Qtr'93 1st Qtr'94 3rd Qtr'94
4th Qtr'93 2nd Qtr'94
Particulates
3rd Qtr'93 1st Glr S4 3rd Qtr'94
4th Qtr'93 2nd Qtr'94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-4. Emission Offsets for Pleasantville, NJ (EPA03).
Notes on System Operation:
3rd Q '93:
Inverter failure limited system to 1/3 power far 43 days. 2/3 power for 5 days.
4th Q '93:
System fully operational
1st Q '94: Inverter failure limited system to 2/3 power 1/18/94 - 3/11/94 and 3/18/94 - 3/31/94.
2nd Q '94: Inverter failure limited system to 2/3 power 4/1/94 - 4/22/94.
3rd Q 8 94: DC disconnect fuse failure limited system to 2/3 power from 7/23/94 to 8/2/94.
8-19
-------
Sulfur Dioxide
3.5
0,0
3rd Qtr'93 tstQir'94 3rd Qtr "94
4th Qtr'93 2nd Qtr'94
Carbon Dioxide
400
350
o
o
w
300
o
£
250
©
200
a.
a>
E
150
s
TO
O
100
£
50
3rd Qtr'93 1stQtr'94 3rdGtr'S4
4th Qtr'93 2nd Qtr'94
Nitrogen Oxides
1.5
O
O
«
o
AS
1.0
s
Q.
W5
E
m 0.5
m
o
0.0
3rd Qlr*93 1st Qtr 94 3rd Qtr'94
4th Qtr'S3 2nd Qtr '94
Particulates
3rd Qtr '93 1st Qtr '94 3rd Qtr '94
4th Qtr'93 2nd Qtr'94
Offsets without
Battery Storage
I
Offsets with
Battery Storage
pq National Average
^ Offset
Figure 8-5. Emission Offsets for Brigantine, NJ (EPA04).
Notes on System Operation
3rd Q '93:
DC disconnect fuse failure prevented generation 8/18/93 - 8/20/94 and again 9/15/94 - 9/20/94.
4th Q '93:
Inverter failure prevented generation 12/6/93 - 12/9/93.
IstQ '94:
System shut down 2/25/94 - 2/28/94. DC injection error prevents generation 3/5/94 - 3/17/94.
2nd Q '94: Building load meter recorded zero load 4/15/94 - 6/18/94.
3rd Q '94:
System fully operational
8-20
-------
Sulfur Dioxide
O 3
o
«
2
o
1
o
0.
0.0
, .VtljjlSsDtt,, ' ¦!
3rd Qtr '93 1st Qtr "94 3rd Qtr '94
4th Qtr'93 2nd Qtr'94
Carbon Dioxide
400
o
350
o
P|Pftg
IB jb 1
iaiilh ®3j
: 91 1 ¦¦
3rd Qtr '93 1st Qtr'94 3rd Qtr '84
4th Qtr'83 2nd Qtr'94
Nitrogen Oxides
1.5
O
0
52
<5 1-0
2
S
a.
01
E
a 0.5
o>
o
O.Q
3rd Qtr'93 fet Qtr'94 3rd Qtr'94
4th Qtr'93 2nd Qtr '94
Particulates
0
„0.
o
O 0.
52.
To
m
5 0
jt
5 o,
a
| 0,
9
5)
o
10
09
08
07
06
05
04
03
02
01
00
3rd Qtr'93 1st Qtr'94 3rdQlr'B4
4th Qtr'93 2nd Qtr'94
Offsets without
Battery Storage
I
Offsets with
Battery Storage
National Average
Offset
Figure 8-6. Emission Offsets for White Plains, NY (EPA05).
Notes on System Operation:
3rd Q "93:
Inverter failure prevented generation 7/22/94 - 8/13/94.
4th Q '93:
System fully operational
1st Q '94;
System fully operational
2nd Q '94: DC disconnect fuse failure prevented generation 6/11/94 - 6/30/94.
3rd Q '94:
System fully operational
8-21
-------
Sulfur Dioxide
3.5
Q
3.0
O
(fl
2.5
(S
5
2.0
as
a.
1 .5
03
E
a
SB
1.0
0
2
0.5
0.0
3rdQ{r'93 1st Qtr'94 3'd Qtr'94
4tHQtr'93 2ndQ!r"94
Carbon Dioxide
3rd Qtr'93 1st Qtr'84 3rd Qtr 94
4th Qtr'93 2nd Qtr'94
Nitrogen Oxides
1.5
O
O
w
S 1.0
a>
a.
u>
B n C
it) 0.5
w
ro
o
0.0
3rd Gir '93 1st Qtr '94 3rd Qtr 94
4th Qtr'93 2nd Gir '94
Particulates
3rd Qtr'93 1st Qtr'94 3rd Qtr'34
4th Qtr 93 2nd Qtr'94
Offsets without
Battery Storage
I
Offsets with
Battery Storage
National Average
Offset
Figure 8-7. Emission Offsets for Scottsdale, AZ (EPA06).
Notes on System Operation:
3rd Q '93:
System fully operational
4th Q '93:
System fully operational
1st 0 '94:
One of two inverters shut down 3/4/94 - 3/6/94. Cause unknown.
2nd Q '94:
System fully operational
3rd Q '94;
System at 1/2 power 7/18/94 - 7/20/94 due to DC disconnect fuse failure.
8-22
-------
Sulfur Dioxide
3.S
3.0
6"
o
« 2.5
o
S 2.0
CM.5
(0
es 1.0
o
'<£.
0.5
0.0
3rd Qtr '93 WQir'94 3rd Qtr'94
4IhQtr'93 2nd Qtr'94
Carbon Dioxide
3rd Qtr'93 fetOtr'34 3rd Qtr'94
4th Qtr '93 2nd Qtr '94
Nitrogen Oxides
1.5
0.0
3rd Qtr "93 1st Qtr 94 3rd Qtr'94
4th Qtr'93 2nd Qtr'94
Particulates
0.
0
o
O 0
«
o o
a
S o
J£
o 0
a.
1°
m
o
10
09
08
0?
06
05
04
03
02
01
00
3rd Qtr '93 1st Qtr '94 3rd Qtr '94
4th Qtr'93 2nd Qtr '94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-8. Emission Offsets for Peoria, AZ (EPA07).
Notes on System Operation:
3rd Q '93:
Data collection began 9/1/93.
4th Q '93:
System fully operational
1st Q '94:
System fully operational
2nd Q "94: System fully operational
3rd Q '94:
System fully operational
8-23
-------
Sulfur Dioxide
3.5
O3
O
C/3
w 2,
o
ra
2/94 4/94 6/94 3/94 10/94 12/94
Carbon Dioxide
2/94 4/94 6/94 8/94 10/94 12/94
Nitrogen Oxides
1.5
O
O
2.
« 1-°
©
CL
to
E
2 0,5
O)
o
0,0
o n
o o
0 0
eg
5 0
JC
5 0
a.
1 0
a
os 0
o
2 o.
10
09
08
07
06
05
04
03
02
01
00
2/94 4/94 6/94 8/94 13/94 12/94
Particulates
2/94 4/94 5/94 8/94 10/94 12/94
Offsets without
Battery Storage
Offsets with
Battery Storage
Cj|j National Average
Offset
Figure 8-9. Emission Offsets for Flagstaff, AZ (EPA 13).
Notes oil System Operation:
1st Q "94: System fully operational
2nd Q '94: System fully operational
3rd Q '94: System fully operational
4th Q '94: System shut down 11/28/94 - 12/1/94 for unknown reason.
8-24
-------
Sulfur Dioxide
3.5
3rd Qtr'93 fetQtr'94 3rd Of "OA
4th Qtr '93 2nd Qtr'94
Carbon Dioxide
400
350
O
52.300
o
CO
250
200
a.
v>
E 150
©
u>
o 100
50
0
3rd Qtr 93 1st Qir '94 3rd Qtr "94
4th Qtr 93 2nd Qtr'94
Nitrogen Oxides
1.5
0.0
3rd Qtr'83 1st Qtr'94 3rd Qtr'94
4th Qtr "93 2nd Qtr 34
Particulates
3rd Qtr '93 feiQtr'94 3rd Qtr'94
4th Qtr'93 2n
-------
Sulfur Dioxide
3rd Qlr'93 fei Qtr'94 3rd Qtr'94
4thQtf93 2nd Qtr'94
Carbon Dioxide
400
3rdQtr'93 fei Qtr'94. 3rd Qtr'94
4th Qtr'93 2nd Qtr'94
Nitrogen Oxides
1.5
0.0
3rd Qtr'93 tit Qtr '94 3rd Qtr '94
4th Qtr'93 2nd Qtr'94
Particulates
3rd Qtr '93 1st Qtr '94 3rd Qtr '94
4th Qtr '93 2nd Qtr '94
j | Offsets without | Offsets with rra National Average
* * Battery Storage ¦ Battery Storage Offset
Figure 8-11. Emission Offsets for Denmark, WI (EPA09).
Notes on System Operation:
3rd Q '93:
System shut down due to inverter failure 7/12/93 - 7/26/93, 9/7/93 - 9/8/93. Building load not monitored until
11/4/93.
4th Q '93:
Building load monitoring begins 11/4/93.
1st Q '94:
System fully operational
2nd Q '94: Inverter shut down 5/16/94 - 5/20/94 due to DC injection.
3rd Q '94:
System fully operational
8-26
-------
Sulfur Dioxide
3.5
Q
3.0
o
t/3
2.5
o
<0
£
2.0
«
a.
«
1.5
E
IB
~s 1,0
0
5
0.5
0.0
3rd Qtr *93 1st Qtr'94 3rd Gtr'94
4th Qtr '93 2nd Qtr'94
Carbon Dioxide
3rd Qtr'93 tetQtr'94 3rd Qtr'94
4th Qtr'93 2nd Qtr'94
Nitrogen Oxides
1.6
O
o
2>
11.0
a
m
E
m 0.5
u>
o
0.0
3rd Qtr S3 "teiOtr'94 3rd Qtr'54
4ih Qtr"93 2nd Qtr '94
Particulates
0
u
O 0,
«
0 0
a
S 0
© 0
a
1 °
«
5.0
o
o.
0
0
10
09
03
07
06
05
04
03
02
01
oo
3rd Qtr '93 1st Qtr '94 3rd Qtr '94
4th Qtr'93 2nd Qtr '94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-12. Emission Offsets for Mirinetonka, MN (EPA 10).
Notes on System Operation:
3rd Q '93:
System fully operational
4th Q *93:
System off-line 10/1/93 - 10/12/93 due ta inverter failure.
1st Q '94:
System fully operational
2nd Q '94: System fully operational
3rd Q '94:
System fully operational
8-27
-------
Sulfur Dioxide
Q 3
O
£0
r2
(B
5-
.* 2
©
E
«
® 1
o
3.5
0
5
0
5
0
0IWIIS
0.0
1st Qtr'94 3rd Qtr"S4
2nd Qlr'84 4th Q!r'94
Carbon Dioxide
1st Qtr'94 3rd QIr'94
2nd Qtr'94 4th Qtr'94
Nitrogen Oxides
1.5
U
O
w
O -
m 1 •
v
a.
m
E
2 0.
Ui
©
0.0
1st Qtr '94 3rd Qtr'94
2nd Qlr '94 4th Qtr'94
Particulates
0
o
0 0.
M
To
to
§ 0
J*
S o.
Q.
1 0.
31
O
2 o.
0
o
10
09
08
07
06
05
04
03
02
01
00
"1st Qtr '94 3rd Qtr '94
2nd Qlr'94 4th Qtr '94
Offsets without
Battery Storage
Offsets with
Battery Storage
National Average
Offset
Figure 8-13. Emission Offsets for San Ramon, CA (EPA11).
Notes on System Operation:
1st Q "94: PV system data acquisition commences 1/7/94. Load measurement commences 3/7/94. System shut down 2/11/94 - 2/14/94.
2nd Q '94: System operated at 2/3 power 5/14/94 - 5/27/94 due to inverter failure.
3rd Q '94: One string disabled after testing. System limited to about 80% 7/1/94 - 8/11/94.
4th Q '94: System operated at 2/3 power 10/19/94 - 10/24/94 due to inverter failure.
8-28
-------
Sulfur Dioxide
a
o
52.
o
<9
3.5
3.0
"2.5
2 2.0
1V93 »94 3/94 5/94 7/94 9/94
Carbon Dioxide
400
350
o
o
w
300
o
co
5
250
JaC
w
O
200
Q.
«
E
150
CO
u>
o
100
X.
50
1V93 V94 3/94 5/94 7/94 9/94
Nitrogen Oxides
1.5
O
o
52.
n 1.0
42
a
0)
E
2 0.5
05
O
0.0
¦V93 V94 3/94 5/94 7/94
Particulates
0,
0.
o „
o o.
»
is
5 o.
o 0.
a.
09 0
o
2 o
o
0
10
09
08
or
06
05
04
03
02
01
00
1V93 -V94 3/94 5/94 7/94 9/84
| Offsets without | Offsets with F3 National Average
j ' Battery Storage ' Battery Storage Offset
Figure 8-14 Emission Offsets for Austin, TX (EPA12).
Notes on System Operation:
4th Q "93:
Properly functioning PV meter not installed until 3/14/94, FV generation prior to that date was simulated using
irradiance data
1st Q '94:
Properly functioning PV meter not installed until 3/14/94. PV generation prior Co that date was simulated using
irradiance data.
2nd Q '94: System at 1/3 power 5/11/94 - 5/25/94 due to inverter failure. off-line6/11/94 - 6/14/94 due to fuse failure.
3rd Q '94: System operated at 2/3 power for total of 12 days due to fuse failure.
8-29
-------
Sulfur Dioxide
Carbon Dioxide
400
350 -
O
o
M 300 -
o
g 250 -
5 200 -
a
«
£ 150
<0
6/94 8/94 10/94 "E/94 2/95 4/95
Nitrogen Oxides
1.5 —
S2.
£ 10
5
JC
2 0,5
Oi
o
6/94 8/94 13/84 G/94 2/95 4/95
Particulates
n Offsets without i Offsets with nn National Average
" Battery Storage ' Battery Storage Offset
Figure 8-15. Emission Offsets for Barstow, CA (EPA 14).
Notes on System Operation:
3rd Q "94: System shut down 7/22/94 - 8/3/94 due to DC disconnect fuse failure. Shading reduces output by 30 to 40 percent.
4th Q "94: System fully operational.
1st Q *95: System fully operational.
2nd Q '95: Work on local distribution circuit causes inverter to operate intermittently 10/25/94 - 6/10/94.
8-30
-------
Sulfur Dioxide
3.5
400
_ 350
O
O
» 300
| 250
® 200
2/94 4/94 6/94 8/94 10/94 12/94
Carbon Dioxide
2/94 4/94 6/94 8/94 10/94 12/94
Nitrogen Oxides
1,5
O
o
m 1-0
©
CL
2 0.5
a>
o
0.0
,,
~y' |j|h ;jL
'iiiilin iihiiiil ¦niiiiiiinr r,m • —
2/94 4/94 S/94 8/94 O/94 W 94
Particulates
0.
0.
o„
O o.
tfl
o 0.
-------
Sulfur Dioxide
3.5
3.
0
5
0
5
0
0.5
0.0
o
O
CO
o
eg
©
CL
« ^
e
S
OS 1
o
400
— 350
O
o
W 300
o
to
250
2/S4 4/94 6/S4 8/94 D/94 "E/94
Carbon Dioxide
Nitrogen Oxides
1.5
O
o
S
o
a
s
1.0
o
a.
2 0.5
a
o
0.0
2/94 4/94 5/94 8/94 10/94 12/94
Particulates
0
o „
o 0
2.
0 0
n
5 0.
JC
® 0.
CL
1 0
2
5> o
o
Si o.
o
0
10
09
OB
07
06
05
04
03
02
01
00
2/94 4/94 6/94 8/94 C/94 E/94
2/94 4/94 6/94 8/94 D/94 12/94
Offsets without
Battery Storage
I Offsets with pp National Average
Battery Storage Offset
Figure 8-17. Emission Offsets for Palm Desert, CA (EPA 16).
Notes on System Operation:
1st Q '94: Data acquisition commences 2/1/94. System fully operational.
2nd Q '94: System fully operational
3rd Q '94: System fully operational
4th Q '94: System fully operational
8-32
-------
Chapter 9
Conclusions
9.1 PV System Hardware
Many lessons were learned during the process of installing the 16 systems for this project. Desirable
modifications became apparent which had the potential to reduce costs, simplify installations, and reduce
field labor.
9.1.1 System Design
The overall system design resulted in an effective combination of FV array and power conditioner, from
the standpoints of current, voltage and power matching. Specific conclusions and recommendations to
aid future system design are identified below.
a) The low profile and shallow tilt angle of the PV array result in snow accumulation and
retention for extended periods on flat-roof installations in the Northeast and Great Lakes
regions. This roof-hugging design was primarily prompted by consideration for wind
loading of the ballast-mounted PV arrays. Access beneath the PV array is extremely
limited, also due to the low profile and shallow tilt angle. This proved inconvenient
when wiring the PV panels. Increasing the tilt angle and raising the PV array will
improve access and snow shedding, at the expense of increased wind loads.
b) Wind loading analysis of the ballast-mounted PV arrays remains an area in need of
clarification. Existing analyses do not specifically address the situation presented by the
ballast-mounted PV array and may or may not be adequate or accurate for assessment of
these installations. A comprehensive study of wind loading and ballast-mounted PV
arrays is highly recommended.
c) No sites in this project are fully compliant with the National Electrical Code. Full
compliance with the National Electrical Code was not possible when these systems were
installed, but is now achievable. The primary reasons for non-compliance are: (1) the
National Electrical Code requires that equipment be listed by Underwriters Laboratory
or other recognized authority, but the inverters, source circuit protectors and connectors
in the systems were not listed at the time of these installations, and (2) PV array open-
circuit voltage, calculated from PV module rating labels, can exceed 600 volts dc (at
Standard Test Conditions). Subsequent to the installation of these systems, Sandia
National Laboratory supported Ascension Technology in a project to submit all
necessary components in these EPA PV-DSM systems to UL for investigation and
listing. A year later all are now listed components, although field retrofitting of listing
labels is not possible due to configuration changes made in the hardware.
9-1
-------
d) Residential rooftops rarely have sufficient south-facing surface area to accommodate a
4kW PV array. Future residential systems will either be designed with a smaller array or
candidate homes will be identified in advance.
9.1.2 Power Conditioning
Twenty-nine 4-kW power conditioners from Omnion Power Engineering were installed for this project
and a myriad of problems developed, as described in chapter 4 and appendix D. To its credit, Omnion
has been responsive in dealing with these problems. They have implemented design changes as required,
established and maintained a 2-day turn-around policy on unit repairs, and extended their product
warranty one additional year for the EPA installations. Omnion's inverters are now UL listed.
a) The two-line display on the front cover of the Omnion inverters has proven useful for
debugging system problems. The display indicates dc voltages and ac power output
during normal operation, and provides various fault messages in the event of utility or
inverter problems. The newest model Omnion inverters make this information available
over a serial communications line for remote datalogging.
b) Inverters that can withstand the outdoor environment are highly desirable. Units were
mounted outdoors at several sites in California, because indoor locations were not
available. Locating the inverter outdoors is desirable in many situations, since this keeps
inverter acoustic noise away from living and working spaces, and the installation can
proceed without entering the building. Omnion's newest power conditioners are
designed for outdoor mounting and operation in virtually all climates.
9.1.3 Mounting Hardware and BOS
Ascension Technology designed and supplied all PV array mounting hardware and wiring balance-of-
systems components. Specific conclusions and recommendations are found below.
a. The RoofJack mounting system for flat-roof buildings provided a convenient means for
installing the PV panels. Installation of the PV panels proved to be the quickest of the
installation steps.
b. The ballast trays were designed to be laid out in long rows, with trays abutting. This can
form a dam that will impede the normal flow of water on the roof and is undesirable on
most roofs. Subsequent to this project, array design has been modified to use trays
rotated to their portrait orientation, so that no two trays are touching.
c. Although the ballast tray system was designed for use on EPDM ballasted roofs, roofing
manufacturers of several other types of roofs have approved this system for use on their
roofs.
d. RoofJacks were installed on the ballast trays using truss-head fasteners inserted through
the tray bottoms. This was a significant inconvenience in the field. Since this project
the ballast trays have been redesigned with pressed-in threaded studs; RoofJacks have
clearance holes to fit over the studs. This way the trays can be laid out and the
RoofJacks attached afterward.
e. The PV module pin assemblies (which fit into slots in the Roofllacks) were attached in
the field rather than the factory, where this work could have been completed more easily.
9-2
-------
In subsequent projects, mounting pins have been attached to module frames prior to
deliver)',
f. During installation of the first system, it became apparent that there was insufficient
clearance in the RoofJacks pipe nipple for the quick-connectors to fit. In redesigned
RoofJacks, the pipe nipple is larger to provide extra clearance,
g. The source-circuit protector is a useful field wiring interface. The source-circuit
protectors have been redesigned since this project; they are now larger to allow more
wiring room inside, have a new terminal block and a new lid to prevent water intrusion
more effectively. The Source Circuit Protectors are now UL listed.
9.1.4 Instrumentation
The Ascension Technology Rotating Shadowband Pyranometer was installed at each PV system site,
with additional transducers to monitor plane-of-array irradiance. building load and PV system AC
output. These Campbell Scientific-based units have proven to be highly reliable and useful in many-
ways for this project.
a. Data from the shadowband has been used to model PV system performance as a means
for assessing system condition; if the measured performance varies appreciably from the
model predictions, then a problem may exist.
b. Utilities at two sites opted to add extra transducers to their data acquisition systems to
monitor PV array DC parameters. This provides the ability to develop and validate PV
array models, useful both for diagnosing problems and also for addressing questions of
PV system operating voltages to help plan future revisions of the National Electrical
Code, Consideration should be given to adding dc instrumentation at one or more sites
in future projects.
c. Ascension Technology has relied upon electric utilities to provide pulse-initiating
kilowatt-hour meters at each site, to monitor PV system AC output and building load. In
many cases there have been significant delays in the installation of these meters and
confusion regarding the calibration constants applied to their pulse outputs. In future
projects, the pulse-initiating kilowatt-hour meter, or an alternative transducer for
measuring PV system output, may be provided by Ascension Technology.
9,2 PV System Operation and Maintenance
The weakness of grid-connected PV systems is clearly the power conditioners. There has not been a
single PV module failure among the 2,436 installed in this project. The Siemens Solar PV modules used
in this project are UL listed and have been a standard, workhorse product for the company for many
years. The other components in the system — mounting hardware, disconnect switches, source circuit
protectors -- are similarly robust. The following conclusions may be drawn regarding PV system
operation and maintenance:
1, Data acquisition and monitoring of PV system demonstration projects is a prudent measure to
help insure their continued and proper operation.
2. Preventative maintenance is not required for grid-connected PV systems.
9-3
-------
3. For PV projects it is essential to identify a site contact, who is familiar with the PV system and
can provide on-site support during periods of system monitoring.
4. Spares of small parts, such as fuses in the power conditioner and disconnect switches, should be
maintained on-site for rapid replacement when needed.
9.3 Host Building Load Impacts
Two general conclusions may be drawn from the results of this study. The first is the relatively self-
evident conclusion that if reduction of customer peak loads is the primary motivation for the installation
of a PV system, it is critical to investigate the correlation of building peak loads to solar irradiance. The
set of host buildings participating in this project included some with loads which were very well matched
to the solar resource as well as some for which the match was very poor. The systems in Ashwaubenon,
WI and Scottsdale, AZ are examples of systems which reduced host building load duration curves
(LDCs) by a substantial fraction of their rated capacity. The highest loads in these buildings occurred
during the midday hours, when the solar resource peaks. The systems in Barstow, CA and Denmark, WI,
on the other hand had very little effect on the host building's LDC, despite ample solar resource. Many
of the highest building loads at these sites occurred near or after sunset.
The second general conclusion to be drawn from the data is that the generation of a PV system during an
individual building's peak load hour provides little information regarding that system's ability to reduce
the building's peak monthly load, or to reduce demand charges. The reason for this is that the PV system
will not be operating at the same power level during all of the highest load hours in a month. If there are
hours for which the gross load level is close to the monthly peak, but during which the PV output is less
than that during the peak hour, net loac for these hours may exceed net load during the hour at which the
gross load attains its monthly peak level. For example, consider a building with a monthly peak load of
300 kW occurring at noon and a second-highest load of 299 kW occurring at 10:00 p.m.. Even though a
10 kW PV system might reduce the building's load by its full rating during the peak load hour, it will not
be generating at all during the second highest load level for the month. The net change in the building's
LDC will be only 1 kW because the hour which was the second-highest load hour without PV system
operation has become the peak load hour. In cases such as this, there may be very little change in the
building's net LDC and correspondingly small changes in demand charges. The monthly peak load will
have simply been shifted to another hour.5
9.4 Utility Coincident Peak Load Reduction
9.4.1 Load Reduction Without Storage
Load matching for PV systems installed in northern States is greatest in the spring and summer months,
with the capacity factor during the highest load hours typically averaging above 40 percent. Several of
these sites acheived capacity factors well in excess of 60 percent of their SOC rating during the highest
load hours in these months. The northern systems invariably generated little or no power during winter
peak hours, most or all of which occurred at night.
5Of course, this may be precisely the intent. Many commercial and industrial electricity consumers face rate structures which strongly discount
charges for peak loads that do not occur during utility-defined peak periods. Usually, these peak periods arc during daytime hours when the PV
system would be capable of reducing building load.
9-4
-------
Utility peak loads in the southern and western parts of the country invariably occurred during the
summer months when the solar resource is greatest, although these peaks consistently occurred in the
mid- to late-afternoon. Most of the systems installed in these regions operated at capacity factors in
excess of 40 percent during the highest load hours in the summer months. Some systems consistently
operated at capacity factors above 60 percent during these hours. The one exception to this is the system
in Flagstaff, which operated at only about 30 percent capacity factor during the peak load hour. This low
result is most likely explained by the fact that the load and weather patterns in Flagstaff are quite
different from those in Phoenix which is about one mile lower in elevation, Loads in the Phoenix area
probably dominate the Arizona Public Service system load.
As did their counterparts in the Midwest and Northeast, systems in the southern and western states
typically operated at a lower level during winter peak hours. With the exception of the systems in
southern California, systems in the West operated at or near zero percent capacity' factor during peak
hours in the first quarter of the year.
9.4.2 Load Reduction With Storage
Except where the power output was limited by a system outage, results from the storage simulation
indicate that storage can provide system operation at the full inverter rating during the peak load hours in
the summer months at all sites. In regions (such as the Northeast Utilities service area) where peak
utility loads are highly correlated to the solar resource, the addition of a dispatchable storage system may
do little to improve the PV system's load matching capability, since it will already be quite good.
Systems in northern states are much less able to provide power during peak load hours in winter due to
the limited solar resource and snow cover. Even with storage, some of these systems were unable to
provide power at more than a few percent of inverter rating during winter peak load hours. However,
daytime generation at other northern sites was sufficient to allow inverter operation well in excess of 50
percent of inverter rating during winter peak hours.
Unlike many of the systems installed in northern climates, the addition of dispatchable storage to
systems installed in the southern and western states would allow them to operate at high capacity factors
during winter peak hours. The results of the simulation indicate that most of the systems installed in this
part of the country would operate at or near 100 percent of inverter rating during the highest winter load
hours.
It is important to recognize that these results are substantially determined by the storage
charging/dispatch algorithm. An algorithm which stores generation from one or more days and
dispatches only when load exceeds a predetermined threshold (as opposed to dispatching during the peak
hours of each day), would substantially improve the load matching characteristics of all systems.
9.5 Emissions
Two primary factors contribute to determine the emission offsets that a PV system will provide in any
given location: the marginal emission rates of the utility system into which it is connected and the solar
resource available for conversion to electricity. The data collected through this project has indicated that
these two factors combine to create tremendous variability in the emission offsets resulting from the
operation of these systems.
9-5
-------
Because marginal emission rates are dependent upon the pollutant content of the fuel used in load
following generators, the combustion temperature and operating efficiency of those generators, and the
presence or absence of pollution control equipment, these rates exhibited considerably greater variability
between utilities than did the solar resource. The ratio of the highest average S02 emission rate
(determined by dividing the total S02 offset over a 12-month period by a PV system's total generation
over that period) to the lowest such rate was nearly 4,000. Corresponding ratios for NOx, CO,, and
particulates were 85. 2.4, and 44 respectively. The variability of CO, emission is relatively low because
C02 emissions are not controlled, and, unlike SO,, NOx, and particulates, the fuel pollutant (carbon)
content for those fuels used in load following plants (coal, oil, and gas) varies by only about a factor of
two.
The second primary factor determining a PV system's ability to offset emissions can be measured by the
system's average annual capacity factor. The capacity factor (calculated by dividing actual generation
for a 12-month period by the product of a system's SOC capacity and the number of hours in a year)
incorporates all factors influencing a PV system's ability to generate power such as the available solar
resource, unplanned outages, and snow cover. There was considerably less variability in the average
capacity factors of the PV systems in this project than in the emission rates of the participating utilities.
The ratio of the largest to smallest average annual capacity factor was only 2.1:1.0 (Scottsdale, AZ to
Pittsburgh, NY).
The consequence of the variability in these factors is that systems with relatively low average capacity
factors may produce offsets well in excess of those produced by systems with very high capacity' factors.
Conversely, PV systems in regions with excellent solar resources may be relatively ineffective at
mitigating pollution if regional emission rates are low. One can find several examples of this by
reviewing Table 8-19, which summarizes the annual offsets of each system, normalized by its respective
SOC capacity, as well as the average annual capacity factor of each system. For example, the system in
Pittsburgh, NY had the lowest annual capacity factor of the systems in the project, yet its annual SO,
offset was among the highest calculated. The three Southern California Edison systems had some of the
highest annual capacity factors, but some of the lowest S02, NOx and particulate offsets.
Analysis of the emission offsets of the simulated dispatchable storage systems indicated that utility
marginal emission rates tend to be lower, and in some cases much lower, during peak load hours than
during off-peak hours. This is due to the fact that many utilities rely on gas-fired combustion turbines
(with high operating costs but relatively low emission rates) to follow load during their highest load
hours. One implication of this is that modeling utility emissions with a constant average rate for all
hours (as was done for NYSEG, NYPA, NSP, and SCE) may result in inaccuracies in the calculated
offsets. Actual emission rates may vary over the course of a day and over the course of a year as units
used to follow load, and the fuel used to operate those units, change. If on any given day a PV system
generates during a utility's highest load hours, the offsets calculated using constant average rates will
likely overstate the actual offsets. Alternatively, if most of the system's generation ocurrs during off-
peak hours, real offsets will likely be greater than calculated using a constant, average rate.
A conclusion that can be drawn regarding PV system emission offsets is that using energy storage in
conjunction with a PV system will necessarily entail an appreciable penalty in the level of offsets. This
is due primarily to the energy lost in charging and discharging the battery. The storage dispatch
9-6
-------
algorithm itself may reduce emission offsets if, as was the case in this study, the algorithm is designed to
maximize the availability of stored energy during the utility's highest load hours, when its marginal
emission rates are likely to be at a minimum. Conceptually, a loss-free storage system could be used to
maximize offsets by delaying the dispatch of PV-generated electricity until the hours at which marginal
emission rates peak. However, the energy lost in charging and discharging any real storage system
would likely reduce offsets to a degree greater than the increase in offsets that might be obtained through
the use of such a dispatch algorithm.
9-7
-------
Appendix A. Quality Assurance Project Plan
PHOTOVOLTAIC DEMAND-SIDE MANAGEMENT DEMONSTRATION
Contract 68-D2-0I48
QUALITY ASSURANCE PROJECT PLAN
December 18, 1992
Prepared for
Judith S. Ford
Quality Assurance Manager
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Prepared by
Ascension Technology, Inc.
P.O. Box 314
Lincoln Center, MA 01773
natures: Philip J. Bolduc
Judith S. Ford
Date:
Date:
A-1
-------
1.0 PROJECT DESCRIPTION
1.1 Abstract
The goal of this program is to document the fuel consumption, pollution mitigation and electric
power generation dispatch impacts of operating utility-interconnected solar photovoltaic power systems on
residential and commercial building rooftops. This will be accomplished using case studies of photovoltaic
systems in ten different cities and operating characteristics of ten electric power companies from New York
to California.
Ascension Technology, Inc. (ATI) will conduct a three-phase program to accomplish the goals of
the Enviromental Protection Agency (EPA).
In Phase 1, Ascension Technology will complete the design of the photovoltaic systems to be used
in this study. The systems will be comprised of photovoltaic modules6 provided by Siemens Solar Industries
of Camarillo, CA and dc-to-ac power converters7 provided by Omni on Power Engineering Corporation of
Mukwonago. WI. Equipment orders for modules and power converters will be/have been placed during
Phase 1 to allow for Phase 2 to begin in early 1993. Building selection criteria will be developed in
conjuction with the EPA, its advisory team and the partner utilities. Specific buildings will be recommended
to the EPA by ATI. A testing plan will be developed and appropriate instrumentation hardware will be
procured and assembled.
In Phase 2, Ascension Technology will fabricate the remaining balance-of-system equipment, install
the photovoltaic systems and performance monitoring instrumentation and execute the Test Plan. The
partner utilities will assist by providing air emissions and generator operations data.
In Phase 3, Ascension Technology will prepare a final report, including recommendations.
1,2 Data Quality Objectives
The fundamental objectives for data quality for this project are that proposed uncertainty limits (see
section 5) are upheld and that the data set be continuous. The proposed monitoring period will begin on
6/01/93 and end on 5/31/94. The primary data sets to be recorded by ATI are the following:
(a) Energy produced by the PV system (EPV)
(b) Energy load of the host building (EL)
In order to provide continual verification of the performance of the PV array and the metering data
acquisition system, ATI will install both a plane-of-array irradiance (IPOA) pyranometer and a Rotating
Shadowband Pyranometer (RSP). Measured IPOA will allow for checking of array performance through the
use of power conditioner and array simulation algorithms. IP0A estimated with the RSP will provide
verification of measured IP0A. For details on measurement verification, please refer to sections 3 and 8.
3 -3 Intended Use of the Data
Data reduction will be used as the primary tool for achieving the research goals of this project.
Summaries of PV system generation, fuel and emissions displacement, and building and grid load profiles
will provide the basis for conclusions regarding the environmental and utility operation impacts of using PV
systems as a DSM generating option,
f E:gh:y-fcur modules will be configured into itar.dttd PV arrays rated a! 4 45 kW dc C i ,000 W-rrr plane-of-array inadiaice. 25 C eel! lemperslurel
Rated at 4 kVV ac
A-2
-------
2.0 PROJECT ORGANIZATION AND RESPONSIBILITIES
Data instrumentation systems will be procured, installed, and monitored by Philip J. Bolduc
(Systems Engineer, ATI). Mr, Bolduc will also be responsible for data retrieval, software processing and
data presentation. He will perform these tasks under the guidance of the Project Technical Manager, Dr.
Edward C. Kern, Jr. (President, ATI).
All ATI personnel will be contributors to the final project report.
3.0 DATA QUALITY INDICATORS (DQIs)
3.1 Data Continuity
The frequency and duration of data gaps will affect the usefulness of the data for long-term analysis.
Incomplete data sets may miss PV system generation peaks and maximum utility loads in addition to
compromising the accuracy of monthly averages. Daily data summaries will identify the data recovery
period for the previous day (15-2400) as well as the number of records retrieved (96). Data loss may occur
in any of three ways:
(a) Loss of EPV data due to meter malfunction
(b) Loss of E, data due to meter malfunction
(c) Loss of IP0A data due to pyranometer malfunction
(d) Loss of RSP data due to head unit malfunction
(e) Loss of all data due to datalogger or power supply malfunction
If event a) occurs, E,>v estimated from measured I,.0A will be substituted. In the event of scenario (b),
El will be calculated as the average of 5. values from the same day of the week from the 4 surrounding
weeks. If (c) occurs, IP0A will be taken from RSP simulated data. Situation (d) will mean a loss of the IP0A
DQI specified in 3.2, but will not incur any interruption in the primary data sets. If (e) occurs, EPV and EL
will be replaced according to the procedure outlined for EL in scenario (b) above.
Records of interruptions in data continuity will be maintained at the ATI office. Each site will have
a contact person who will be of service to ATI if any basic remedial action is needed. Major system
problems will be accompanied by a site visit from ATI personnel. Data recovery is expected to exceed 98
% (358 days out of 365).
3.2 Ln, and En»: Measured vs. Calculated Bias and RMS Errors
Fifteen minute averages of measured plane-of-array irradiance and array energy output will be
checked against estimated values derived from ATI RSP processing algorithms. On both daily and monthly
bases, the bias and root-mean-square (RMS) errors (pP0A, PpV RMS^0A, RM§,V) between measured and
estimated quantities will be calculated according to procedures outlined in section 8.
The primary function of IP0A data will therefore be to provide EPV data verification and a means for
calculation of PV system efficiency. ATI will work on improving the IP0A and FPV estimation code,
originally written based on the methods in PVFORM8. The opportunity provided by this project for further
development of the simulation software will eventually allow ATI to perform a quantitative uncertainty
analysis of its PV simulation package.
The viability of the two bias errors as effective DQIs is very secure. Spot shading would affect a
large array surface and a point source measurement entirely differently. The likelihood of two independent
gD F.Memcuccr, J P Fernandez. "User's Manual for PVFORM A Phoiovoltaic System S.mulaticm Pragran For Sland-Alone and Gnd-InieracLve Applications' Sar.dia National
Laboratories April 198S.
A-3
-------
measurement schemes, one a watt-hour meter and the other a photocell and shunt resistor, to malfunction
concurrently and produce identically erroneous data is exceptionally low.
3.3 PV System Efficiency
Thorough analysis of the correlation between measured and estimated EPV will in effect be a check
on system efficiency. Estimated and measured values of EPV will only agree if system efficiency
assumptions are upheld. Nevertheless, a daily calculation of system efficiency based on measured IPOA and
EPV will be performed in order to provide secondary verification of proper system operation. Details on
the efficiency calculation algorithm wi'l be presented in section 8.
4.0 SAMPLING PROCEDURES FOR CRITICAL MEASUREMENTS
4.1 Er.. and E, Measurement
EPV and EL will be measured with General Electric IW-70-IS or VW-64S kilowatt-hour meters with
type D-72 pulse initiators (or comparable Sangamo or Westinghouse meters) for single and three phase
systems, respectively. Meters will be installed by the host utility according to standard utility metering
procedures. High resolution pulse per revolution ratios will be selected for all metering, A Campbell
Scientific, Inc. (Logan, Utah) CR10 datalogger will be used as the primary on-site data acquisition tool. For
information on meter calibration, see section 5.
4.2 Measurement
IP0A will be measured with a LI200SZ pyranometer (LiCor Corporation, Lincoln, ME). The
datalogger will measure millivolt signals produced by the pyranometer photocell and a shunt resistor.
Multiplication by a calibration factor (CF) W/m2/mV will produce W/rh . This calibration factor is
determined using Ohm's Law and the calibration coefficient (CC) |iA/(kW/m2) supplied by LiCor for each
pyranometer.
CF = 10< W/r"2
R X CC mV
4.3 Solar Resource Measurements
A standard ATI Rotating Shadowband Pyranometer system will be installed at each site to monitor
temperature and provide data support for the PV systems (DQTs, etc). ATI RSPs measure global horizontal,
direct normal and diffuse horizontal irradiance as well as temperature. With the potential refinements in
simulation code expected through the duration of this project, RSP data would enable analyses of alternate
array configurations. Further investigations into the relationship between insolation, temperature and
building and system load will be possible with the data sets provided by the RSP.
4.4 Datalogger Operation
Measurements of the primary quantities EPV and EL will be performed once per second. RSP global
horizontal and calculated direct normal readings are taken every second and diffuse measurements are taken
once per minute. Every 15 minutes the datalogger will log a record in its final storage area in the following
format;
A ~4
-------
Table 1
Datalogger Record Format
DC
Datalogger Code
JD
Julian Day
HM
Hour-Minute at End of Interval
GH
Average Global Horizontal Irradiance During Interval
DN
Average Direct Normal Irradiance During Interval
DH
Average Diffuse Horizontal Irradiance During Interval
T
Average Temperature During Interval
EPV
Total PV Array Energy During Interval
fpoA
Average Plane of Array Irradiance During Interv al
el
Total Building Load During Interval
Fifteen minute data intervals will provide compatibility with utility system data.
4.5 Instrumentation Maintenance
Instrumentation systems have been designed to require minimal maintenance. Pyranometers will
be cleaned once per month by a contact person designated for each site. ATI will call each contact person
every month to check on the appearance of the system and to confirm that cleaning has been performed. In
the event of a system malfunction, ATI will work with the contact person to determine the problem and
remedy the situation. If necessary, ATI will arrange for a personal site visit.
A-5
-------
5.0 ANALYTICAL PROCEDURES
5.1 Uncertainty Analysis
Table 2 lists the uncertainty specifications for instrumentation components. Meter uncertainties were
provided by General Electric; if alternative meters are used, their specifications will be provided.
Uncertainties associated with LI2O0SZ pvranometers (IP04 and RSP data) are given in accordance with a
1992 National Renewable Energy Laboratory review of the ATI RSP system9. Uncertainty specifications
shall be requested of the utilities for their data, although we do not expect to be able to exercise control over
their measurement procedures.
The NREL RSP review encompassed a detailed breakdown of the sources of uncertainty in RSP and
raw LI200SZ data. Final conclusions (p.60) indicated that if averaging intervals greater than 5 minutes are
used, RSP data is within the uncertainty limits of the NREL Baseline Monitoring System (BMS), i.e. ± 4.03
% for global horizontal and direct normal and ± 4.80 % for diffuse horizontal. The specification for global
horizontal irradiance measurements are directly applicable to IP0A measurement uncertainty, as they are
subject to the same calibration, alignment, temperature, spectral and datalogging sources of error.
NREL BMS uncertianty specifications are given according to the U9S model, where measured values
lie within the uncertainty window about the true value 95 % of the time.
U„ = sj$2 + (2 Sf
where P is the bias error and S is the precision error.
J StotTel, C Ricrdaii. and J Bigger "Evaluation of Solar Raaiatign Measurement Systems EPRUNREL Final Test Report". Volume 1. NR.EL-TP-463-4771 November 1992 The
specifications on the LiCor pyranometer were chosen from this report in lieu of manufacturer's specrfica-ions due to an improved uncertainty characterization
A-6
-------
Table 2
Uncertainty Specifications for Critical Measurements
Epv
PV Array Energy
Watthour Meter10
± 0.5 %
El
Building Load
Watthour Meter5
± 0.5 %
'pOA
Plane of Array Irradiance
LI200SZ
± 4.03 %
Meters will produce switch closures which are not susceptible to datalogger analog measurement
uncertainty. High pulse/revolution ratios and extended (15 min) data intervals will ensure precise energy
totalizing by the datalogger. Calculated values of IP0A and EPV will be expected to satisfy ± 10 % uncertainty
at the commencement of the project.
5.2 Recalibration
Neither instrument will require recalibration during or at the conclusion of the project. LiCor
specifies a maximum 1 % change in sensor readings after one year and General Electric lists a maximum
increase in uncertainty of 0.5 % over one year for pulse-initiator applications. ATI research on LiCor sensor
soiling indicates, under normal air quality conditions, a reduction in sensor response not greater than 2 %
with monthly cleaning.
General Electric accuracy specifications are traceable to the National Bureau of Standards GE reia:ns 2 meters periodically checked by ibc Bureau with Vrhich to calibrate their meters
Meters are shipped from lhe factory with a ± 0 2 % uncertainty. Due to the unpredictability of shipping and installation impacts on meter accuracy, GE increases its specified uncertainty to
± 0.5 % Additionally, utilities periodically run calibration checks on their meters Participating utilities will be asked to provide ATI with calibration data.
A-7
-------
6.0 DATA REDUCTION METHODS, VALIDATION AND REPORTING
6.1 Data Retrieval
Data records will be retrieved from the datalogger with phone line modern links using Campbell
Scientific PC208 software and a PC at the ATI office. Data retrieval will take place each night just after
midnight local standard time for each site. New data will be appended to preexisting data files for one-month
periods. With the amount of data contained in each datalogger record for this project, the circular datalogger
storage will not overwrite data records until they are typically over 32 days old. The PC data storage
resolution capabilities are greater than those selected for the datalogger (4 significant digits) and as such do
not represent any compromise in data accuracy.
6.2 Daily Processing
Ascension Technology will write custom software to provide daily data summaries and calculation
of DQIs. Each work day the following quantities will be reviewed:
(a)
Data recording date and period (MM/DD/YY 15-2400)
(b)
Number of records (96)
(c)
Peak EPV and EL kWh (with time of peak)
(d)
Total EPV and EL kWh
(e)
lPOA and EPV bias and RMS errors
(0
PV system efficiency
Calculation of (e) and (f), all of which are DQIs, will be outlined in section 8.
A-8
-------
6,3 Monthly Processing
At the end of each month data files will be reviewed for proper system performance. New data files
will be generated with software which aggregates the 15 minute data records into hourly averages and
merges them with data provided by the host utilities. The utilities have agreed to provide records of which
load-following power plants (marginal or "swing" plants) are operating, along with total system load for
hourly intervals. They will also supply emissions data for all potential marginal plants in their generating
mix. The data format will be as follows:
Table 3
EPA Data File Format
JD
Julian Day
H
Hour
Epv
PV Array Energy-
El
Building Load
p
Marginal Power Plant Identifier Number
Eu
Utility System Load
P is the identifier code from the table of marginal plants and their fuel and emissions information.
Ey is the total utility system load. Hourly data files for each month will be appended to one another in three-
month groupings.
Monthly processing will also include investigation of the possibility of further reducing building
peak loads using battery storage and dispatch simulations. In situations where the capacity charge is a
significant factor in a building's utility bill, the value of PV generation may be significantly greater when
used in timed, dispatchable units.
A-9
-------
6.4 Quarterly Processing
Every three months the data files of hourly records will be processed to calculate the following
quantities:
(a) Energy produced by the PV system (kWh)
(b) Avoided power plant fuel consumption due to (a)
(c) Avoided emissions due to (b)
(d) Reduced capacity requirements due to (a)"
Items (b) and (c) will be straightforward calculations of displaced emissions and fuel per kilowatt
of installed photovoltaic capacity at each site. Calculations of (d) will represent, for each installed kW of
PV, the reduction in peak load for the host building. This will be achieved by identifying EPV at the time of
the maximum building load (EL + Epy)^, System load capacity reductions per kW of PV will be calculated
in a similar manner by finding E,.v at (Ey + EPV)mM.
6.5 Programming Validation
All processing programs will be verified with spreadsheet calculations.
6.6 Data Reporting
Data reductions outlined in 6.4 will be presented in hardcopy form to the EPA following the end of
each three month period. Final analysis and conclusions will be presented following the conclusion of the
4th quarter. The quarters will be designated as:
Summer June 1993 - August 1993
Fall September 1993 - November 1993
Winter December 1993 - February 1994
Spring March 1994 - May 1994
1: At utility system and building load levels, both with and without batiery storage/dispatch
A-10
-------
7.0 SYSTEM AND PERFORMANCE AUDIT PLANS
ATI will assess system performance continually using daily data processing and screening programs.
EPA quality assessment and control methodology audits shall be performed as deemed appropriate by the
EPA QA Officer.
8.0 CALCULATION OF DATA QUALITY INDICATORS
Data quality indicators will be calculated and monitored daily through the use of the processing
programs developed for this project (as outlined in sections 3 and 6). The first DQI, data continuity, will be
demonstrated by the number of records retrieved each day (96) and the time series continuity check on all
records performed by the daily processing program (all records increase sequentially with respect to time).
The next four data quality indicators, EPV and IP04 bias and RMS errors (pj>v, pt>0A.. RMSV, and
RMSpoJ, will be calculated both over the previous day and for the entire current month. The formulas for
calculation of bias and RMS errors between two quantities x and y are given by:
p = ^ E
N j=i
RMS
\
— £ (xryO2
N i .
where N is the number of records for the day or the month. Tn both cases the measured values will be the
x variable.
A-ll
-------
PV system efficiency (T|pV) will be calculated according to
, n E
1 ¦r-, PVl
N Ui h0A. • n
POA • * Array
Data Quality Indicators will be presented with each quarterly report, along with the operational log
taken for all sites. Any iterruptions or inadequacies in the data stream will be brought to the attention of the
EPA QA Officer,
9.0 CORRECTIVE ACTION PROCEDURES
Corrective action required by unsatisfactory DQIs (to extents greater than outlined in section 3) shall
be decided upon by ATI personnel in collaboration with the EPA QA Officer.
A-12
-------
Appendix B. Quality Control Evaluation Report
SUMMARY
The approved Quality Assurance Project Plan (QAPjP) specified two data quality objectives (DQOs) for
the project: 1) that the proposed uncertainty limits be upheld; and 2) that continuous data sets be
collected from each site. For the primary data sets collected by this project, PV system generation and
host building load, the QAPjP specified measurement uncertainty of ± 0.5%. Ascension Technology
believes that this goal has been satisfied. Additional data sets containing hourly system-level power
demand, the load level of individual generating units, and the emission and fuel consumption rates of
individual generating units (or proxies therefore) were provided by each of the utilities participating in
the project. These data sets were used to determine pollutant offsets and utility system-level peak load
reductions resulting from the operation of the PV systems. Because project staff were unable to obtain
uncertainty limits for these utility-supplied data sets, the uncertainty associated with the resulting
analyses is unknown.
For 10 of the 17 sites, continuous data sets were collected throughout the study period. Completeness of
the data sets at the remaining sites was very high as well, although delayed initiation of metering at three
sites resulted in building load data sets less than 80 percent complete.
DATA QUALITY
Precision and Accuracy of Measurements
The primary measurements obtained in this study were those of PV system power output and host
building load. Both of these measurements were made using General Electric IW-70-1S or similar
kilowatt-hour meters with pulse initiators. Discussions with utility personnel indicated that these meters
are adjusted to a maximum uncertainty of ±0.5 %. Because meter pulses are not subject to datalogger
measurement uncertainty, the total measurement uncertainty for these measurements is equal to that for
the meters.
Measurements of solar irradiance were also collected at each site as a means of verifying PV system
operation. Plane-of-array (1P0A) irradiance was collected and used in a PV system simulation program to
verify system performance and identity' any system outages. An Ascension Technology Rotating
Shadowband Pyranometer (RSP) was also installed at each site to measure global horizontal, direct
normal and diffuse horizontal irradiance. These data, in turn, were used to verify the IP0A measurements.
Both IP0A and RSP measurements were made with L1COR LI200SZ pyranometers. Uncertainties
associated with LI200SZ pyranometers are given in accordance with a 1992 National Renewable Energy
Laboratory review of the ATI RSP system12. The NREL RSP review encompassed a detailed breakdown
of the sources of uncertainty in RSP and raw LI200SZ data. Final conclusions indicated that if averaging
intervals greater than 5 minutes are used, RSP data is within the uncertainty limits of ± 4.03 % for global
horizontal and direct normal and ± 4.80 % for diffuse horizontal. The specification for global horizontal
irradiance measurements are directly applicable to 1P0A measurement uncertainty, as they are subject to
the same calibration, alignment, temperature, spectral and data logging sources of error.
'2 T Sioffcl, C Riordtn. and J Bigger ''Evaluation of Solar Radiauer, Measurement Systems EPRINREL Final Test Report" Volume I N'REL/TP-463-4.77] N'o^ember 1992 The
specifjcaticn.5 on the LICOR pyranometer were chos«ii from this report in lieu of manufacturer's specifications due to art Improved uncertainty characleniatton
B-l
-------
Completeness of Data Sets
Completeness of the primary data sets can be measured by the ratio of the number of data records
collected to the total number of records possible, following initiation of data acquisition, Data
completeness was quite high for all sites, averaging 99.8 percent for PV system output and 95.7 percent
for building load data. Completeness of building load data is lower than that for PV due to the fact that
building load monitoring commenced well after the official monitoring start date at sites EPAG8, EPA09,
and EPA11. Aside from these three sites, the lowest completeness ratio for building load data was 97.5
percent, a result of datalogger power interruption which prevented data collection for 9 days. The
completeness ratios for each site are listed in Table A-l. These results compare favorably with the data
quality objective of 98 percent data recovery stated in the quality assurance project plan,
Table B-2 Completeness of PV System and Building Load Data Sets
Site
PV Data
Building Load
Data
EPA01
98.11
97.5
EPA02
100
100
EPA03
100
100
EPA04
99.26
99.26
EPA05
99.99
99.99
EPA06
100
100
EPA07
100
100
EPA08
100
78.12
EPA09
99.05
76.52
EPA 10
100
100
EPA11
100
75.5
EPA 12
100
100
EPA13
100
100
EPA 14
100
100
EPA15
100
100
EPA16
99.65
99.65
B-2
-------
Representativeness of Data
The data collected through this project and the analyses thereof provide a series of case studies of the
performance of PV systems and their impacts on customer and utility-level power demand and utility
pollutant emissions. Throughout this report, the attempt has been made to express results in a manner
which facilitates extrapolation to larger systems located at or near the existing system. Due to the
geographical variation in the solar resource, data on the performance of any of these systems may not be
representative of the performance of other PV systems located at a substantial distance from those
included in this study. Furthermore, it should be noted that factors such as array orientation, tracking
ability (or lack thereof), the relative magnitudes of PV array and power conditioner capacities, and local
microclimates may have dramatic effects on the performance of a PV system. In attempting to apply the
results of this study to other PV systems, differences in any of these factors should be carefully
considered.
LIMITATIONS OR CONSTRAINTS ON DATA APPLICABILITY
In addition to the limits on data representativeness discussed in the previous section, the following
caveats are in order:
1) The results of the emission offset analysis are predicated on the assumption that the small scale
of the PV system had no effect on the participating utilities' dispatch of conventional generating
units. If PV systems became an appreciable portion of a utility's generation mix, the hourly
contribution of PV would likely be forecast, and this forecast used to alter that utility's dispatch
decisions. In this situation, PV generation would likely displace intermediate-level or cycling
generation, rather than generation at the margin as assumed in this study. Because the generating
units used to serve peak loads are often among the least polluting of the units operated by a
utility-', the effect of moving PV generation further down in the dispatch order would likely be an
increase in the emission offset achieved per kW of PV capacity.
2) The emission offset results calculated in this study are specific to each individual utility and the
current mix of generating resources each uses to follow load. If utilities install equipment to
reduce pollutant emissions as a result of the Clean Air Act or other regulation, emission offsets
could decline in the future.
3) The building-level DSM results contained in this report apply only to the participating end-use
customers, over the course of the 15-month study period. The results contained herein should be
indicative of future building level demand reduction for these customers, provided that their load
patterns remain relatively constant.
4) The utility system-level DSM results reported herein apply only to the individual utilities for
which they were calculated. In areas of the country where peak loads are driven primarily by
weather patterns, the load patterns of utilities with adjacent service territories may be similar. In
such cases, it may be possible to extrapolate system-level DSM results to neighboring utilities.
RESULTS OF TECHNICAL SYSTEM AUDIT
On April 18th and 19th 1994, the EPA conducted a technical audit of Ascension Technology's data
measurement, collection and processing procedures for this project. Overall, the auditors found that
project personnel and management were technically qualified and highly motivated to produce reliable
B-3
-------
data. There were no audit findings that would seriously impact the usability of PV measurement data.
There were, however, audit findings related to comparability of data supplied by the different utilities,
sensor calibrations, and data management procedures and documentation. Following this audit, project
staff were provided with a list of tasks necessary for the fulfillment of the data quality objectives stated
in the QAPjP. This list is presented below, along with actions taken in response to the concerns of the
auditors.
Finding #1: The information supplied by the different utilities was not comparable. Some utilities
reported hourly data for all load-following plants in the "pool" from which the utility buys or sells
electricity. Other utilities reported hourly data for only their own plants. One utility responded with a
single set of emission factors (not hour data) that the utility had estimated for the year 2003 based on
planned pollutant reductions. In addition to the lack of comparability, the utilities had not supplied any
quality assurance or validation information supporting the emissions or fuel consumption data.
Response: Ascension Technology personnel discussed this finding with the EPA project officer, and
agreed to try to obtain the quality assurance information cited as missing. On May 9, 1994, Ascension
Technology sent a memo to all participating utilities requesting copies of meter calibration certificates
for both PV and building load meters, calibration records of instruments used to measure emission and
fuel consumption of the generating units used for load-following, and a letter on company letterhead
explaining the method and rationale for emissions and fuel consumption rate reporting, along with the
rates themselves. While most utilities responded to this memo by providing calibration certificates for
the PV and building load meters, few provided the requested description of the method or rationale for
the determination of marginal emission rates, and none of the utilities supplied calibration records for the
instruments used to measure emissions or fuel consumption. After the conclusion of the study period, a
second attempt was made to obtain this information from the utilities. This attempt was also
unsuccessful in obtaining the desired QA information.
Finding #2; The preferred method for calculating the offsets had not been established. As discussed in
chapter 8 of this report, project staff developed a variety' of methods for the calculation of emission and
fuel offsets. The selection of a method for a given utility was dependent on the information available to
a utility and its willingness to share information for the purposes of this project. The auditors felt that it
would be highly desirable to adopt a single uniform methodology for the calculation of emission offsets
for all utilities.
Response: Ideally, data sets from continuous emission monitors (CEMs) installed on the single load-
following unit or on all units on automatic generation control would have been used to determine hourly
marginal emission rates. In reality, however, CEMs are installed on only a small fraction of generating
units, typically the large baseload units which are subject to emissions limits under the Clean Air Act.
Frequently, the emission rates of the load-following units whose generation was displaced by that from
the PV systems is not measured at all, but rather calculated based on algorithms and factors contained in
the EPA publication Compilation of Air Pollutant Emission Factors - Volume 1: Stationary Point and
Area Sources (a.k.a. the AP-42 report). Furthermore, in the process of determining a suitable marginal
emissions model for each utility, many utilities stated that it was difficult or impossible to identify a
single marginal unit in each hour. As a result, it was agreed that Ascension would apply the best
information available from each utility. Variations in the methodology applied to calculate emissions
and fuel offsets at each site reflect variations in the type and extent of data available from each utility.
B-4
-------
Finding #3: Calibration certificates were not available for the power meters used to measure the PV
array output and the building power consumption.
Response: As noted above, meter calibration certificates were requested and received from most of the
participating utilities. The certificates showed that the meters had been recently calibrated to within the
0.5% uncertainty limit specified in the QAPjP as of the start of the study period.
Finding #4: Calibrations for some of the plane-of-array (POA) and RSP pyranometers would expire
before the project was finished. The auditors noted that data from these pyranometers were not
designated as critical measurements, ar.d suggested three options for addressing this concern; 1)
replacing individual sensors as their calibrations expired; 2) performing a field recalibration using a
recently calibrated sensor; and 3) recalibrating the sensors at the conclusion of the monitoring period.
Response: After discussing this matter with die EPA Project Officer, it was determined that because the
data collected from these sensors were not critical measurements but were instead used in the
development of data quality indicators for critical measurements, it was unnecessary to conduct the
sensor replacement or recalibration recommended by the auditors. It should be recalled that the data
collected by the RSP was used to verify plane-of-array irradiance measurements taken by the POA
pyranometer. Irradiance data from the POA sensor was used in tum to simulate the performance of each
PV system. Measured PV generation was compared to simulated generation on a daily basis to verify
proper system operation. The simulation was intended solely to identify partial or total system outages,
rather than as a precise estimate of expected PV generation.
Finding #5: The auditors found some weaknesses in routine data management procedures and
documentation. Specifically, they found that 1) screening of the daily error logs generated at time of
data retrieval had not been done in sufficient detail to catch problems such as inaccurate drift in the
datalogger time clocks; 2) network-wide clock checking and resetting procedures were not being
conducted at regular intervals: and 3) checklists for routine procedures (such as clock-setting) were
lacking.
Response: In response to these concerns, Project staff took two actions. First, modifications were made
to the daily screening software so that it would generate a single file summarizing any and all error
messages recovered from the dataloggers on a daily basis. Second, a highly accurate time clock card was
purchased and installed in the computer responsible for calling each datalogger to retrieve data. A
standard operating procedure was implemented to use this clock to reset datalogger clocks on a monthly
basis. This procedure was highly effective in maintaining the accuracy of datalogger clocks, which were
found to drift by no more that 30 seconds per month.
Finding #6: No computerized data validation information was being stored. At the time of the audit,
notes on malfunctions of the PV system, irradiance sensors, and datalogger were written in a log book.
The auditors felt that this information should be associated with the computerized data to facilitate access
to data qualifications and limitations.
Response: Shortly after the audit, Ascension Technology procured Paradox database software and began
using it to record all system events.
B-5
-------
Finding #7: There were no instructions for using the computer programs that had been developed for
data management. The auditors felt that written instructions would be valuable if new personnel were
brought in to operate the software.
Response: Although written instructions for the use of the computer software were not developed, the
operation of this software is quite simple, and easily explained verbally. Several personnel were trained
and became proficient in the use of this software.
Finding #8: Computer programs were lacking sufficient in-line documentation and comments. The
auditors were concerned that the software developed for this project lacked introductory comments
regarding the purpose of each program and the history of revisions. The auditors felt that such comments
would be very helpful in the event that software modifications were later determined to have introduced
errors.
Response: Introductory descriptive comments were added at the beginning of each program. In
addition, all program modifications were described in the program documentation.
B-6
-------
Appendix C. Operation of the Rotating Shadow-band Pyra no meter
Solar Resource Monitoring
Ascension Technology, Inc.
Telephone; (617) 890-8844 • Facsimile (617) 890-2050* email: ekem@mit.edu
Mail; P.O. Box 314* Lincoln Center • Massachusetts 01773 USA
Offices: 235 Bear Hill Road • Waltham • Massachusetts 02154 USA
Solar Electric Power Planning
Photovoltaic (PV) power is universally
regarded as a desirable source for electric power
generation in the future. During the 1980s,
research projects
demonstrated PV power's
reliability and ability to
complement
conventional electric
power generation and
delivery systems. In the
mid-1990s, efforts to
commercialize utility PV
power have have been
launched by the U.S.
electric power industry.
PV Module
Air Temperature
Sensor and Shield
Ascension Technology's (AT) Rotating
Shadowband Pyranometer (RSP) measures
direct normal and horizontal diffuse irradiance.
The RSP and associated data acquisition
equipment form a
rugged, integrated
system that simplifies
field measurement of
solar resources.
( J'
Enclosure;
• Datalogger
• PV Controller
• Battery
To determine PV's value
electric power planners
need solar energy
resource data for specific areas. Solar energy
reaches the earth's surface along two paths —
irradiance directly from the sun and diffuse
irradiance from the sky. Both components are
needed to estimate the energy produced by-
various fixed-orientation and tracking PV
system designs.
RSP Head Unit.
• Pyranometer
Shadowband
Motor The Ascension RSP
unit replaces a solar
tracker, a pvrheliometer,
and pyranometer at a
lower cost. The RSP is
easily installed and
requires much less
maintenance than
conventional
instruments. The RSP
system includes the head
unit, manufactured by
AT, a datalogger
manufactured by
Campbell Scientific, and proprietary software.
The multi-purpose datalogger adds
flexibility' to ATs RSP system. With added
sensors the datalogger can also record watt-hour
meter pulses and transducer outputs for watts,
volt-amps, current, wind speed and direction,
rainfall, and relative humidity.
C-l
-------
PHOTOVOLTAIC GENERATION ESTIMATES
Monthly AC ganeration (kWh/kWp)
¦ Concentrator cz Fixed 30 deg tilt
s North south axis a Two-axis tracking
AT RSP solar monitoring stations are
operating at over 150 locations in the United
States, Brazil, Mexico and Pakistan. They are
meeting the irradiance data acquisition
requirements of the U.S. National Renewable
Energy Laboratory, the U.S. Department of
Energy's Sandia National Laboratories, the
Electric Power Research Institute, the U.S.
Environmental Protection Agency and over
thirty U.S. electric power companies.
PV Generation Estimates
Direct normal and horizontal diffuse solar
irradiance data and temperature data are used to
estimate the energy produced by fixed and
tracking PV systems. These estimates support
utility planning and PV system design studies
by comparing the performance of various fixed
and tracking array concepts. RSP units have
been deployed with over 30 utility photovoltaic
systems to verify system performance with
comparisons of measured and simulated power
generation.
AT estimates PV systems energy production
using the amount of sunlight energy, or
insolation, received by fixed and tracking PV
arrays, and the temperature at which the PV
cells operate. These data can then be used to
compute the dc and ac power output for any
fixed or tracking PV system. The above figure
shows PV systems energy production estimates,
expressed as monthly ac kWh per kW of PV
capacity, for Riverhead, NY, based on RSP data.
AT provides these estimates for the most
frequently studied PV array designs.
RSP Network Operation
Routine Ascension Technology operation
of an RSP system includes daily data retrieval,
quality checking and a single-page summary
comparing measurements from all network
stations. At the end of each month, customers
are provided with PV system generation
C-2
-------
estimates, data summaries and data files on
diskette.
Ascension Technology's personal computer
software allows customers to monitor RSPs,
other meteorological measures, and PV systems
in real time. In addition AT offers a proprietary
graphic software package allowing users to scan
and check RSP data files quickly and display
direct normal, global horizontal and horizontal
diffuse irradiance, ambient temperature, and PV
system data in a convenient graphical format.
-------
FUNDAMENTALS OF OPERATION
1,000
800
600
400
200
Measured irradiance (Watts/sqrn)
(A)
B)
(C)
(D)
_ _
Total Horizontal
Th
_
Direct normal * cos(zenilti),
_ —
_
_
Dirn cos{Z)
_. — -
_ —
-
-V
-
diffuse horizontal, DrJ^
i
!
I I
0.2 0.4 0.6 0.8 1
Time {seconds)
Fundamentals of Opera tion
The AT RSP uses a single light sensor
(pyranometer) to determine total horizontal,
direct normal and horizontal diffuse irradiance.
These are related by the equation:
Th = Difh + Dirn cos(Z)
which expresses Th, the total irradiance
measured on a horizontal surface; Difh, the
diffuse irradiance (skylight) on a horizontal
surface; Dirn, the direct normal irradiance
(sunlight directly incident on a surface facing
the sun); and Z, the sun's zenith angle, the angle
measured from straight overhead down to the
center of the sun.
Once per minute the shadowband circles
over the light sensor, taking approximately one
second for this motion. During this one-second
period the pyranometer signal is sampled about
700 times. The minimum pyranometer reading
occurs
when the solar disk is completely occluded by
the shadowband. The stream of high-sample-
rate irradiance data is processed to measure the
horizontal diffuse irradiance. With Th, Difh and
Z known, Dir„ is calculated.
The above graph illustrates measurements
during a single shadowband rotation on a clear
day. The pyranometer views the full sky while
the shadowband travels from its stowed position
below the sensor, up to the horizon (A). As the
band traverses its path above the horizon, it
blocks a small strip of the sky, reducing the
diffuse irradiance falling on the sensor (B). The
irradiance measurement drops dramatically
when the shadowband shades the sensor from
direct sunlight (C). A symmetrical pattern
occurs as the shadowband completes its
revolution, ending in its stowed position below
the pyranometer (D).
C-4
-------
SPECIFICATIONS AND PRICES
Ascension Technology's RSP has been
thoroughly tested and evaluated by both the
National Renewable Energy Laboratory and by
Sandia National Laboratories. Both
organizations have reported their findings in
reports documenting the comparative
performance of RSP versus conventional
radiation monitoring instruments. These reports
are available on request.
Specifications
Rotating Shadowband Pvranometer Head Unit:
Licor 200SZ pyranometer. shadowband, drive
mechanism, wiring harness and mounting
bracket.
Air temperature sensor: rated from -35C to
+50C, gill radiation shield, wiring and
mounting.
Photovoltaic power supplv: 18-watt
photovoltaic module, wiring harness and
mounting bracket.
Instrument enclosure: Fiberglass enclosure (12"
wide x 14" high x 6" deep), mounting hardware,
Campbell Scientific CR10 micrologger, 1200-
baud telephone modem, charge regulator and a
32 amp-hour gel-cel type battery.
Ascension Technology software license for
CR10 code to control the RSP and determine
global, direct and diffuse irradiance and ambient
air temperature.
Mounting pipe: Pipe with a threaded end for
mounting the RSP head unit and hardware for
mounting the pipe to a wall or flat roof using a
ballast tray.
C-5
-------
Appenix D. Events Affecting PV System Performance.
Table D-l. PV System Events for Plattsburgh. NY
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
7/01/93 - 7/16/93 One of three inverters shut down due to an inverter control card 420
malfunction.
7/23/93 - PV system was the suspected source of noise on building's electrical 3,150
10/05/93 distribution system. System was shut down until DC filters were added to
the inverters.
November 1993 Snow cover. 50
December 1993 Snow cover. 150
January 1994 Snow cover. 410
February 1994 Snow cover. 340
2/1/94 - 4/25/94 Inverter Failure (DC injection), 930
March 1994 Snow cover. 600
6/5/94 - 6/15/94 DC disconnect fuse failure. 240
6/18/94-6/30/94 DC disconnect fuse failure. 230
8/21/94 - 8/29/94 DC disconnect fuse failure. 150
Annual system generation®: 9,744
Generation loss due to inverter-related malfunction: 4,500
Generation loss due to snow cover: 1,550
Generation loss due to other causes: 620
3 Sum of monthly generation data between October 1993 and September 1994.
D-l
-------
Table D-2. PV System Events for Berlin, CT
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
7/17/93 - 8/06/93 Inverter failure, system shut down. 370
December 1993 Snow cover 20
January 1994 Snow cover 130
February 1994 Snow cover 180
Annual System Generation3: 4,813
Generation loss due to inverter-related malfunction; 370
Generation loss due to snow cover: 310
Generation loss due to other causes: 0
0 Sum of monthly generation data between October 1993 and September 1994.
D-2
-------
Table D-3. PV System Events for PleasantviHe, NJ
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
7/01/93 - 7/15/93 Inverter IGBT failure-system at 2/3 power. 330
7/5/93 - 7/21/95 Water infiltration leading to diode failure in source-circuit 390
protector.
7/15/93 - 8/13/93 Inverter control board fuse failure. 1.640
9/16/93-9/22/93 Inverter shut down due to excessive noise. 60
December 1993 Snow cover. 80
January 1994 Snow cover. 140
1/18/94 - 3/11/94 Inverter failure-DC injection 440
February 1994 Snow cover. 110
3/17/94 - 4/22/94 Inverter removed to replace faulty inverter in Brigantine, NJ 600
system.
7/23/94-8/2/94 DC disconnect fuse failure. 150
Annual system generation8: 13,932
Generation loss due to inverter-related malfunction: 2,470
Generation loss due to snow cover: 330
Generation loss due to other causes: 1,140
¦ Sum of monthly generation data between October 1993 and September 1994.
D-3
-------
Table D-4. PV System Events for Brigantine, NJ
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
8/18/93 - 8/20/93 DC disconnect fuse failure. 40
8/20/93-8/22/93 Data gap. 50
9/16/93-9/20/93 DC disconnect fuse failure. 60
12/6/93- 12/9/93 Inverter shut down. 50
December 1993 Snow cover. 10
2/25/93 - 2/28/93 System shut down, cause unknown, 80
3/4/93-3/17/93 Inverter failure (unspecified). 190
Annual system generation3: 5,461
Generation loss due to inverter-related malfunction: 240
Generation loss due to snow cover: 10
Generation loss due to other causes: 230
3 Sum of monthly generation data between October 1993 and September 1994.
Table D-S. PV System Events for White Plains, NY
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
7/22/93- 8/13/93 Unspecified inverter failure. 360
December 1993 Snow cover. 30
January 1994 Snow cover. 110
February 1994 Snow cover. 110
6/11/94 - 6/31/94 DC disconnect fuse failure. 320
8/14/94 - 8/18/94 DC disconnect fuse failure, 60
Annual system generation3: 4,168
Generation loss due to inverter-related malfunction: 360
Generation loss due to snow cover: 250
Generation loss due to other causes: 380
3 Sum of monthly generation data between October 1993 and September 1994.
D-4
-------
Table D-6 PV System Events for Peoria, AZ
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
No outages during study period.
Annual System Generation®; 13,325
Generation loss due to inverter-related malfunction: 0
Generation loss due to snow cover: 0
Generation loss due to other causes: 0
3 Sum of monthly generation data between October 1993 and September 1994.
Table D-7. PV System Events for Scottsdale, AZ
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
3/4/94 - 3/6/94 One of two inverters shut down, cause unknown, 60
3/8/94 PV system shut down entirely, cause unknown. 40
3/20/94 - 3/21/94 PV system shut down entirely, cause unknown, 70
7/18/94 - 7/20/94 DC disconnect fuse failure. 40
Annual System Generation3: 5,983
Generation loss due to inverter-related malfunction: 60
Generation loss due to snow cover: 0
Generation loss due to other causes: 150
a Sum of monthly generation data between October 1993 and September 1994.
D-5
-------
Table D-8. PV System Events for Ashwaubenon, W1
Estimated
Event Event Generation
Date(s) Description Loss
(kWb)
8/9/93 - 9/10/93 Inverter failure-DC injection. 470
9/28/93- Inverter failure-DC injection. 140
10/11/93
10/30/93 - System shut down for installation of instrumentation. 70
11/1/93
December 1993 Snow cover. 70
January 1994 Snow cover, 700
February 1994 Snow cover. 630
3/11/94 System shut down for metering troubleshooting. 40
April 1994 Snow cover. 100
7/30/94 - 8/5/94 System rewired for open-circuit and short-circuit testing. 260
9/24/94 - 9/30/94 DC disconnect fuse failure. 80
Annual System Generation8: 13,814
Generation loss due to inverter-related malfunction: 610
Generation loss due to snow cover: 1,500
Generation loss due to other causes: 450
¦ Sum of monthly generation data between October 1993 and September 1994.
D-6
-------
Table D-9. PV System Events for Denmark, WJ
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
7/12/93 - 7/26/93 Unspecified inverter failure, system shut down. 430
9/7/83 - 9/8/93 Inverter control board fuse failure. 40
January 1994 Snow cover. 90
February 1994 Snow cover. 100
March 1994 Snow cover. 20
5/14/94 - 5/15/94 Power interruption to datalogger resulted in data loss. 40
5/16/94 - 5/20/94 I nverter failure -- DC injection. 150
Annual system generation8; 5,128
Generation loss due to inverter-related malfunction: 820
Generation loss due to snow cover: 210
Generation loss due to other causes: 40
* Sum of monthly generation data between October 1993 and September 1994.
D-7
-------
Table D-10, PV System Events for Minnetonka, MN
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
10/1/93- 10/4/93 DC disconnect fuse failure. 60
10/5/93- Unspecified inverter failure. 120
10/12/93
November 1993 Snow cover. 40
January 1994 Snow cover. 90
February 1994 Snow cover. 180
Annual system generation*: 4,870
Generation loss due to inverter-related malfunction; 120
Generation loss due to snow cover: 310
Generation loss due to other causes: 60
a Sum of monthly generation data between October 1993 and September 1994.
Table D-l 1. PV System Events for San Ramon, CA
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
2/12/94-2/14/94 PV system shut down - cause unknown. 50
5/2/94 - 5/4/94 Partial system shut down -- cause unknown 60
5/14/94 - 5/27/94 Inverter failure - DC injection. 270
6/30/94 - 7/1 /94 System shut down far testing. 80
7/1/94-8/11/94 One half-string not reconnected after testing. 370
9/12/94-9/14/94 DC disconnect fuse failure. 60
10/19/94 - Inverter failure - DC injection. 80
10/23/94
Annual system generation8: 18,445
Generation loss due to inverter-related malfunction: 350
Generation loss due to snow cover: 0
Generation loss due to other causes: 620
" Sum of monthly generation data between February 1994 and January 1995.
D-8
-------
Table D-12 PV System Events for Austin, TXa
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
3/14/94 - 5/11/94 Inverter failure — two inverters operating intermittently, 1,190
5/12/94 - 5/25/94 Inverters removed for repair. 870
8/11/94 - 6/13/94 DC disconnect fuse failure. 50
8/14/94-6/17/94 System shut down, cause unknown. 170
7/7/94 - 7/8/94 DC disconnect fuse failure. 40
8/2/94 - 8/11 /94 DC disconnect fuse failure. 180
9/12/94 - 9/13/94 Inverter failure - DC injection. 30
10/9/94 - Inverter failure - DC injection. 80
10/12/94
11/29/94 - 1/9/95 Inverter failure - DC injection. 390
1/9/95-2/17/95 Inverter failure -- blown transistor 440
3/7/95 - 3/8/95 Inverter failure - DC injection 40
Annual system generation11: 13,406
Generation loss due to inverter-related malfunction: 3,040
Generation loss due to snow cover: 0
Generation loss due to other causes: 440
* PV metering malfunction prevented accurate recording of system performance prior to March 14,1994.
Analysis of system events has therefore been extended through March 1995.
6 Sum of monthly generation data between April 1994 and March 1995.
D-9
-------
Table D-13 PV System Events for Flaggstaft AZ"
Event
Date{s)
Event
Description
Estimated
Generation
Loss
(kWh)
October 1994
Snow cover.
20
November 1994
Snow cover.
40
December 1994
Snow cover.
20
January 1995
Snow cover.
240
2/7/95 - 2/28/95
AC disconnect fuse failure.
390
Annual system generation";
Generation loss due to inverter-related malfunction:
Generation loss due to snow cover:
Generation loss due to other causes:
6,031
0
320
390
a System operation commenced on 12/1/93, but PV meter was not installed until 2/12/94. Analysis of system
events has therefore been extended through February 1995.
b Sum of monthly generation data between March 1994 and February 1995.
Table D-I4 PV System Events for Baistow, CA»
Event
Date(s)
Event
Description
Estimated
Generation
Loss
(kWh)
6/1/94 - 5/30/94
Afternoon shading reduces system output by up to 40 percent in
1,500
summer months.
7/22/94 - 8/3/94
DC disconnect fuse failure.
210
5/25/95 - 5/31/95
AC line voltage exceeded inverter operating window, causing
100
inverter to shut down.
Annual system generation1":
Generation loss due to inverter-related malfunction:
Generation loss due to snow cover:
Generation loss due to other causes:
5,270
0
0
1,810
a System operation commenced 6/9/94. Analysis of system events has therefore been extended through
May 1995.
6 Sum of monthly generation data between June 1994 and May 1995.
D-10
-------
Table D-1S PV System Events for Edwards Air Force Base, CA3
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
12/20/94- DC disconnect fuse failure. 130
1/11/95
Annual system generation13: 6,276
Generation loss due to inverter-related malfunction: 0
Generation loss due to snow coven 0
Generation loss due to other causes: 130
a Data collection commenced 2/1/94. Analysis of system events has therefore been extended through
January 1995.
" Sum of monthly generation data between February 1994 and January 1995.
Table D-16 PV System Events for Palm Desert. CAa
Estimated
Event Event Generation
Date(s) Description Loss
(kWh)
No events affecting system performance between February 1994
and January 1995.
Annual system generation13: 17,694
Generation loss due to inverter-related malfunction: 0
Generation loss due to snow cover: 0
Generation loss due to other causes: 0
a Data collection commenced 2/1/94. Analysis of system events has therefore been extended through
January 1995.
b Sum of monthly generation data between February 1994 and January 1995.
D-ll
-------
TECHNICAL REPORT DATA
{Phase read JnOructions on the reverse before comp
llllilllllllll
PE97-1
III IIIIIN III
%. REPORT MO ]2
EPA-600/R-96-130 |
ill IIUK III
17613 )/
. l ^
96
A. TITLE AND SUBTITLE
Demonstration of the Environmental and Demand-side
Management Benefits of Grid-connected Photovoltaic
Power Systems
S. REPORT DATE
November 19
6. PERFORMING ORGANIZATION CODE
7, AUTHORlSl
Edward C. Kern, Jr. and Daniel L. Greenberg
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING OROANIZATION NAME AND ADDRESS
Ascension Technology, Inc.
P. O. Box 314
Lincoln Center, Massachusetts 01773
10. PROGRAM ELEMENT NO.
11, CONTRACT/GRANT NO.
68- D2-0148
12, SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
13. TYPE OF REPORT ANO PERIOD COVERED
Final; 8/82-1/96
14. SPONSORING AGENCY CODE
EPA/600/13
"¦supplementaryNOTES APPCD project officer is Eonald ^ Spiegel, Mail Drop 63, 919/
541-7542.
is. abstract rep0r£ g£ves results of an investigation of the pollutant emission reduc-
tion and demand-side management potential of 16 photovoltaic (PV) systems instal-
led across the U.S. in 1993 and 1994. The investigation was sponsored by the U. S.
EPA and 11 electric utilities.- The report presents analyses of each system's ability
to offset emissions of sulfur dioxide, nitrogen oxides, carbon dioxide, and particu-
lates, and to provide power during peak load hours for the individual host building
and the utility. Results of simulations of battery storage systems powered by each
PV system are also presented. The analysis indicates a very broad range in the sys-
tems' abilities to offset pollutant emissions, due to variation in the solar resources
available and the marginal emission rates of the participating utilities. Use of dis-
patchable storage would reduce emission offsets due to energy losses in charging and
discharging the batteries. Each system's ability to reduce building peak loads depen-
ded on the correlation of that load to the available solar resource. Most systems
operated in excess of 50% of their capacity during building peak load hours in the
summer months, but well below that level during winter peak hours. Similarly, many
systems operated above 50% of their capacity during utility peak load hours in the
summer months, but at a very low level during winter peak hours.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Pollution Carbon Dioxide
Photovoltaic Cells Particles
Electric' Power
Generation
Sulfur Dioxide
Nitrogen Oxides
Pollution Prevention
Stationary Sources
Particulates
13 B
14 B 14G
IDA
07B
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This ReportJ
Unclassified
21. NO. OF PAGES
207
20. SECURITY CLASS (This pagej
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
------- |