Sand2014 2883

SANDIA REPORT
SAND2014-2883
Unlimited Release
Printed May 2014
Performance Assessment of the PNM

Prosperity Electricity Storage Project
A Study for the DOE Energy Storage Systems Program
Dakota Roberson, James F. Ellison, Dhruv Bhatnagar, and David A. Schoenwald
Prepared by
Sandia National Laboratories
Albuquerque, New Mexico 87185 and Livermore, California 94550
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation,
a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energys
National Nuclear Security Administration under contract DE-AC04-94AL85000.
Approved for public release; further dissemination unlimited.

Issued by Sandia National Laboratories, operated for the United States Department of Energy
by Sandia Corporation.
NOTICE: This report was prepared as an account of work sponsored by an agency of the United
States Government. Neither the United States Government, nor any agency thereof, nor any
of their employees, nor any of their contractors, subcontractors, or their employees, make any
warranty, express or implied, or assume any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, product, or process disclosed, or rep-
resent that its use would not infringe privately owned rights. Reference herein to any specific
commercial product, process, or service by trade name, trademark, manufacturer, or otherwise,
does not necessarily constitute or imply its endorsement, recommendation, or favoring by the
United States Government, any agency thereof, or any of their contractors or subcontractors.
The views and opinions expressed herein do not necessarily state or reflect those of the United
States Government, any agency thereof, or any of their contractors.
Printed in the United States of America. This report has been reproduced directly from the best
available copy.
Available to DOE and DOE contractors from

U.S. Department of Energy
Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831
Telephone: (865) 576-8401

Facsimile: (865) 576-5728
E-Mail: reports@adonis.osti.gov
Online ordering: http://www.osti.gov/bridge
Available to the public from

U.S. Department of Commerce
National Technical Information Service
5285 Port Royal Rd
Springfield, VA 22161
Telephone: (800) 553-6847

Facsimile: (703) 605-6900
E-Mail: orders@ntis.fedworld.gov
Online ordering: http://www.ntis.gov/help/ordermethods.asp?loc=7-4-0#online
NT OF E
ME N
RT
ER
A
DEP
GY

IC A
U NIT
ER
ED
ST M
A TES OF A
2
SAND2014-2883
Unlimited Release
Printed May 2014
PNM Prosperity Electricity Storage

Project Evaluation
A Study for the DOE Energy Storage Systems Program
Dakota Roberson, James F. Ellison, Dhruv Bhatnagar, and

David A. Schoenwald
Sandia National Laboratories
Electric Power Systems Research Department
P.O. Box 5800
Albuquerque, NM 87185-1140
Abstract
The purpose of this study is to characterize the technical performance of the PNM Prosperity
electricity storage project, and to identify lessons learned that can be used to improve similar
projects in the future. The PNM Prosperity electricity storage project consists of a 500 kW/350
kWh advanced lead-acid battery with integrated supercapacitor (for energy smoothing) and a 250
kW/1 MWh advanced lead-acid battery (for energy shifting), and is co-located with a 500 kW
solar photovoltaic (PV) resource. The project received American Reinvestment and Recovery Act
(ARRA) funding. The smoothing system is effective in smoothing intermittent PV output. The
shifting system exhibits good round-trip efficiencies, though the AC-to-AC annual average efficiency
is lower than one might hope. Given the current utilization of the smoothing system, there is an
opportunity to incorporate additional control algorithms in order to increase the value of the energy
storage system.
keywords: electrical energy storage, ARRA projects, technical evaluation, solar photovoltaic
smoothing.
3
Acknowledgment
This research was sponsored by the Office of Energy Delivery and Energy Reliabilitys Energy
Storage Program at the U.S. Department of Energy. The authors would like to thank Dr. Imre
Gyuk and his colleagues at the Energy Storage Program at the U.S. Department of Energy for their
funding and support of this project.
4
Contents
Abstract 3
Acknowledgement 4
Executive Summary 7
Acronyms and Abbreviations 9
1 Introduction 11
1.1 ARRA Energy-Storage Demonstration Projects . . . . . . . . . . . . . . . . . . . . 11
1.2 PNM Prosperity Project Description . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Battery Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4 Battery Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2 Evaluation Metrics 19
2.1 PV Smoothing Effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Round-Trip Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Balance-of-Plant Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Estimated Remaining Battery Lifetime . . . . . . . . . . . . . . . . . . . . . . . . . 22
3 Analysis 24
3.1 PV Smoothing Effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Shifting Battery Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Balance-of-Plant Load Characterization . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4 Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5 Estimated Remaining Battery Lifetime . . . . . . . . . . . . . . . . . . . . . . . . . 35
4 Recommendations for Battery and Photovoltaic Systems 37
5 Conclusions 40
References 42
Bibliography 42
A ARRA Energy Storage Projects 43

A.1 Grid-Connected Demonstration Projects . . . . . . . . . . . . . . . . . . . . . . . . 43
A.2 Research Level (Not Grid-Connected) Demonstration Projects . . . . . . . . . . . . 48
B Solar PV Smoothing Metric Derivation 51
C AGC Signal Assumptions for Smoothing Battery 59
5
List of Figures
1 PNM Prosperity energy-storage project. . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Battery storage-system one-line diagram. . . . . . . . . . . . . . . . . . . . . . . . . 15
3 System one-line diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4 UltraBattery configuration (Lam et al., 2006). . . . . . . . . . . . . . . . . . . . . . 17
5 PV shifting, 26-Jun-2013 data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6 High PV variability, 15-Apr-2012 data. . . . . . . . . . . . . . . . . . . . . . . . . . 25
7 Moderate PV variability, 22-Jun-2013 data. . . . . . . . . . . . . . . . . . . . . . . . 26
8 Mild PV variability, 13-Apr-2012 data. . . . . . . . . . . . . . . . . . . . . . . . . . 26
9 Shifting-batttery action, 23-June-2013 data. . . . . . . . . . . . . . . . . . . . . . . 29
10 Shifting-battery efficiency, by month. . . . . . . . . . . . . . . . . . . . . . . . . . . 30
11 Availability (source of data for calculations: Ecoult) . . . . . . . . . . . . . . . . . . 35
12 Smoothed power profile, 7-Mar-2012 data. . . . . . . . . . . . . . . . . . . . . . . . 52
13 Moving average versus original signal . . . . . . . . . . . . . . . . . . . . . . . . . . 53
14 Residue R(n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
15 Probability mass function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
16 Power demanded (kW) over time by the sample AGC signal from PJM. . . . . . . . 59
17 Cumulative Energy Flows (kWh) resulting from the sample AGC signal from PJM . 60
List of Tables
1 Shifting battery efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 ARRA grid-connected demonstration projects. . . . . . . . . . . . . . . . . . . . . 13
3 ARRA electricity storage research and development projects. . . . . . . . . . . . . 13
4 Reduction in PV Output Variability Due to Smoothing Battery . . . . . . . . . . . 27
5 Shifting Battery Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6 Round-Trip Efficiency over Sampled Days . . . . . . . . . . . . . . . . . . . . . . . 29
7 Round-Trip Efficiency over Sampled Days, Including BoP Load . . . . . . . . . . . 32
8 AC-to-AC Round-Trip Efficiency of Shifting-Battery System . . . . . . . . . . . . . 33
9 Shifting battery efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6
Executive Summary
The Public Service Company of New Mexico (PNM) Prosperity electricity storage project consists
of a 500 kW/350 kWh energy-smoothing battery and a 250 kW/1 MWh energy-shifting battery,
and is co-located with a 500 kW solar photovoltaic (PV) resource. The smoothing battery was
designed to smooth rapid fluctuations in solar PV output due to intermittent cloud cover, and the
shifting battery was designed to shift the PV resources output to better coincide with evening peak
load. The project received American Reinvestment and Recovery Act (ARRA) funding.
To evaluate how well the smoothing battery system performed, several metrics were proposed to
evaluate the systems ability to mitigate photovoltaic energy intermittency. The smoothing battery
is effective at reducing PV output volatility. For the days examined, when using the smoothing
battery system, the PV power output standard deviation was reduced by about 60%, and the
PV power ramp-rate standard deviation was reduced by about 55%. For an alternate measure,
the maximum-minimum adjusted PV power output swing was reduced by about 65%, and the
maximum-minimum adjusted PV power ramp rate swing was reduced by roughly 70%, when the
smoothing battery was used.1
Two factors are of primary interest when evaluating a shifting battery: the ability to charge and
discharge in blocks as commanded, and the round-trip efficiency. The system performance was
nearly flawless with respect to delivering the commanded charge/discharge energy. Therefore, this
report focuses on the round trip efficiency to evaluate the shifting battery performance.
Efficiency is calculated on both an AC-to-AC and DC-to-DC basis. In order to determine the
effects of the Balance of Plant (BoP) load on the efficiency of the shifting system, differing energy
components are utilized in the efficiency calculations. The efficiencies are determined over two
different time intervals: 24 hour and 365 day periods, representing daily and annual efficiencies,
respectively. These results are summarized in Table 1.
Table 1: Shifting battery efficiency.
Measurement Includes BoP Losses? Round-Trip Efficiency Annual Efficiency

DC-to-DC No 89% 85%
Yes (measured BoP) 83% 69%
AC-to-AC Yes (measured BoP) 76% 59%
Another way to evaluate efficiency is to compare the energy used to charge the battery versus the
energy output from the battery over some time period. The annual average efficiency (on a DC-to-
DC basis, not including measured BoP loads) over the twelve months through June 2013 was 85%.
If measured BoP load is included, the annual (AC-to-AC) efficiency this number drops to about
59%.
While the round-trip measurements are in-line with our expectations, the annual efficiency estimates
including BoP losses are lower than expected. The reason for this is that while the shifting battery
is being used for a fraction of the time, there is BoP load the entire time being examined. In other
words, BoP load when the battery is not being used (every night, as well as on days the battery
is not dispatched) counts against its annual efficiency. Operating the battery system more often
1
Ten days were examined.
7
would improve this efficiency, as would drawing less BoP load when the battery system is not in
use.
Placing the storage systems and the PV array on the same DC bus would likely improve the AC-
to-AC efficiency measure. PV energy must take two more one-way trips through an inverter than
it would if the two systems shared a DC bus. Assuming an average inverter efficiency of 96.5%
for one-way conversion, these two additional trips amount to nearly 7% higher losses to store PV
energy than would be the case if the systems shared a DC bus. One measure that could have been
taken during construction to reduce the air-conditioning portion of the BoP load is orienting the
containers East-West instead of North-South, exposing less wall area to the harsh West exposure
during the afternoon hours. Providing shade for the battery containers would likely reduce BoP
load and improve system efficiency.
Even though the storage systems were designed to smooth and shift PV output (and have been
almost exclusively operated in this manner), they are not limited to this application. The storage
systems are merely co-located with the PV arraythey can perform applications that do not depend
on PV output. Given this, it makes sense to examine whether the battery systems might provide
additional value by performing other applications.
From a bulk-grid perspective, providing frequency regulation by following an automatic generation
control (AGC) signal may be a better application than smoothing PV output for this location. If
smoothing PV output, the smoothing battery system will necessarily be tasked to operate only
during the day, meaning that it will be idle at night. In addition, New Mexico skies are generally
clear, meaning that on most days the battery will receive little use. Battery throughput data
shows that, assuming cumulative discharge is the limiting factor, the smoothing battery could last
another 100 years at the current useage rate. This confirms that the smoothing battery system
use is low compared with the battery systems lifetime capacity for work. Having the smoothing
battery system follow an AGC signal would increase the battery utilization and be useful over the
entire 24-hour period. Following an AGC signal is also of more benefit to the grid, as it means the
battery will always be acting in a way to balance generation and load. In contrast, smoothing PV
output can be counterproductive if there is a momentary drop in load at the same time there is a
momentary drop in PV output.
Moreover, using the energy-shifting battery system to deliver power when most needed and withdraw
power from the grid when it is cheapest may provide more value than shifting PV output. To add
value shifting energy, the system benefit of shifting off-peak energy to on-peak energy must outweigh
the cost of losing energy to storage round-trip inefficiencies. Shifting PV output means taking power
generated earlier in the day and shifting it to later in the day. Decoupling the storage resource from
PV output allows it to charge at night or during the day, increasing the operators ability to charge
when energy is cheapest. Using the shifting battery to provide spinning reserve is another option
that may add more value than PV shifting, and is worth study.
The PNM Prosperity project is one of a few grid-scale battery storage systems in the U.S., is one
of two storage systems in the U.S. using the UltraBattery (with the other being at the East Penn
factory in Pennsylvania), and is the only grid-scale storage system using the Deka Synergy battery.
The Prosperity project has provided, and continues to provide, significant lessons learned from its
operations. In addition, the PNM Prosperity and East Penn storage projects have given East Penn
experience in manufacturing the UltraBattery, resulting in increased manufacturing efficiency and
reduced cost.
8
Acronyms and Abbreviations
AGC automatic generation control
AGM absorbed glass mat
ARRA American Reinvestment and Recovery Act of 2009
BESS battery energy storage system
BoP balance of plant
CAES compressed air energy storage
CDF cumulative density function
CES community energy storage
ESS electricity storage system
KS Kolmogorov-Smirnov (test)
kW kilowatt
kWh kilowatt-hour
MW megawatt
MWh megawatt-hour
PG&E Pacific Gas and Electric Company
PJM a regional transmission organization that coordinates wholesale electricity generation in all
or parts of 13 states and the District of Columbia
PMF probability mass function
PSoC partial-state-of-charge
PV photovoltaic
RSDP reduction in standard deviation of power
RSDR reduction in standard deviation of ramp-rate
SCE Southern California Edison
SGDP Smart Grid Demonstration Program
SoC state-of-charge
V volts
VAC volts alternating current
VDC volts direct current
VRLA valve-regulated lead-acid
WSS wide-sense stationary
9
This Page Intentionally Left Blank
10
1 Introduction
This study focuses on evaluating the technical performance of the Public Service Company of New
Mexico (PNM) Prosperity projects storage resource, which is one of the electricity storage demon-
stration projects funded by the American Reinvestment and Recovery Act (ARRA) of 2009. Con-
tributing partners are: University of New Mexico, Northern New Mexico College, Sandia National
Laboratories, East Penn Manufacturing Company, and Electric Power Research Institute.
The PNM Prosperity electricity storage project consists of a 500 kW/350 kWh energy-smoothing
battery and a 250 kW/1 MWh energy-shifting battery, and is co-located with a 500 kW solar
photovoltaic (PV) resource. The smoothing battery was designed to smooth the PV resources
rapid output fluctuations due to intermittent cloud cover, and the shifting battery was designed to
shift PV resource output to better coincide with evening peak load.
Time-of-day shifting of solar PV output is useful, as the PNM Prosperity solar arrays peak energy
production occurs before the evening load peak. Having an energy system to shift (i.e., store then
discharge) solar production to better coincide with the evening peak is therefore likely to benefit
the bulk grid. Likewise, smoothing rapid PV output fluctuations, due to intermittent cloud cover,
is useful. At the bulk grid level, higher PV output variability requires more operating reserve to be
set aside, which is costly. At the distribution level, rapidly varying PV output can cause voltage
flicker and voltage excursions outside of the desired band.
This evaluation attempts to identify the technical value of the PNM Prosperity project, that is,
the benefit that the battery system provides as a grid asset. This includes energy, capacity, and
ramp-rate mitigation of the associated PV system amongst other services. It includes the deferment
of the use of other resources to provide these services. It does not evaluate system economics in
this useeither independently or relative to other resources.
A publicly-available, independent PNM Prosperity storage project evaluation serves several pur-
poses. First, any utility or developer considering a storage resource of this type will want to
understand the nature, and magnitude, of the technical risks associated with the project. A storage
system performance evaluation should contribute to a clearer understanding of technical risk. Sec-
ond, an evaluation helps provide perspective on the role of ARRA funding in facilitating advances
in electricity storage, specifically in the type of storage used at the PNM Prosperity site.
This report, and others that follow, will serve to identify the performance-based value proposition
for the ARRA energy storage demonstration projects. They will provide the public, utilities, and
potential owners an in-depth understanding of energy storage uses as demonstrated by the ARRA
demonstration projects, and address utility and developer concerns about using energy storage on
their systems.
After a brief introduction to the ARRA Energy-Storage Demonstration Projects, the PNM Pros-
perity storage project will be discussed in detail.
1.1 ARRA Energy-Storage Demonstration Projects
ARRA brought significant government stimulus funds into private industry, with about $4.5 billion
allocated to the U.S. Department of Energy to fund projects to modernize the electric grid. The
11
Smart Grid Demonstration Program (SGDP) is one primary avenue through which this funding has
been distributed.
The SGDP is authorized by the Energy Independent Security Act of 2007 Section 1304 and amended
by ARRA to demonstrate the technical, operational, and economic feasibility of existing and emerg-
ing smart-grid technologies. The DOE selected projects for this program through a competitive
solicitation, where up to 50% of the project cost was provided through the program.
Approximately $140 million in funding was allocated for the U.S. Department of Energy to fund
16 energy-storage demonstration projects. The remainder of the funding for these projects was
provided by state and private investment. The grid-connected demonstration projects are listed in
Table 2, and the research and development demonstration projects are listed in Table 3. A more
complete description of each of the projects is provided in Appendix A.
The SmartGrid.gov website provides further information on these projects: their value, current
status, and other relevant background information. As a whole, these projects attempt to demon-
strate the value of energy-storage technologies in a number of electric system functions and validate
their performance and long term capabilities in providing these functions. The overall intention of
this effort is to promote the deployment of these technologies by addressing the technical risk that
prevents utilities and other developers from deploying energy-storage technologies as solutions to
grid problems.
Using this resource and other resources available from the Department of Energys Energy Storage
Program and other publicly available sources, a potential developer can evaluate the performance
value proposition and the economic value proposition when considering solutions to addressing
electric grid needs. A number of modeling tools, regulatory support documentation, a cost database
and economic evaluations for potential deployments are available.
12
Table 2: ARRA grid-connected demonstration projects.
Project Name Location Project Description Funding ($M) Project

ARRA/Total Status
PNM Prosperity Albuquer 750 kW of advanced lead-acid 2.3/6.1 Operational
-que, NM batteries for PV smoothing/shifting
East Penn Lyons 3 MW UltraBAttery storage facility 2.5/5.1 Operational
Station, PA for frequency regulation
Duke Energy - Notrees, 36 MW/24 MWh Xtreme storage 22/44 Operational
Notrees TX facility for wind smoothing/shifting
Southern Tehachapi 8 MW/32 MWh lithium-ion battery 25/55 Under
California Edison CA storage facility for wind smoothing construction
City of Painesville, 10 MW/80 MWh Vanadium redox 4.2/9.5 In Progress
Painesville OH battery storage for peaking power
Primus Power - Modesto, 25 MW/75 MWh Zinc flow-battery 14/47 In Progress
EnergyPods CA storage for wind firming/peak shaving
DTE - Comm. Detroit, MI 1 MW community energy storage (CES) 5.0/11 Operational
Storage system to improve grid reliability
PGE CAES CA 300 MW/3000 MWh CAES for 25/360 In Study
renewable generation support
Beacon Power Hazle 20 MW flywheel for frequency 24/52.5 Operational
Twp, PA regulation
NYSE&G - NY 150 MW CAES 1.4/2.9 Discontinued
Advanced CAES
Premium Power MA Two 500 kW zinc-bromide flow- 6/12.5 Discontinued
Dist. Storage batteries for PV and wind support
Table 3: ARRA electricity storage research and development projects.
Project Name Location Project Description Funding ($M) Project

ARRA/Total Status
Aquion Energy PA Low-temperature sodium-ion battery 5.2/10.4 Complete
development for grid-scale storage
Ktech Corp NM EnerVault 250 kW/1 MWh redox flow 4.8/9.5 In progress
battery coupled with 180 kW PV system
Amber Kinetics CA Developing lower-cost flywheel 3.7/7.5 In progress
SEEO CA Solid-state electrolyte lithium-ion battery 6.2/12.4
SustainX NH Developing an isothermal CAES device 5.4/13.0
using pressure vessels for storage
13
1.2 PNM Prosperity Project Description
The Prosperity Project is located in Albuquerque, New Mexico, just south of the Albuquerque In-
ternational Airport on land owned by PNM, and is in PNMs service territory. An aerial photograph
of the project site is displayed in Figure 1. There are eight climate-controlled shipping containers
containing batteries. These are connected to a grid-tied inverter with a 750 kW/1.5 MVA rating.
Separately, there is a 500 kW PV installation, connected to a 500 kW grid-tied inverter. (The PV
installation and associated inverter were not part of this energy-storage project, and did not receive
DOE funding).
Figure 1: PNM Prosperity energy-storage project.
The Ecoult/East Penn Manufacturing storage system at the PNM Prosperity site is comprised
of:
A 0.25 MW/1 MWh energy-shifting battery system
Six shipping containers of advanced lead-acid batteries (Deka Synergy
R batteries)
A 0.5 MW/0.35 MWh power-smoothing battery system

Two shipping containers of advanced lead-acid batteries with integrated capacitors (Ul-
traBatteries)
A bi-directional grid-tied inverter with a 750 kV/1.5 MVA rating
Connected to each battery system through bi-directional DC converters
Inverter is capable of power factor modification
Also part of the storage system is an Ecoult battery management and monitoring system, a battery
power conditioning system, and a data acquisition and control system. A one-line diagram of the
battery system can be found in Figure 2.
14
/ ^

W
/EsWh

s/d
^W^

^^ ^
W
& KEs
Wh d
Wh
^

s & ^ ^
/ d
D ^ Wh
KEs
Wh

W
WEDZdh

^^ W^D
^

d
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0000000 Wh
W
s
W ^W 00000000000000000000000000000
000
W
Figure 2: Battery storage-system one-line diagram.
The two containers housing the energy-smoothing battery system are considered a single string.
This string is termed an Ultrabattery Storage Block, and is labeled as an UBer USB-320/500
kW storage block. The string voltage is nominally 320 VDC, though actual voltage can fluctuate
between 270 VDC and 400 VDC. The string has a power rating of 500 kW and a capacity of 1
MWh. Roughly 13 of this amount, or 0.35 MWh, can be made available for use without adversely
impacting battery life.2 This is termed the maximum energy band available for application use.
This can only be achieved at average power levels below the peak power rating.
The six containers housing the energy-shifting battery system are considered three strings. Each
string is comprised of two containers, and is termed a Deka Synergy Storage Block, and is labeled
as an UBer SSB-320/250 kW storage block. As with the smoothing battery system, the string
voltage is nominally 320 VDC, though actual voltage can fluctuate between 270 VDC and 400 VDC.
Each shifting battery string has a power rating of 250 kW and a capacity of 1 MWh. Roughly 1/3
of the storage capacity of a string (0.35 MWh) can be made available for use without adversely
impacting battery life. The three shifting battery strings are connected in parallel, yielding a
three-string system which has the same power rating of a single string, but triple the storage
capacity. Therefore, the total useable storage volume of the three shifting-battery strings is roughly
1 MWh.
DC-to-DC converters connect each system (smoothing and storage) to a grid-tied inverter. This
2
Page 24 of Ecoult brochure.
15
is done mainly to keep the voltage, as seen by the inverter, constant and to ensure that if both
battery systems are operating at the same time, they are sending power to/receiving power from
the inverter at the same voltage.
The PV system is connected to the grid through its own 500 kV inverter. Because the PV system
and storage system have separate inverters, they are therefore separately connected to the grid. This
is shown in Figure 3. Both inverters feed the low-voltage side of a 12.47 kV/480 V transformer.
While PNM would have preferred to have a single inverter with a common DC bus serving the PV
and battery storage systems, they encountered obstacles to doing so.
s W s Z
D / s
^t
WD, ^W
&
W
<s ^
^^
E
W
D
^
^^
s E
^
s &

Ws

s D D
/ s Ws/ t
t
WsW
s
W

&
t t s
^ ^
Ws/
D D
^ ^

Figure 3: System one-line diagram.
1.3 Battery Description
Both the Deka Synergy battery and the UltraBattery are advanced valve-regulated lead-acid (VRLA)
batteries.
VRLA batteries, also known as sealed batteries, use a one-way pressure-relief valve, allowing them
to be recombinant. This means that the oxygen produced on the positive plate is absorbed by the
negative plate, suppressing the production of hydrogen and producing water instead. The battery
thus never needs for water to be added. The name valve regulated refers to the pressure release
16
valve which will activate if hydrogen is given off too quickly (faster than the rate at which hydrogen
and oxygen combine within the battery). This can happen if cell is charged at too high a voltage
for a sustained period. VRLA batteries can be further classified as absorbed glass mat (AGM)
batteries and gelled electrolyte batteries (gel) (Reddy (2011)).
The Deka Synergy battery is manufactured by the East Penn Manufacturing Company, Inc. It is
essentially the same as their standard VRLA battery (called Unigy II , R which is an AGM-type
battery), with the exception of the electrode plates, which are enhanced with carbon.3 VRLA
batteries can develop crystalized lead sulfate deposits inside the negative plate, which shortens
battery useful life, when operated continuously in a partial-state-of-charge (PSoC) regime. Frequent
conditioning (or overcharge) cycles act to dissolve the sulfate deposits. Adding carbon to the
electrodes enables operation in a PSoC regime, as it greatly reduces sulfate deposit formation when
operated in a PSoC regime. Reducing sulfate deposit formation has the potential to increase battery
performance and lifetime (U.S. Department of Energy, 2012).
The UltraBattery is a lead-acid battery coupled with an asymmetric supercapacitor. An asymmetric
supercapacitor is comprised of a lead-dioxide positive plate and a carbon-based negative plate, while
a lead-acid battery is comprised of a lead-dioxide positive plate and a lead negative plate. Placing
the lead and carbon-based negative plates in parallel allows both electrodes to share the same
lead-dioxide positive plate. This is illustrated in Figure 4.
Figure 4: UltraBattery configuration (Lam et al., 2006).
The capacitor incorporated into the Ultrabattery helps to minimize heating and negative plate
sulfation at high-rate-PSoC cycling. This battery has shown good longevity when operated within
a state-of-charge (SoC) of roughly 40 to 60% (Coppin and Wood, 2011). This battery was devel-
oped by Australias Commonwealth Scientific and Industrial Research Organization (CSIRO), and
manufactured by East Penn Manufacturing Company.
3
Conversation with Chuck Mathias, East Penn.
17
1.4 Battery Operation
The project goal was to use the UltraBattery system solely to moderate fluctuations in PV power
output due to intermittent cloud cover and to use the Deka Synergy battery system solely to provide
time-of-day shifting.
It should be noted that the UltraBattery system can provide time-of-day energy shifting, just as
the Deka Synergy battery system can offer smoothing. However, each application requires a certain
balance of energy and power, which each battery is designed to meet. Shifting, for example, is
a service that benefits directly from each additional unit of energy that can be stored while only
requiring enough power to store and release that energy in the desired time window. Smoothing,
in contrast, is a service that benefits directly from each additional unit of power (up to some
proportion of the rated power of the PV installation), while only requiring enough energy to keep
from saturating the battery either at full charge or at empty.
The optimal temperaure for using stationary lead-acid batteries is between 20 C to 30 C. While
temperatures up to 50 C can be tolerated, high-temperature operation increases self-discharge and
reduces cycle life. Operation at cold temperatures is also problematicavailable storage capacity
drops as electrolyte temperature decreases (Reddy, 2011, pp.16.63-16.64). Therefore, all of the
containers at the PNM Prosperity site have air conditioners/heaters. When cooling is needed, the
targeted temperature is between 24 and 27 C. When heating is needed, the target temperature is
between 21 and 24 C.4
PNM targets operating the shifting battery between 30% and 70% SoC.5 In addition, both battery
systems undergo conditioning charges. These bring the SoC up to 100%, and are performed, on
average, once every two months.6 These charges help to mitigate lead sulfate crystal formation on
the battery cell anodes.
4
E-mail from Steve Willard of PNM, 11-Sep-2013.
5
Communication from Steve Willard of PNM, 17-Jan-2014.
6
E-mail from Simon Boland of Ecoult, 9-Sep-2013.
18
2 Evaluation Metrics
In this paper, we are interested in characterizing how the storage system is performs. The char-
acteristics of interest include: PV smoothing performance, round-trip efficiency, balance of plant
load quantity and variability, system availability, and operational requirement impact on system
life. Because there are various ways of quantifying these characteristics, it is necessary to define
how the calculations are performed. This is the purpose of this section.
2.1 PV Smoothing Effectiveness
PV system power-output variability is primarily a function of time of day, cloud cover, and season.
In this study, however, we are mainly interested in the rapid PV power-output fluctuations caused
by intermittent cloud cover. The term PV smoothing is used to denote the use of an electricity
storage system (ESS) to mitigate variable PV outputs rapid power fluctuations. These power
fluctuations are an unfavorable PV-system characteristic. At the feeder level, these fluctuations can
cause voltage flicker and feeder voltage excursions outside of the desired band. These fluctuations
can result in increased switching of tap changing transformers and capacitor banks, resulting in
premature failure of components. At the bulk grid level, PV variability requires additional operating
reserve to be set aside, and causes conventional generation to cycle more than otherwise. By using
an ESS to absorb or supply power at the appropriate times determined by an automatic control
system, PV power output can effectively be smoothed, lessening the burden on the system caused
by the rapid fluxes. Pumped-storage hydro (PSH), battery systems, flywheels, and compressed-air
energy storage (CAES) are some of the technologies used for energy storage.
The following metrics were developed to characterize the degree to which photovoltaic power is
effectively smoothed by an energy storage system:
1. Percent reduction in standard deviation of power (RSDP)
2. Percent reduction in standard deviation of ramp-rate (RSDR)
3. Max-min reduction
These three metrics are used in this study to quantitatively describe an energy-storage systems
ability to smooth the fluctuations of a given PV power profile. By using the solar irradiance cate-
gorization described in Trueblood et al. (2013), a determination of how well a given ESS performs,
the smoothing operation for various solar irradiance profiles can be made.
Mathematically, the variance of power can be defined as:
N 1
2 1 X
= (R(n) E[R(n)])2 (1)
N 1 n=0
where n = 0, 1, ..., N-1; E[R(n)] is the expected value of R(n); and N = length of R(n) in the
sample. The standard deviation of power is simply the square root of the variance of power.
The ramp-rate variance can be defined as:
N 1
2 1 X
= (Y (n) E[Y (n)])2 (2)
N 1 n=0
19
X(n+k)X(n) P
where n = 0, 1, ..., N-1; Y (n) = k
, and E[Y (n)] = 1/N Y (n). The SDRR is the square
root of the ramp-rate variance.
Finally, the max-min reduction can be defined as the reduction in:
Pmax swing = max|R(n)| (3)
A detailed discussion of these metrics can be found in Appendix B.
2.2 Round-Trip Efficiency
Round-trip efficiency is the efficiency of a single charge and discharge cycle, and can be measured on
a DC-to-DC or an AC-to-AC basis. The lower the storage systems round-trip efficiency, the higher
the value of the services provided by storage (here, smoothing and time-of-day shifting) needs to
be in order to compensate for the losses.
A key question is whether BoP load should be taken into account in calculating the round-trip
efficiency. Both the battery-system control equipment and the climate control for the battery
containers are AC-powered. They are not fed from the same DC feeder that connects the battery
systems to the grid-tied inverter, but from a separate, AC line. This BoP load therefore does not
show up as energy in or out of the batteries (as measured directly downstream of the inverter in
DC, or directly upstream of the inverter in AC), but it is captured by the Station Meter reading.
On one hand, showing efficiency excluding the BoP load is useful, as it isolates the battery string
efficiency. On the other hand, the BoP load should be included, both because it is a part of battery
operations, and because the efficiency measurement becomes a metric that can be improved through
BoP load reduction.
Figure 5 illustrates how the shifting battery has typically been employed at the PNM Prosper-
ity project to shift PV output. The battery charges during the morning and discharges in the
afternoon.
Storage-device efficiency can be calculated by simply dividing the amount of energy discharged by
the amount of energy charged.
Eout
= 100 (4)
Ein
For the PNM prosperity site, because BoP load is not included as part of the energy input mea-
surement, BoP load must be explicitly taken into account (Equation 5).
Eout
= 100 (5)
Ein + BoP load
Any difference between amount of energy charged and discharged must represent energy lost in the
charge/discharge cycle.
Eloss = Ein Eout (6)
Storage-system inefficiency is due to heat loss and any electrical, mechanical, or chemical losses
in the conversion and reconversion processes. Sources of such inefficiencies are electrical inverters,
20
600
PV Power
500 Smoothed Power
Primary Meter
Battery Power
400
BoP Load
300
Power (kW)
200
100
100
200
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Time (hours)
Figure 5: PV shifting, 26-Jun-2013 data.
chemical reactions in batteries, bearings and friction in mechanical storage systems, and resistance-
in-wire losses associated with high current.
Again, by taking the BoP load into account, it can be considered as contributing to the energy lost
in the round-trip cycle.
Eloss = Ein + BoP load Eout (7)
Expressing DC-to-DC efficiency excludes inverter losses, while the AC-to-AC efficiency accounts
for the associated inverter loss. By excluding the inverter losses, only inefficiency in the chemical
conversion and connected copper is represented. However, for a more holistic look at what the
system efficiency is, inclusion of the inverters is necessary. It is also necessary to examine these
quantities across different periods of time. Individual round trips, characterized by a single charge
and a single discharge cycle, tend to be more efficient than that of a year-round total charge vs.
total discharge cycle.
2.3 Balance-of-Plant Load
The Balance-of-Plant (BoP) load is an important consideration in calculations of storage system

efficiency.
At the PNM Prosperity project, BoP load is not withdrawn from the batteries instead, it is taken
from the grid, as AC power is needed for the BoP loads. These loads include air conditioning and
heating for the containerized battery strings, control systems, and data measurement systems.
21
The BoP load can either be directly measured, or indirectly calculated. The method we have
chosen for evaluating shifting battery efficiency in this report is direct measurement. At the PNM
Prosperity site, the BoP load is measured by a meter called the Station Meter. This meter is on a
low-voltage distribution circuit. According to PNM, it measures most, but not all, of the BoP load.
(There is an additional 1 to 2 kW load not on this same circuit, according to PNM.) While there
is the risk of under-counting BoP load, we know that we will not be including loads and losses at
the Prosperity site that are unrelated to the storage systems.
The other option is to indirectly calculate the BoP load. In this case, this can be done by taking the
Primary Meter reading and subtracting PV output (as measured at the PV meter on the AC side
of the circuit) as well as the Storage Battery output (also measured on the AC side of the storage
systems). While this calculation is more comprehensive, there is the risk that loads and losses not
related to the storage systems would be counted against storage system performance.
2.4 Availability
It is useful to understand how often the storage-system has been available. Because we are interested
in the storage system performance, it seems reasonable to exclude outages for reasons external to
the storage system (such as a system power outage, fault on the feeder connecting to the storage
system, etc.). Therefore, battery system forced outages and maintenance requiring the system to
be taken off-line were counted against the systems availability. Specifically, downtime would be
considered the length of time the storage system recorded any of the following:
1. An inverter fault,
2. A battery string fault, or
3. Any instance the system was taken off-line for maintenance or manual charging/discharging.
Note that this definition of availability does not require that there be any charge or discharge
commanded to the battery systemit measures the amount of time the storage system was available
to be used, whether or not a command to charge or discharge was given.
2.5 Estimated Remaining Battery Lifetime
For the UltraBattery USB-320/500 kW (as is installed at the PNM Prosperity site) under recom-
mended operating conditions, Ecoult estimates the total energy available before battery replacement
to be 2 GWh (Ecoult, 2013, p. 24).7 The energy output of the smoothing battery on a cloudy day
can be compared with this estimated lifetime energy available to determine on how many cloudy
days the battery system might be expected to help mitigate PV output fluctuations.
Instead of being tasked to smooth PV output, if the UltraBattery installed at the PNM Prosperity
site were given an automatic generation control (AGC) signal to follow its expected lifetime may
differ. A representative AGC signal can be used to calculate how much battery energy output
would be required, which can then be used to calculate an expected lifetime when following an
AGC signal.
7
Here, available energy means the total amount that the battery is capable of discharging before it fails. It is not
throughput, which would also count energy used to charge the battery.
22
For the Deka Synergy SSB-320/250 kW storage block, we assume that the overall battery throughput
expected lifetime is the same as that for the Unigy II battery.8 While the Unigy II battery is a
standard VRLA battery, the primary difference is that it needs to be operated near its maximum
SoC, whereas the Synergy battery was designed to operate in a PSoC regime. The Unigy II is rated
at 600 cycles at a 100% depth of discharge (East Penn, 2011). Each Deka Synergy Storage Block
has a total capacity of 1 MWh. As there are three blocks in parallel, the total capacity is 3 MWh.
The total energy available before replacement for the shifting battery system, therefore, should be
about 1.8 GWh.9
For the estimated remaining lifetime, we will use data from Ecoult on the monthly amount of
energy charged and discharged from each battery system, and compare the estimated GWh output
remaining with how much the battery is currently being utilized.
8
The Deka Synergy battery cell is the same as that of the Unigy II, except that the Synergy battery plates are
impregnated with carbon.
9
Calculation: 1 MWh capacity 3 storage blocks 600 cycles = 1.8 GWh
23
3 Analysis
This section contains results for the PV smoothing, shifting battery efficiency, BoP load character-
ization, availability, and battery estimated lifetime.
3.1 PV Smoothing Effectiveness
We begin by analyzing the smoothing function of the UltraBattery at the PNM Prosperity site. The
Prosperity site is fitted with various meters and measurement equipment in order to fully measure
and analyze the systems performance. The data used for this analysis is recorded at a rate of
1 sample per second. All data sets are one days worth of sampling, from sun-up to sun-down.
Characteristic days were chosen based on the methods described in Trueblood et al. (2013), and
span approximately 15 months worth of data collection.
It is helpful to introduce two metrics in order to help classify daily solar irradiance profiles. The
first is the Daily Clearness Index, and it is given as the ratio of solar energy measured on a given
surface to the theoretical maximum energy on that same surface during a clear sky day (Trueblood
et al., 2013, p. 35). It is described in equation 8.
measured solar insolation

Daily Clearness Index = CI = (8)
calculated clear-sky solar insolation
The second metric is described in equation 9. It is called the Daily Variability Index, and is
defined as the variability in measured irradiance, relative to the variability of the calculated clear
sky irradiance, with each quantified by the length of the irradiance versus time plot for the day,
where the curve length between two measurements is determined using a line segment. Typical
values for daily variability index range from 1 to 30 (Trueblood et al., 2013, p. 35).
length of measured irradiance plot

Daily Variability Index = V I = (9)
length of clear-sky irradiance plot
This analysis contains results for days that fit categories of high, moderate, and mild variability.
The clear and overcast variability classifications also described in Trueblood et al. (2013) will
not be used in this analysis. The purpose of smoothing PV power is to smooth large fluctuations,
so if the day is classified as clear, there is no reason to smooth it. This is because the PV power
output will be smooth, because its V I < 2, and the ESS will not need to be used. An overcast
day has a V I < 2 as well, so it is also unnecessary to analyze days that fall in this classification.
This analysis will focus on days with mild to highly variable irradiance.
A 60-second time period was used for all ramp-rate calculations, resulting in units expressed in
[W/minute].
High Variability
A day with high variability is defined as having a variability index of V I > 10. For more informa-
tion on this classification system, see (Trueblood et al., 2013). Figure 6 represents the unprocessed
PV power, the processed (smoothed) PV power, and the battery power on a characteristic day
24
with high variability. All days analyzed with high variability display similar, highly sporadic power
output.
600
PV Power
Smoothed Power
500 Battery Power
400
300
Power (kW)
200
100
100
200
9 10 11 12 13 14 15 16 17 18
Time (hr)
Figure 6: High PV variability, 15-Apr-2012 data.
Moderate Variability
A day with moderate variability is defined as having a variability index of 5 V I 10. Figure
7 represents the unprocessed PV power, the processed PV power, and the battery power on a
characteristic day with moderate variability. All days analyzed with moderate variability display
similar, moderately sporadic power output. Some of the days with moderate variability contain
large sections of power that is much lower than its clearness index would suggest. This indicates
heavy cloud cover for an extended period of time, and the battery system is not designed to smooth
these types of power discrepancies. They are too slow for the power system to track.
Mild Variability
A day with mild variability is defined as having a variability index of 2 V I 5. Figure 8
representing the unprocessed PV power, the processed PV power, and the battery power output on
a characteristic day with mild variability. All days analyzed with mild variability display similar,
slightly sporadic power output. Some of the days with mild variability contain large sections of
power that is much lower than its clearness index would suggest. This indicates heavy cloud cover
for an extended period of time, and the battery system is not expected or designed to smooth these
types of power discrepancies. They are too slow for the power system to track.
The results in Table 4 are given for ten sample days over the course of approximately 15 months.
Many of the results are quite consistent with one another, both across time and across metrics. It
appears that the data is quite consistent across the RSDP and RSDR from the beginning of 2013
until the last sampled day in July. However, older data from 2012 suggests that a parameter of the
25
600
PV Power
Smoothed Power
500 Battery Power
400
300
Power (kW)
200
100
100
200
9 10 11 12 13 14 15 16 17 18
Time (hr)
Figure 7: Moderate PV variability, 22-Jun-2013 data.
600
PV Power
Smoothed Power
500 Battery Power
400
300
Power (kW)
200
100
100
200
9 10 11 12 13 14 15 16 17 18
Time (hr)
Figure 8: Mild PV variability, 13-Apr-2012 data.
control algorithm may have been changed. This is suggested because ESS performance seems to
have increased significantly over time.
26
Table 4: Reduction in PV Output Variability Due to Smoothing Battery
4/13/12 5/12/13 6/21/13 6/28/13 7/5/13 3/6/12 6/22/13 4/15/12 2/11/13 5/27/13
Classification mild mild mild mild mild mod mod high high high
RSDP 32.03 58.82 60.38 65.86 55.22 17.93 60.3 30.48 72.03 63.01
RSDR 30.25 61.23 55.79 63.6 44.68 11.09 57.45 28.81 67.86 60.61
MM Power 32.39 68.78 63.19 69.13 78.21 0 62.78 22.46 76.85 73.14
MM Ramp Rate 41.1 71.21 69.37 72.52 75.76 0.02 64.4 36.54 82.33 73.38
Consistency across the two standard-deviation calculations and the max-min reduction is also seen.
This gives confidence in all of the metrics, and lends credibility to the analysis results. The outlier is
the date 6-Mar-2012, because it has poor performance. However, by inspecting the plot, it appears
that the ESS is offline during a large portion of the day. The time it spends offline contains high
variability in power, so the max-min calculation being near-zero is expected, as well.
Perhaps most importantly, note that the system appears to be consistent across the variability of
solar irradiance, as well. It does not tend to favor one type of variability over another. This implies
that the system can smooth the most variable, rapidly changing cloudy days about as well as it can
smooth the less cloudy or potentially partially cloudy days (i.e., thunderstorm moves in late in the
afternoon).
The ESS achieved significant reductions in the volatility of solar power output across all conditions
(mild-, moderate-, and high-variability days).
3.2 Shifting Battery Efficiency
The efficiency of the shifting battery was estimated by using the metered amount of power input
to and output from the battery system. One efficiency measure examines a single charge and
discharge cycle, or round-trip efficiency. Efficiency can also take into account the losses associated
with keeping the equipment on 24 hours a day, seven days a weekwe call this annual efficiency.
Both of these efficiency measures can include or exclude BoP load. They can also both be done on
either a DC-to-DC basis or on an AC-to-AC basis. These measures are all valid and help understand
battery efficiency from different perspectives. The DC-to-DC round-trip efficiency excluding BoP
load, for example, quantifies how the battery system at its most basic level is performing. This
measure reflects what is under the control of the battery maker/storage system provider. Including
BoP load and site losses is also valid, as this reflects a true cost of operating the system.
These calculations were made using data directly from the PNM Prosperity project. Table 5 contains
a summary of the results.
Table 5: Shifting Battery Efficiency

DC-to-DC No 89% 85%
It follows that in all cases, the annual efficiency is lower than the round-trip efficiency. One reason
27
for this is because the annual efficiency reflects losses due to battery self-discharge over time. In
addition, when BoP loads are considered, the annual efficiency measurement takes into account
BoP loads at night (for round-trip efficiency, only BoP loads and losses during the charge-discharge
cycle are taken into account). Self-discharge of the battery is not estimated in this report.
In addition, when calculating annual efficiency with BoP loads, then the more energy charged and
discharged over the year, the greater the efficiency. This is because BoP loads are relatively fixed,
and are added to the total amount of energy used to charge the batteries in the efficiency calculation.
In other words, using the battery system more spreads the fixed BoP load out over more MWh,
increasing the annual efficiency.
The 7% drop in round-trip efficiency going from DC-to-DC (including measured BoP) to AC-to-AC
(as measured at the storage systems meter on the AC-side, including measured BoP) is reasonable.
The inverter losses, which are probably on the order of 3.5% in each direction (AC-to-DC and
DC-to-AC),10 are considered in the AC-to-AC calculation, but not in the DC-to-DC calculation.
Two trips through the inverter would completely account for the lower round-trip efficiency in the
AC-to-AC measurement. (We note that with the PNM Prosperity configuration, inverter losses
apply to grid or PV energy that is used to charge the batteries, as well as energy discharged from
the batteries.)
Below is a discussion of each of the efficiency metrics and how the efficiency numbers were calcu-
lated.
DC-to-DC Efficiency, excluding BoP Load
The DC-to-DC efficiency, excluding BoP load, is an important metric, as it allows one to assess
how efficient the battery systems themselves are at storing and releasing energy.
Round-Trip Efficiency
Data from ten days (over a period from December 2012 through June 2013) were used to assess the
round-trip efficiency of the smoothing battery.
Figure 9 demonstrates the action of the shifting battery on one of these days, 23-Jun-2013. The
shifting battery, shown by the green line, first charges and then discharges later in the day.
Table 6 reports the efficiency of the shifting battery, as calculated from the total charge and dis-
charge energy. All of the days chosen feature a single charge-discharge cycle, in which the shifting
battery system is charged from the PV generation in the morning through early afternoon, and
then discharge takes place in the late afternoon or evening.
Over the eleven days reported on in Table 6, the average round-trip efficiency not including the
BoP load is 89%.
The sampled days show a range of round-trip efficiencies from 85% to 92%. The reason for this
rather large difference is likely because of differences in starting and ending SoC. While we chose
sample dates where the starting and ending SoC were the same or close together, any errors in the
data on SoC level would have a large impact. We believe that taking the average efficiency from
this set of data greatly reduces errors in calculating efficiency associated with SoC measurement.
The power lost in the conversion is approximately 10% of the total stored, and is attributed to the
10
The 3.5% losses for one-way conversion is based on the reported weighted efficiencies of similar-sized inverters in
a California Public Utilities Commission report (Malashenko et al., 2013).
28
600
PV Power
500 Smoothed Power
Primary Meter
Battery Power
400
BoP Load
300
Power (kW)
200
100
100
200
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Time (hours)
Figure 9: Shifting-batttery action, 23-June-2013 data.
Table 6: Round-Trip Efficiency over Sampled Days
Year: 2012 2013

Date: 12/1 12/3 12/15 1/14 3/22 3/23 3/24 6/19 6/23 6/26
Efficiency (%, ex-BoP load) 90 86 92 89 92 90 85 90 87 90
Charge Energy (kWh) 892 884 811 799 813 820 788 880 925 898
Discharge Energy (kWh) 801 763 747 711 745 738 668 796 801 807
Max Load Power (kW) 39.3 92.1 49.0 94.9 66.4 89.6 86.9 63.4 61.8 59.2
Average Load Power (kW) 23.3 22.4 28.5 33.4 21.0 24.7 25.6 22.5 21.4 23.7
Std. Dev. Load Power 6.9 5.9 6.8 7.2 10.9 5.3 6.7 8.5 8.3 9.1
heat and conversion processes mentioned previously. Overall, it appears that the shifting battery
performs as commanded. It shifts energy stored at a rather slow rate early in the day, and releases
it at a higher rate later in the day, when the load is expected to be greater.
Annual Efficiency
One year of hourly average power output (and input) data was obtained from the PNM Prosperity
data server. This data measures the average power to and from the shifting battery, and is measured
on the DC side of the system. The data is from 1-Jul-2012 through 31-Jun-2013.
It is also important to insure that the battery system was at the same SoC at the beginning
and ending points selected. If the battery system is at higher SoC at the endpoint, then this
energy that went into increasing the state of charge will appear to be energy that was lost in the
efficiency calculation. Likewise, if the battery system is at a lower SoC at the endpoint, then energy
simply withdrawn from the battery will be counted as energy that increases the battery systems
29
efficiency. It is possible to adjust the calculation for the situation where the SoC is not equal at
the start and end times, but days were chosen that follow the above criteria so that no estimation
is necessary.
According to the PNM data, the shifting battery was at a 60.9% SoC on 10-Jul-2012 at 1pm, and
was also at a 60.9% SoC on 26-Jun-2013 at 3pm. Over this time frame, 136 MWh were discharged
from the shifting battery, and 161 MWh were used to charge the battery. The resulting efficiency
is approximately 85%.
Data provided by Ecoult covering the period of July 2012 through the end of June 2013 was utilized
to calculate that 138.5 MWh were discharged and 162 MWh were used to charge the battery system
over this timeframe. The resulting efficiency is 86%. Figure 10 shows the efficiency by month based
on the Ecoult data.
100
95
90
Efficiency (%)
85
80
75
Aug12 Sep12 Oct12 Nov12 Dec12 Jan13 Feb13 Mar13 Apr13 May13 Jun13 Jul13
Month
Figure 10: Shifting-battery efficiency, by month.
We note that when summing energy inputs and outputs over a month, errors in efficiency estimates
can result from the fact that the SoC at the months end may differ from that at the beginning of the
month. This can give artificially high or low efficiency estimates. At the same time, shifting-battery
efficiency shows a clear downward trend.
While these two estimates are close, we believe that the 85% efficiency estimate is probably the
more accurate, given that we were able to match the beginning and ending SoC when performing
the calculation.
Note the downward trend shown in figure 10. It is expected that the battery efficiency will tend
to decrease over extended periods of service time. This is a widely accepted phenomena, explained
in part by the lead-acid batterys tendency to have a decreased specific power over time in service
(Reddy (2011) p. 15.9 - 15.18). Lower specific power is related to increased internal resistance of the
30
battery. If the internal resistance of the battery is increased, then the I 2 R loss associated with the
battery is also increased (assuming constant current). This power loss makes the electrochemical
conversion process less efficient over time, as shown in figure 10.
31
DC-to-DC Efficiency, including BoP Load
The DC-to-DC efficiency, including measured BoP load, is designed to take into account battery-
container climate-control load and battery control-system load. It is reasonable to penalize battery
system efficiency for these loads, as the battery system would not be able to operate without them.
In addition, it is possible to decrease these loads. Including BoP load in an efficiency metric ensures
that it is measured, and provides incentive to reduce this load.
The reason we term these loads measured BoP load is that they are measured at a meter called
the station meter at the PNM Prosperity project. The station meter is behind the primary meter,
and is on a distribution circuit. The station meter is behind the primary meter on the distribution
circuit, and includes the battery container climate control as well as the battery control system.
It does not include the data acquisition system.11 Since the shifting-battery efficiency is being
analyzed here, the station meter load must be weighted by 75%, as we are interested in the load
supporting the shifting battery alone.12
Using the same ten days, the measured BoP load is taken during the charge-discharge cycle,13 and
used 75% of this amount to reflect the amount of BoP load attributable to the shifting battery.14
We then added this BoP load to the total amount of energy used to charge the battery when
calculating the efficiency. The BoP loads, as well as the resulting individual daily efficiencies, are
in Table 7.
Table 7: Round-Trip Efficiency over Sampled Days, Including BoP Load
Year: 2012 2013

Date: 12/1 12/3 12/15 1/14 3/22 3/23 3/24 6/19 6/23 6/26
Efficiency15 (%) 84 80 87 82 85 84 78 84 80 83
Charge Energy (kWh) 892 884 811 799 813 820 788 890 925 898
Discharge Energy (kWh) 801 763 747 711 745 738 668 796 801 807
BoP Load16 (kWh) 63 64 46 70 65 58 71 71 71 70
The average efficiency including BoP load for a round-trip is 83%.

Annual Efficiency
The average daily measured BoP load is 140 kWh, based on measurements available from 1-Sep-2012
through 30-Jun-2013. Because we are interested in 75% of this amount, this gives us a measured
BoP assigned to the shifting-battery system of 105 kWh/day. Taking this to be the average load
over the period used for our DC-to-DC efficiency calculation that excludes BoP load, this would
amount to 37 MWh.17 Adding this amount to the energy used to charge the battery yields a
11
E-mail from Steve Willard on 23-Oct-2013.
12
The station meter measures the load required to support all eight battery containers. We weight by 75% because
there are six shifting battery containers out of the total of eight.
13
Here, we are interested in round-trip efficiency. We therefore want the BoP load only while the charge-discharge
cycle is being performed. BoP load at night is not part of this measurement.
14
As six of the eight battery containers are for the shifting battery, we can approximate the BoP load due to the
shifting battery by multiplying the total BoP load by 75%.
15
Including BoP load.
16
Attributable to shifting batterywhich is 75% of the total measured BoP load.
17
The period from 10-Jul-2012 through 26-Jun-2013 contains 351 days. 140kW h 351days = 49M W h.
32
shifting battery round-trip efficiency of 69% over this same 10-Jul-2012 through 26-Jun-2013 time
period.
The drop in efficiency (including measured BoP load) from 83% on a charge-discharge cycle basis to
69% on an annual basis is significant. This drop in efficiency is due to accounting for the measured
BoP load not only during the charge-discharge cycles, but also at times when the shifting battery
is not performing work (such as at night). The more frequently and heavily the battery system
is used, the higher the annual efficiency will be, since the fixed BoP load will be spread out over
a large volume of energy charged/discharged. In addition, battery self-discharge over time acts to
further decrease the storage system efficiency when measuring continuously over time.
AC-to-AC Efficiency, including BoP Load
This measurement of efficiency is based on the AC meter nearest the battery storage systems. Since
this meter measures the power flows to and from both battery systems, it is necessary to adjust the
power flows to remove the impact of the smoothing battery system (since here we are interested in
the efficiency of the shifting battery system). For this adjustment, we simply subtracted out the
measured DC flows to and from the smoothing battery system. For the Shifting Battery BoP Load,
the total BoP load is multiplied by 75%, since six of the eight battery containers are for the shifting
battery.
The following formula is used for this efficiency calculation:
Adjusted AC Discharge Energy
AC-to-AC Efficiency = (10)
Adjusted AC Charge Energy + Shifting Battery BoP Load
The AC-to-AC Round-Trip Efficiency is calculated using equation 10. To obtain a round-trip
efficiency measure, the Adjusted AC Discharge Energy and the Adjusted AC Charge Energy are
measured from the beginning to the end of a charge-discharge cycle, taking care to insure that the
SoC at the beginning and end are the same or close to each other. The BoP is measured over
this same charge-discharge cycle, and is multiplied by 75% to find the amount attributable to the
shifting battery. Table 8 shows the resulting calculated BoP, and round-trip efficiency, for the same
ten days examined earlier.
Table 8: AC-to-AC Round-Trip Efficiency of Shifting-Battery System
Year: 2012 2013

Date: 12/1 12/3 12/15 1/14 3/22 3/23 3/24 6/19 6/23 6/26
Efficiency (%) 90 86 92 89 92 90 85 90 87 90
Efficiency, AC-to-AC18 (%) 78 76 78 75 75 76 71 77 76 77
AC Charge Energy19 (kWh) 924 907 883 801 799 837 812 915 942 930
AC Discharge Energy20 (kWh) 772 735 723 656 648 680 627 758 770 773
Shifting Battery BoP Load (kWh) 63 64 46 70 65 58 71 71 71 70
18
AC Efficiency of the shifting battery system, as measured in front of the storage systems, net of BoP load
attributable to the shifting batery system.
19
The smoothing battery charge volumes were subtracted from the total AC charge energy to obtain the AC charge
attributable to the shifting battery system.
20
The smoothing battery discharge volumes were subtracted from the total AC discharge energy to obtain the AC
discharge attributable to the shifting battery system.
33
The average AC-to-AC efficiency, including calculated BoP load, over the sampled days is 76%.
Annual Efficiency
Measuring from the AC Battery Meter (which measures power flows to and from both battery
systems), we see that 194,500kWh was used to charge the two battery systems, and 133,300 kWh
was discharged from the two battery systems over the 350 day period from 10-Jul-12 and 26-Jun-
13.
Since here we are interested in isolating flows to and from the shifting battery, we adjust the charge
and discharge numbers by subtrating power flows due to the smoothing battery. We then get an
adjusted charge of 173,100 kWh, and an adjusted discharge of 123,800 kWh.
We then add 75% of the measured BoP over this same period, which was 36,800 kWh (or 140 kWh
per day), to the adjusted charge in order to reflect the total energy inputs. The annual average
AC-to-AC efficiency including BoP load, then, is 59%.21
3.3 Balance-of-Plant Load Characterization
In the shifting battery efficiency analysis, measured BoP loads were used. This measurement at the
PNM Prosperity site is at what they have called the Station Meter. This meter is on a distribution
circuit at low voltage. PNM states that it captures most of the BoP load, but that there is a device
drawing 1 to 2 kW of power that is not on this same circuit. Therefore, the measured BoP load may
understate actual BoP load. This measured BoP load attributable to the shifting battery averaged
4.4 kW per hour for the period from 29-Aug-12 through 30-Jun-13.
For that same period, the calculated BoP load attributable to the shifting battery system averaged
14 kW per hour, or more than three times greater than the measured BoP load. Here, the term
calculated balance-of-plant load refers to the difference between energy recorded at the Primary
Meter and the AC meter for PV output and the AC Battery Systems Meter. (It is calculated
becuase it is determined by a calculation rather than a direct meter reading). This difference must
be the result of BoP load and site losses.
While the calculated BoP is a more comprehensive measure, we did not use it for the efficiency
calculations. Our goal was to evaluate the performance of the storage resource, not the entire
site. Since the calculated BoP is significantly larger than the measured BoP, it may be accounting
for losses in the site that are not related to the battery systems. While this probably merits
investigation, we did not feel it was fair to penalize the storage systems for losses that are likely
taking place elsewhere in the Prosperity site.
3.4 Availability
Ecoult has been keeping data since November 2011 on storage-system downtime. The monthly
system availability, measured as the number of hours the system is available in a month divided by
21 AdjustedDischarge 123, 800kW h
= = 59%
AdjustedCharge + BoP 209, 900kW h
34
the total number of hours in that month, is shown in Figure 11. Based on this data, the availability
from November 2011 through July 2013 was calculated to be approximately 91%.
Figure 11: Availability (source of data for calculations: Ecoult)
We should keep in mind that because the storage system here has been used for PV shifting and
smoothing, it has been used almost exclusively during the day. Maintenance has generally been
performed in the evening, when the system has not been asked to respond. The availability number
here, therefore, does not reflect the percentage of time that the battery system was commanded to
perform and not available toit reflects whenever the battery system was off-line, whether or not
the system was needed. If availability instead is viewed as the percentage of time the battery was
needed but was off-line, the availability would be significantly higher.
3.5 Estimated Remaining Battery Lifetime
Integrating, with respect to time, the battery data (supplied by Ecoult) for both the total smoothing-
battery and total shifting-battery output, we find that from November 2011July 2013, the smooth-
ing battery has discharged approximately 32 MWh, and the shifting battery has discharged approx-
imately 156 MWh.
Based on the estimated total lifetime output of the two batteries, through July 2013 the smoothing
battery has expended about 2% of its total lifetime resource, whereas the shifting battery has
expended roughly 9% of its total lifetime resource.
35
While these calculations are rough, they provide us with a sense of how hard the batteries are being
utilized, and how much longer they might last if they are required to function similarly to what they
are currently. The smoothing-battery system, compared with its capability, is being tasked at a low
level. Based on its lifetime discharge capacity, it could continue to be operated as it has historically
for roughly another 100 years.22 (To be clear, other factors will likely prevent the batteries from
operating this long. The point made here is that compared to their lifetime discharge capacity,
the batteries are being utilized at a low level.) The shifting battery, however, has been used more
heavily, and could be expected to last another 12 years if required to perform similarly to what it
has been doing.23
If the shifting battery were used each day for 0.85 MWh of output (85% of recommended total
useable capacity), then the battery would be expected to last for another 5 years, approximately.24
If the smoothing battery were used for frequency regulation, as opposed to smooth out the variability
of PV output, the system would have a much higher utilization. Assuming that following the AGC
signal caused an output of 1.0 MWh each day,25 the smoothing battery would also be expected to
last for another 5 years.26
22
Smoothing battery had an average output of 1.6 MWh/month from Aug-12 through Jul-13. Given a 2000 MWh
maximum estimated output, and given that about 32 MWh have already been output, this would yield a total
additional lifetime of 100 years.
23
Shifting battery had an average output of 11 MWh/month from Aug-13 through Jul-13. Given a 1800 MWh
maximum estimated output, and considering that about 156 MWh have already been output, this would yield a
total additional lifetime of 12.4 years.
24
1800 MWh total lifetime 156 MWh already output = 1644 MWh remaining. 1644 MWh/(0.85 MWh/day) =
1934 days of life remaining at 0.85 MWh/day. 1934 days/ (365 days/year) = 5.3 years remaining.
25
Please see Appendix C for more information on this supposition.
26
2000 MWh total lifetime 32 MWh already output = 1968 MWh remaining. 1968 MWh/(1.0 MWh/day) =
1968 days of life remaining at 1.0 MWh/day. 1968 days/(365 days/yr)=5.4 years remaining.
36
4 Recommendations for Battery and Photovoltaic Systems
The following observations and suggestions apply to the PNM Prosperity project, as well as other
similar projects that are planned for the future. Some observations are based on this reports
analysis; others are based on more general project considerations, as well as on putting the storage
and PV plant into the broader context of the bulk grid.
PV Smoothing
When using the battery system to smooth PV output, it would make sense to also take the BoP load
into account in the smoothing algorithm. The BoP load is highly variable due to the air conditioning
and heating load. The battery systems were installed, in part, to reduce PV output variabilitybut
in addition to doing this, they introduce a new variability of their own. Even if primarily tasked
with PV smoothing, the battery smoothing algorithm could be adjusted to take both PV output
and battery system BoP load into account. Another option is smooth the feeder voltage, which
would, in effect, take BoP load movements into account (as well as taking into account load swings
on the feeder itself).
A broader issue is whether PV smoothing is the most valuable use for the smoothing battery.
Instead, it could be used for frequency regulation (possibly in combination with PV smoothing).
This would increase utilization, and perhaps provide a higher value-added to the grid. It should be
kept in mind that the storage systems are merely co-located with the PV arraythey do not have
to perform either PV smoothing or shifting. They can be used for other applications, such as those
that provide greater benefit to the bulk grid. Responding to an AGC signal instead of PV smoothing
is discussed below in the Battery Lifetime and Recommended Applications subsection.
Battery Lifetime and Recommended Applications
Battery cells have a finite lifetime. While that lifetime depends on multiple factors, two of the
most important are the cumulative battery output and battery cell age. Storage-resource owners
would like to extend the lifetime of their resource as much as possible. A tendency to use that finite
resource sparingly is not surprising.
However, it is possible to have too much conservation. If battery cells have a shelf-life of 20 years,
for example, and the batteries are being operated such that they would have a lifetime of 100 years
based on usage, then they will still fail after 20 years. At a minimum, it would be better to operate
them such that their lifetime based on usage matches their shelf-life.
In areas that generally have clear skies, such as New Mexico, one would expect that storage systems
designed for smoothing PV output may be used sparingly. This does appear to be the case. This
does not mean that the PV smoothing application has little value. It does mean, however, that the
storage resource may be underutilized.
Having the smoothing battery follow an AGC signal means that it can be called upon to operate
24-hours a day, as opposed to only during daylight hours when used for PV smoothing. It can be
used no matter what the weather. In addition, when performing PV smoothing, there can be a
momentary drop in customer load at the same time there is a momentary drop in PV outputin
which case it would be better (from the overall grids perspective) not to inject power from storage
to compensate. If an AGC signal is followed instead of PV smoothing, this counterproductive action
would not be taken. Our recommendation, therefore, is that frequency regulation be considered as
an additional grid service.
37
PV Shifting
When using the shifting battery to shift PV output forward in time to better coincide with peak
system load, one must keep in mind that this comes at a cost. Some of the energy generated will be
lost due to inefficiencies in the round-trip storage cycle. The amount of energy lost could be as little
as 10%, or as great as 35%, depending on how the round-trip efficiency is calculated (discussed below
in the Battery System Efficiency, System Configuration, and Balance-of-Plant Load subsection).
One question is whether the benefits to the system of shifting power output forward a few hours
are worth the cost of losing some PV output to round-trip storage cycle inefficiencies. A production
cost model analysis could help answer this question.
When considering a storage system to shift PV output, a partial substitute worth evaluating is that
of using a tracking PV array. Installing a tracking PV array results in higher PV outputespecially
in the afternoon (Trueblood et al., 2013)at the expense of additional capital and maintenance
costs. Installing a tracking PV array can therefore be thought of as a partial substitute to adding
a shifting storage system to a fixed PV array. The PV array at the PNM Prosperity site is fixed
converting it to a tracking array at this point is not a realistic option. However, for future projects,
the costs and benefits of a fixed PV array plus shifting storage system versus a tracking PV array
can be compared.
The incremental cost of making a fixed PV array tracking can be compared with the cost of installing
a shifting battery system. As for the benefits, the utility to the grid of a tracking PV system versus
a fixed PV system plus shifting battery can be compared by using a production cost model that uses
the power production profile of each as an input. The annual production cost (from conventional
power plants) difference between the two scenarios will reflect the difference in value to the bulk
grid between the two options.
Of course, the shifting battery at the PNM Prosperity site does not have to be used to shift PV
output. Instead, it could supply a block of power when it is most needed, and charge when energy
is cheapest (i.e., arbitrage). This application may have higher value than PV shifting, because
the battery could charge at night, when energy is least expensive. Using the shifting battery to
provide spinning reserve is another option that may add more value than PV shifting, and is worth
study.
Battery System Efficiency, System Configuration, and Balance-of-Plant Load
Battery system round-trip efficiency, measured on a DC-to-DC basis, was found to be 89%. Includ-
ing BoP load, the DC-to-DC round-trip efficiency drops to 83%. When measuring on an AC-to-AC
basis (which factors in inverter losses and BoP load), the round-trip efficiency drops to around
76%.
While the round-trip measurements are in-line with our expectations, the annual efficiency estimates
including BoP losses are lower than expected. The reason for this is that while the shifting battery
is being used for a fraction of the time, there is BoP load the entire time under consideration. In
other words, BoP load when the battery is not being used (every night, as well as on days the
battery is not dispatched) counts against its annual efficiency. Operating the battery system more
often would improve this efficiency, as would drawing less BoP load when the battery system is not
in use.
Future projects may be able to reduce the level of losses due to BoP load and inverter losses exhibited
here. Orienting the battery containers East-West instead of North-South, exposing less wall area
38
to the harsh West exposure during the afternoon hours, would have reduced the air-conditioning
load for the PNM Prosperity project, and can be taken into account for future projects. Locating
the energy storage containers underground may also be a possibility for future projects, in order to
decrease temperature regulation costs. In addition, the number of trips through inverters could be
reduced (when shifting or smoothing PV power) by placing the PV array and battery systems on
a common DC bus.
To store PV-generated energy, the PV generation must be first converted to AC at the PV array,
and then be converted back to DC to charge the battery-storage system. Therefore, PV-generated
energy must pass through inverters three times in order to be stored and put on the grid (once at
the PV array, a second time to charge the battery, and a third time to discharge the battery to the
grid).
If the PV array and battery-storage systems were connected on a DC bus, PV power would only
need to go through an inverter once, even if the battery were used to shift PV power. Assuming
an average inverter efficiency of 96.5% for one-way conversion,27 these two additional trips amount
to nearly 7% lower efficiency (two additional passes through an inverter, at about 3.5% losses each
pass) to store PV energy than would be the case if the systems shared a DC bus.
While the PNM Prosperity project planners would have liked to have connected the two systems
via a DC bus, there were technical issues preventing this. For future projects, it would be preferable
to plan for the PV and battery resources at the same time, in order to insure that they can be
connected via a DC bus. Even if the storage system were to be used to respond to an AGC signal, it
could charge during the day from the PV generation in DC. Charging directly from a DC resource
would avoid an extra trip through the inverter, reducing losses.
Finally, the large difference between measured BoP load and calculated BoP load plus losses war-
rants investigation, in our view. While this difference may not be due to the storage systems, the
site losses are real (assuming there is no systematic data error). A site loss / BoP load audit may
uncover the root cause of the losses.
Smoothing- vs. Shifting-Battery System Selection
As the smoothing battery (UltraBattery) can both perform high ramp-rate smoothing and can store
as much energy as the shifting battery (Deka Synergy), it is the more versatile of the two. When the
PNM Prosperity storage project was built, the UltraBattery cost significantly more to manufacture
than the Deka Synergy battery. Since then, East Penn has been able to produce the UltraBattery
more efficiently, reducing the cost differential between the two batteries. Additional experience in
producing the UltraBattery may further reduce the cost.
We believe that the additional capability an UltraBattery system can provide allows it to provide
greater benefits to the grid than a Deka Synergy-based system. How much more depends on the
specifics of the grid where it will be installed. In a number of the power grids studied at Sandia
National Laboratories, reserve provisionespecially regulating reserve provisionappears to be more
valuable than time-of-day shifting.28 Given that storage value depends on other system factors
(such as the conventional generating fleet, amount and variability of renewables, load variability,
etc.), we recommend evaluating storage for the specific grid under consideration. The UltraBattery
systems capability to do both smoothing and shifting may well justify the additional cost.
27
Estimate derived from Malashenko et al. (2013), pp. 89.
28
For an example of this, see Ellison et al. (2013).
39
5 Conclusions
The Public Service Company of New Mexico (PNM) Prosperity electricity storage project consists
of a 500 kW/350 kWh energy-smoothing battery and a 250 kW/1 MWh energy-shifting battery,
and is co-located with a 500 kW solar photovoltaic (PV) resource. The site demonstrates its ability
to accomplish both smoothing and energy-shifting tasks as required. It is dispatchable, acting as a
distributed generator when commanded. The smoothing functions mitigate high-frequency power
swings when cloud cover rapidly changes the PV sites output.
In order to determine the degree to which the battery smooths the output, a set of metrics were de-
veloped. These metrics compare the pre-smoothed output of only the PV site to the post-smoothing
output of the battery and PV site combined by a statistical relationship of the fluctuations of their
power. This allows for a quantitative demonstration of the systems effectiveness.
Two factors are of primary interest when evaluating a shifting battery: the ability to charge and
discharge in blocks as commanded, and the round-trip efficiency. As no problems were reported
with the system charging and discharging as commanded, round-trip efficiency was chosen as the
metric to evaluate smoothing battery performance.
Because the shifting battery acts as a dispatchable energy source as commanded without fault,
the focus of the analysis of this battery system shifts to the efficiency with which it dispatches
the energy. The round-trip efficiency of the charging and discharging cycle is studied, across two
different time horizons. The results from this study show that the battery system can charge
and discharge relatively efficiently for one cycle, but over the course of a year, is significantly less
efficient. The reason for this is that the shifting battery is only being dispatched for a small portion
of the entire time horizon, during which there is a BoP load. To clarify, the BoP load is present
when the battery is not being used (all night, as well as on days the battery is not dispatched),
which decreases its annual efficiency. Table 9 illustrates the results of the shifting battery efficiency
evaluation. The two ways to increase this efficiency are: dispatch the battery more frequently, and
decrease the BoP load.
Table 9: Shifting battery efficiency.

DC-to-DC No 89% 85%
Various suggestions are discussed in order to increase efficiency. Placing the storage and PV array
on the same DC bus would likely improve the AC-to-AC efficiency, since the PV energy would avoid
one extra trip through the AC inverter. Also, an East-West instead of North-South orientation of
the battery containers may increase efficiency by exposing less wall area to the harsh west exposure
during the afternoon hours. Taking this a step further would be a shading canopy for the battery
containers, for an additional increase in efficiency (passive cooling).
While the battery system is conveniently co-located with a PV array, it is not limited to performing
actions related to the PV output. It is able to provide additional value by performing other applica-
tions, such as the following of an automatic generation control (AGC) signal. Using the smoothing
battery to follow an AGC signal will allow it to be useful over a full 24-hour cycle, instead of only
40
the days where clouds cause intermittent output on the PV array. As discussed, the battery can
be utilized more often than it is currently, making it more valuable to the utility (vs. standing by).
Also, following the AGC signal is useful by balancing load and generation on the grid; smoothing
PV output can be counterproductive if there is a simultaneous drop in load and PV output.
For the shifting battery, using the system to dispatch power when most needed and withdraw power
from the grid when it is cheapest may provide more value than simply shifting the PV output (i.e.,
arbitrage). To add value shifting energy, the system benefit of shifting off-peak energy to on-peak
energy must outweigh the cost of losing energy to storage round-trip inefficiencies. Decoupling the
shifting battery from the PV output allows it to charge at night or during the day, allowing the
operator to charge the battery at a time when energy is the least expensive, and sell it when it is
worth the most. Using the battery as spinning reserve may also be beneficial to the utility.
This project was partially funded by the American Recovery and Reinvestment Act, in conjunction
with its partners: University of New Mexico, Northern New Mexico College, Sandia National Labo-
ratories, East Penn Manufacturing Company, and Electric Power Research Institute. As one of the
first distribution-scale, grid-connected demonstration systems, the PNM Prosperity site continues
to provide valuable information for future storage projects of this scale.
41
Bibliography
Reddy, Thomas B., editor Lindens Handbook of Batteries Fourth Edition, McGraw Hill, 2011.
Coppin, Peter and John Wood. Ultrabattery Storage Technology and Ad-
vanced Algorithms at the MW-Scale EESAT Conference Proceedings, 2011,
http://www.sandia.gov/eesat/2011/papers/EESAT 2011 UltraBattery Storage Coppin Wood.pdf,
accessed December 3, 2013.
East Penn Manufacturing Co., Inc. Deka Unigy II Modults brochure,
http://www.dekabatteries.com/assets/base/0919c.pdf, accessed December 3, 2013.
Ecoult. Ecoult Megawatt Scale Energy Storage brochure, http://www.ecoult.com/technology/brochure-
downloads/, accessed September 5, 2013.
Ellison, James F., Dhruv Bhatnagar, Nader Saaman, and Chunlian Jin. NV Energy Electricity
Storage Valuation: A Study for the DOE Energy Storage Systems Program. Sandia National
Laboratories, SAND2013-4902, June 2013.
Kay, Steven M. Fundamentals of Statistical Signal Processing, Volume III, 1st ed., Prentice Hall,
April 5, 2013.
Lam, L.T. and R. Louey. Development of ultra-battery for hybrid-electric vehicle applications. Jour-
nal of Power Sources, vol. 158, 2006.
Malashenko, Elizaveta, Stephen Appert, and Wendy al-Mukdad. Advanced Inverter Technologies
Report. Grid Planning and Reliability, California Public Utilities Commission (CPUC), January
18, 2013.
Proakis, John G. and Dimitris K. Manolakis. Digital Signal Processing, 4th ed. Prentice Hall, April
7, 2006.
Trueblood, Chris, Steven Coley, Tom Key, Lindsey Rogers, Abraham Ellis, Cliff Hansen, and Eliza-
beth Philpot. PV Measures Up for Fleet Duty. IEEE Power and Energy Magazine, February 20,
2013.
U.S. Department of Energy, Office of Electricity Delivery and Energy Reliability. Energy Storage
Program. Carbon-Enhanced Lead-Acid Batteries. http://energy.gov/sites/prod/files/CELA.pdf,
accessed December 3, 2013.
U.S. Department of Energy. Recovery Act Smart Grid Programs, Project Information.
http://www.smartgrid.gov/recovery act/project information, accessed December 3, 2013.
Kay, Steven M., Fundamentals of Statistical Signal Processing, Volume II: Detection Theory, 1st
ed., Prentice Hall, February 6, 1998.
42
A ARRA Energy Storage Projects
This section highlights the funded projects to provide a context from which to consider the individual
project discussed in this report. The Smart-Grid Energy-Storage Demonstration Projects are being
managed by the National Energy Technology Laboratory for the DOE Office of Electricity Delivery
and Energy Reliability. The information below was primarily derived from project fact sheets
contained on the smartgrid.gov website (U.S. Department of Energy, 2013).
A.1 Grid-Connected Demonstration Projects
PNM Prosperity Project

ARRA Funding: $2,305,931
Total Project Funding: $6,113,433
PNM installed a 500 kW/1 MWh advanced lead-acid battery and a 250 kW/500 kWh advanced
lead acid with supercapaitors battery with a 500 kW solar PV plant to create a dispatchable,
distributed generation resource. This hybrid resource provides simultaneous voltage smoothing and
peak shifting through advanced control algorithms and switches between two configurations, end-of-
feeder and beginning-of- feeder. Data collection and analysis produce information for a wide range
of applications including grid upgrade deferral. The project has also yielded modeling tools used to
optimize battery-system control algorithms and further the understanding of feeders with storage
and distributed generation. The site is located in southeast Albuquerque.
Goals/Objectives
Demonstrate that intermittent, renewables-based, distributed generation and storage can mit-
igate voltage-level fluctuations and enable peak shifting
Quantify and refine performance requirements, operating practices, and costs associated with
the use of advanced storage technologies
Achieve 15% or greater peak-load reduction through a combination of substation-sited PV
and storage
East Penn Lyons Station
Project Description
East Penn Manufacturing will design and construct an energy storage facility consisting of an
array of UltraBatteryTM modules integrated in a turnkey battery energy storage system (BESS).
In addition to the UltraBatteriesTM , the BESS will include a power conditioning system, a master
programmable controller, and a battery monitoring system. The completed energy-storage system
will be designed to sell up to 3 MW of frequency regulation to PPL EnergyPlus, a designated load-
serving entity within PJM. In addition to frequency regulation, the system will provide demand-
management services to Met-Ed during specified peak power periods. These services will provide
up to 1 MW for 14 hours to meet the requirements of PA Act 129. The UltraBatteryTM is uniquely
suited to these applications because it was designed for high-rate PSoC cycling.
43
The system is sized to maintain the batterys SoC between 70% and 30% for a maximum 40%
depth of discharge for continuous regulation services. The UltraBatteryTM is a hybrid energy-
storage device that combines an asymmetric ultracapacitor and a lead-acid battery in one unit cell.
The UltraBatteryTM is expected to provide the same benefits as lead-acid battery systems, such as
low initial cost, full recyclability, plus increased cycle life by incorporating ultracapacitor technol-
ogy within the battery. To demonstrate modularity and portability, self-contained, containerized
UltraBatteryTM system will be designed and included as a subset of this project.
Goals/Objectives
Integrate advanced energy storage technology into an existing utility grid
Demonstrate the economic and technical viability of an UltraBatteryTM BESS for frequency
regulation ancillary services and demand management
Establish the cost of the UltraBatteryTM and all of the controlling power electronics required
for a utility grid management application
Duke Energy Notrees
ARRA Funding: $21.8 M
Total Project Funding: $43.6 M
Project Description
The Notrees Project will analyze and discern how, when integrated with wind power, energy storage
can compensate for the inherent intermittency of this renewable power generation resource. Incor-
porating both existing and new tools, technologies, and techniques, this demonstration project will
provide valuable information regarding wind energy storage and serve as a model for other entities
to adapt and replicate. The energy-storage system will be designed and constructed using fast
response, advanced lead-acid batteries configured to provide 36 MW output capabilities with a
storage capacity of 24 MWh.
Goals/Objectives
Store energy during non-peak generation periods and re-issue the power to meet demand
Quantify the value of wind storage
Demonstrate the reliability and dispatchability of wind storage
Use the storage system for system balancing
Determine if energy storage solutions are commercially viable to support wind generation
Southern California Edison Tehachapi
Project Description
The Tehachapi Wind Energy Storage Project will be located at Southern California Edisons (SCE)
Monolith Substation in Tehachapi, California, and the 8 MW 4 hr (32 MWh) BESS and 2 4
MW/4.5 MVA smart inverters will be housed in a 6,300-square-foot facility. The project will evalu-
ate the performance of the BESS to improve grid performance and assist in the integration of large-
scale variable-energy-resourced generation. Project performance will be measured by 13 specific
operational uses: providing voltage support and grid stabilization, decreasing transmission losses,
44
diminishing congestion, increasing system reliability, deferring transmission investment, optimizing
renewable-related transmission, providing system capacity and resources adequacy, integrating re-
newable energy (smoothing), shifting wind-generation output, frequency regulation, spin/non-spin
replacement reserves, ramp management, and energy-price arbitrage. Most of the operations ei-
ther shift other generation resources to meet peak load and other electric-system needs with stored
electricity, or resolve grid stability and capacity concerns that result from the interconnection of
variable-energy resources. SCE will also demonstrate the ability of lithium-ion battery storage to
provide nearly instantaneous maximum capacity for supply-side ramp rate control to minimize the
need for fossil fuel-powered back-up generation.
Goals/Objectives
Validate the performance and effectiveness of lithium-ion technology
Demonstrate the integration of intermittent wind energy
Develop a smarter, more efficient electric grid
Advance market readiness of utility-scale storage
City of Painesville
Project Description
The City of Painesville, Ohio and its partners will demonstrate vanadium redox battery-storage
capacity at the 32 MW, coal-fired Painesville Municipal Electric Plant. Using stored power enables
the facility to attain the same daily output requirement, more efficiently and with a lower carbon
footprint. When the project is fully implemented, the plant will operate at a constant 26 MW, 80%
of rated capacity. The long-term goal is to scale the battery system in stages, ultimately upgrading
the facility to 10 MW of capacity with up to 80 MWh of storage. In the first phase, 1 MW of power
with 68 hours of storage will be installed. This capacity is sufficient for Painesville to evaluate the
benefits of energy storage, assess its uses in optimizing power-generation efficiency, and facilitate
American Municipal Power with leveling the peak demands of the system. A bi-directional, four-
quadrant inverter, with a rated capacity of 1.0 MW and 1,440 amps at 480 VAC, will be used
to provide AC/DC and voltage conversions. The battery will be constructed with two parallel
electrolyte flow systems providing the total net electrical capacity of 1.0 MW. Each subsystem will
be comprised of sixty four 10 kW stacks. The subsystems will be arranged in parallel to supply
peak operating loads. Each stack subsystem will have their electrolyte flow into two 15,000 gallon
polymer tanks, at rates ranging between 500 and 1500 gallons per minute. The battery components
will be produced in the United States and the stacks will be assembled in Painesville before being
installed at the Painesville Municipal Electric Plant Battery Building.
Goals/Objectives
Demonstrate power storage to provide spinning reserves in a grid environment with expansion
to a larger scale
Establish a template that can be introduced throughout the United States
Provide data on the active use of storage to manage peak requirements in the most efficient
manner
45
Primus Power, Modesto Energy Pods
Project Description
Primus Power is deploying a 25 MW/75 MWh EnergyFarmTM in the Modesto Irrigation District
substation in California that consists of a series of EnergyPodsTM ; a plug-and -play zinc-flow bat-
tery combined with off-the-shelf components and power electronics housed inside a standard ship-
ping container. The modular design and operation will be field tested at Pacific Gas & Electric
(PG&E) with support from Sandia National Lab and the Electric Power Research Institute. The
EnergyFarmTM will displace a planned $78M fossil-fuel plant. EnergyFarmsTM can firm intermittent
wind energy and mitigate energy price spikes. During a spike the Farm switches to full dispatch
power, reducing energy purchases during costly periods, capturing multiple revenue streams, and
shortening return on investment. EnergyFarmsTM are scalable from 330 kW/1 MWh up to systems
larger than 100 MW/300 MWh.
Goals/Objectives
Develop a distributed, mobile energy storage module based on a zinc-flow battery technology
that can be mass produced
Reduce system capital costs and footprint
Enhance application flexibility
Validate module performance and functionality
Beacon Power - Hazle Township, Pennsylvania
Plant has been delayed because of Beacon Power bankruptcy and reorganization.
Detroit Edison community-scale energy storage (CES)
Project Description
Detroit Edison will design, construct, and install an aggregated 1 MW CES system in their service
territory at the Hager substation in Michigan to demonstrate the potential of CES systems to
strengthen grid reliability. The performance data of the CES devices and control systems under
in-service operating conditions will be analyzed and used to identify gaps and facilitate how the
devices can be standardized for use across the U.S. The project will also integrate the utility-owned
500 kW solar system to the energy-storage device; provide proof-of-concept testing for an integrated,
centralized communication system; and test the use of secondary-use electric vehicle (EV) batteries
as CES devices.
Goals/Objectives
Demonstrate peak shaving, demand response voltage, and emergency load relief of the CES
devices when integrated to the utility grid
Explore remote and automatic monitory and control responses
46
Develop and verify advanced modeling and simulation methods for system planning and op-
erations based on existing utility practice and expanded to include photovoltaic systems in-
tegration
Demonstrate intentional islanding of CES devices with a utility distribution circuit and how
they can aid in frequency regulation
Detroit Edison is an electric distribution utility serving approximately 2.2 million customers in
Michigan. Detroit Edison will work with selected subrecipients, consultants, contractors, and ven-
dors to demonstrate the use and benefits of CES systems in a utility territory and also to test the
ability to integrate secondary-use EV batteries into the CES demonstration effort. This project will
install 1820 S&C supplied 25 kW/50 kWh CES units and two Chrysler-supplied secondary-use
EV battery CES units into a system that includes a 500 kW storage device integrated into a solar
PV system that was installed as part of project cost-share through the Michigan Public Service
Commission (MPSC) Smart Grid Storage Program. The CES units will be coupled with the utility
scale device to demonstrate a variety of applications. The goal of this project is a proof of concept
to demonstrate the use and benefits of CES systems in a utility territory and also to test the abil-
ity to integrate secondary-use EV batteries into the CES demonstration effort. This project will
install 2022 CES units into a system that includes a 500 kW storage device integrated into a solar
system. The 1820 S&C-supplied CES units will be 25 kW/50 kWh devices and will be coupled
with the utility scale device. Two Chrysler-supplied automotive battery CES units will be installed
to demonstrate the same set of applications a year and a half after the installation of the S&C CES
devices.
PG&E CAES no grid connection (if it goes forward) before 2015
Project Description
(PG&Es advanced underground CAES demonstration project is intended to validate the design,
performance, and reliability of a CAES plant rated at 300 MW with up to 10 hours of storage.
The CAES demonstration project is scoped to test the suitability of a porous rock formation as the
storage reservoir in California, and demonstrate the technological improvements in the design of
such plants. Porous rock formations are much more plentiful than the salt domes now used by the
two operational plants in Alabama and Germany. If this geology is proven viable, this technology
has the potential to be replicated throughout California and elsewhere in the United States. The
project is also differentiated by its potential use of a new CAES plant design that is much more
efficient than first-generation Alabama and German designs. This project is comprised of three
phases. Phase I includes site selection, reservoir testing, preliminary plant design, an environmental
assessment, and a competitive solicitation to determine if there are interested and viable parties for
plant construction, ownership, and operations/maintenance. Phase I is estimated to last 4.5 years.
Phase II, which includes obtaining approval to proceed with the construction and commissioning of
a full CAES plant, has an estimated 6-year duration. Phase III includes operations and monitoring
and is expected to occur over 2 years.
Goals/Objectives
Verify the technical performance of advanced CAES technology using a porous rock formation
as the underground storage reservoir
Integrate intermittent renewable resources
47
Maintain emergency spinning/non-spinning reserve and perform volt-ampere reactive/voltage
support
A.2 Research Level (Not Grid-Connected) Demonstration Projects
Aquion Energy
Project Description
Aquion Energy and its partners will demonstrate a low cost, grid-scale, ambient-temperature
sodium-ion energy-storage device. The energy-storage chemistry in this device uses an electro-
chemical couple that combines a high-capacity carbon anode with a sodium intercalation cathode
capable of thousands of deep discharge cycles over extended periods of time. The proposed aqueous
sodium-ion technology includes the use of thicker electrodes, less expensive separator and current
collector materials, and the use of benign materials for electrodes and electrolyte salts. This project
will progress the work from bench-scale to pilot-scale, enabling multiple ampere-hour cells to be
manufactured and assembled into test batteries. Aquion plans to site units with a capacity be-
tween 10 kWh and 100 kWh that have the ability to perform medium-to-long-duration (more than
2 hours) charge and discharge functions with greater than 95% DC-DC efficiency. The units will
be safe and environmentally benign. Testing will characterize the energy storage capacity of the
units, the response to various signals, compliance with utility interconnection standards, battery
and power-conversion-system efficiency, and effectiveness under various cycles typical of the appli-
cations being validated. Utility application-level testing will test the functionality of the unit with
respect to its ability to respond to external control signals and properly interact with electric grid
in carrying out relevant sequences. The pilot line will be commissioned for production at the end
of the project.
Goals/Objectives
Projected capital cost less than $250/kWh at pack level
Deep discharge cycle life of >10,000 cycles
Volumetric energy density of >20 kWh/m3
Calendar life of over 10 years
Build and install multiple 100 kWh batteries
Ktech Corp
Project Description
Raytheon Ktech and EnerVault will integrate EnerVaults Vault-20 battery energy-storage system
(250 kW/1 MWh) with a Helios dual-axis tracker 180 kW PV system. The system will be deployed
at an agricultural site in Californias Central Valley. It will store the energy generated and dispatch
power to run an irrigation pump and inject energy back into the utility grid during peak times to help
offset demand from a section representing 4% of Californias electricity demand. System modularity
provides scalability for multi-megawatt deployments. The Vault-20 consists of electrolyte tanks
48
and transportainers, which house stacks, pumps, control system, and power-conditioning systems.
Technology development will progress from 15 15 cm lab-scale cells and 20-layer stacks, to a
25 kW prototype system, then a 30 kW alpha system, concluding with a 250 kW beta system.
EnerVault plans to begin manufacturing flow battery stacks in its Northern California plant within
12 months of project completion.
Goals/Objectives
Develop a modular system rated at 250 kW and 1 MWh that fits inside a standard shipping
containers to minimize onsite deployment time and cost
Integrate a battery energy-storage system with a variable renewable energy resource
Reduce cost and environmental impacts
Amber Kinetics not grid connected
Project Description
Amber Kinetics is developing a flywheel system from sub-scale research prototype to full-scale
mechanical flywheel battery and will conduct a commercial-scale demonstration. The goal is to
deliver a cost-effective prototype flywheel system that can be grid connected and electrically charged
and discharged. The flywheel stores energy in a spinning rotor that is connected to an electric
motor that converts electrical energy into mechanical energy. To recover the energy, the motor is
electrically reversed and used as a generator to slow down the flywheel converting the mechanical
energy back into electrical energy. Amber Kinetics will improve the traditional flywheel system by
engineering breakthroughs in three areas, resulting in higher efficiency and radically reduced cost:
bearings, low-cost rotor, and high-efficiency motor generator. This technology can also be used to
optimize existing infrastructure.
Goals/Objectives
Deliver a prototype system that can be grid connected and electrically charged and discharged
Develop a commercial-scale prototype of the flywheel technology
Provide a plan to scale the system to cost-effective price points
Achieve energy storage efficiencies greater than 85%
SEEO not grid connected
Project Description
SEEO and its partners are demonstrating a large-scale prototype of a solid-state electrolyte lithium-
ion rechargeable battery for use in smart-grid energy storage applications. SEEO seeks to vali-
date this technology to address the needs of CES systemssmall (less than 100 kW) distributed
energy-storage systems alongside pad-mounted and pole-mounted transformers. The battery pack
is more than a 50% improvement in weight and energy density; has 1015+ year operating life with
3,0005,000 or more cycles; has no volatile or flammable components; and will be 35% cheaper
than existing lithium-ion batteries. This approach allows independent control over mechanical and
electrical properties. The cell can withstand temperatures as high as 150 C and voltages of 10 V
49
without incident. An independent analysis of the environmental and economic impact of battery
improvement will also be conducted.
Goals/Objectives
Develop and deploy a prototype battery system that validates SEEOs technology
Improve battery installation and maintenance
Produce a plan for manufacturing and commercializing the technology at utility scale
Sustain X not grid connected
Project Description
SustainX is developing and demonstrating a modular, market-ready energy-storage system that
uses compressed air as the storage medium. SustainX uses a crankshaft-based drivetrain to convert
electrical energy into potential energy stored as compressed air. SustainXs ICAES system captures
the heat from compression in water and stores the captured heat until it is needed again for ex-
pansion. Storing the captured heat eliminates the need for a gas combustion turbine and improves
efficiency. SustainX achieves isothermal cycling by combining patented innovations with a design
based on mature industrial components and principles. The system is designed for a 20-year life-
time. It achieves full-power output from start-up in less than one minute, and it does not use toxic
chemicals.
Goals/Objectives
Demonstrate the viability of isothermal compressed air technology to provide cost- effective
energy storage
Validate scalability for applications in both low- and medium-voltage distribution or sub-
transmission grids
50
B Solar PV Smoothing Metric Derivation
Justification for developing these metrics stems from statistics theory. The standard deviation from
the mean power output represents the average power variation that the power system is likely to
experience during the days particular weather conditions. The percent reduction in the standard
deviation of unsmoothed vs. smoothed power shows smoothing ESSs effectiveness by describing
the improvement of how far the power is likely to deviate from its mean.
By applying the same statistical treatment to calculated ramp-rates (i.e., discrete-time derivative
with varying t of the same data), a new (but similar) metric can be described. The percent
reduction in ramp-rates, however, represents the ESSs ability to dampen the rate at which power
fluctuations occur. While the two metrics are very similar, they are not interchangeable.
The third and final metric represents an intuitive measurement of ESSs usefulness by quantifying
how well the days largest power swing is diminished. This particular quantitys mitigation is
important to a system because these large swings are of particular concern to the utility connected
to the storage system. Combined with the previous two metrics, one can perform a complete
quantitative assessment of a PV smoothing energy-storage system.
Mathematically Quantifying Smoothing
In order to calculate the first- and second-order statistics of a given PV power profile, the mean
must be calculated and removed. Due to the nature of the diurnal trend associated with PV power,
it can be shown that the power output is not wide-sense Stationary (WSS).
Let X(n) = Ppv and Fs = 1 [sample/sec] and have a typical shape similar to the waveform in Figure
12.
For X(n) to be WSS:
E[X(n)] = x ; < n < + (11a)

Cx [n1 , n2 ] = g(|k|); < k < + (11b)
where E[ ] is an expectation operator, C[ ] is a covariance operator, and k = n1 n2 (with <

n1 < + and < n2 < +).
While applying an autocorrelation function or formal unit root test to the data would reveal that
it is nonstationary in time (due to the autocorrelation functions slow decay), it is more easily
and intuitively shown by example that X(n) random process does not meet the requirements to be
considered WSS. By visual inspection, the sample mean estimator of E[X(n)] for 0 < n < 5000 [sec]
is not equivalent to E[X(n)] for 30000 < n < 35000 [sec]. This indicates that the temporal average
varies with respect to time. Thus, the mean cannot be removed by simply subtracting the sample
mean point-by-point. Because of this, the signal is disqualified from the consideration of WSS and
all of the assumptions that go along with the designation. Therefore, a time-varying mean removal
method must be developed in order to approximate the signal as being WSS.
Note that the use of a full days worth of data is analyzed and assumed non-stationary. Because
the trend that creates the non-stationarity is shaped (mostly) by the diurnal trend of the sun, the
assumption that the sample mean of a sub-set of points limited in length is constant with respect
to time can be made. This assumption is justifiable because the maximum rate at which the sun
causes appreciable power output change is roughly 2 orders of magnitude slower than the rate at
51
500
PV Power
Smoothed Power
Battery Power
400
300
Power (kW)
200
100
100
9 10 11 12 13 14 15 16 17 18
Time (hr)
Figure 12: Smoothed power profile, 7-Mar-2012 data.
which the cloud cover can cause appreciable change in power output. The fluctuations of interest
are on the order of magnitude of [W/second], while the sun causes a much slower maximum rate of
change, somewhere on the order of [W/10 min] in mid-morning and late-afternoon, and even slower
during the mid-day time period. If a time window is chosen appropriately such that the power
differences associated with the suns movement are eliminated but the rapid fluctuations are left
intact, this is a strong assumption. Further research may be performed in the future to determine
optimal window lengths, depending upon time of year and other variables.
Now that the window length has been appropriately chosen, one can perform a moving average.
N
2
1 X
Y (n) = X(n) (12)
N +1 N
n= 2
where N = window length in samples, X(n) = PV power signal, and Y(n) = moving averaged
PV power signal. Note that the above moving-average calculation is a modified version of Proakis
and Manolakis (2006) in that it is centered about m. This is typically referred to as a central
moving average. Therefore, the process is a noncausal filter and can only be realized as a post-
processing utility. (i.e., real-time application is impossible because the current output relies on
future inputs).
The resulting moving average calculation vs. its original signal is shown in Figure 13.
52
Moving Averaged vs. Raw PV Power
500
PV Power
Power (kW)
400 PV Moving Avg
300
200
100
0
8 10 12 14 16 18 20
Time (hr)
Moving Averaged vs. Smoothed PV Power

500
Smooth Power
400 Smooth Moving Avg
Power (kW)
300
200
100
0
8 10 12 14 16 18 20
Time (hr)
Figure 13: Moving average versus original signal
By subtracting Y(n) from X(n), as in Equation 13:
R(n) = X(n)Y (n) (13)
where R(n) = residue of X(n), X(n) = PV power signal, and Y(n) = moving averaged PV power
signal. The resulting signal contains the PV systems high-frequency fluctuations R(n) (residues),
without the low-frequency movement caused by the diurnal sun trend. R(n) comprises the fluctu-
ations of interest because their mitigation is the goal of the ESS.
A first- and second-order statistic-based comparison of the unsmoothed vs. smoothed residues
results in the mathematical representation of how well the ESS performs power smoothing; hence
the first metric.
N 1
1 X
E[R(n)] = R(n) (14)
N n=0
The temporal or sample mean, described in Equation 14, is used to verify the central moving
averages effectiveness in removing the mean from the data. The temporal average is shown to be
a minimum variance unbiased estimator in the literature. Intuitively, the expected value of the
53
Ramp Rate PV
200 Ramp Rate Smoothed PV
150
100
Ramp Rate (kW/min)
50
50
100
150
200
250
8 10 12 14 16 18 20
Time (hr)
Figure 14: Residue R(n)
residue function R[n] should be approximately equal to 0. If it isnt, a modification to the window
length of the central moving average function is necessary to ensure proper mean removal from the
data.
Note that the residue function R(n) will have a varying statistical distribution based upon the
profile of solar irradiance that the PV array is subject to on any given day. However, because
the data has been made approximately WSS, it is of interest to attempt to describe the resulting
random process as a known random variable distribution. Of course, the first logical assumption
is that the resulting distribution is of a Gaussian type with its standard deviation describing the
days cloud cover and a zero mean. This can be visualized by thinking of the width of the Gaussian
curve to be much wider on days that experience large, rapid fluctuations, and very narrow on a
clear, near-cloudless day.
A statistical distribution comparison can be performed between the calculated cumulative density
function (CDF) of R(n) and the CDF of a Monte-Carlo simulation of a known distribution with the
same mean and standard deviation as R(n) to approximate R(n)s distribution type. A standard
method of comparing the two distributions is described in the Kolmogorov-Smirnov (KS) test, given
in Equation 15.

Dn = N sup[|F (n) G(n)|] (15)
nS
where N = length of R(n), sup = the supremum operator, G(n) = CDF of the Monte-Carlo
54
simulation of the estimated distribution, and F[n] = CDF of R(n). The KS statistic Dn itself
is quite arbitrary, but when comparing the distribution of R(n) to other known distributions, it
allows for a quantitative assessment to be made. The KS test is a classical tool that allows for
the comparison of distributions to be made quite easily. However, it should be noted that if a
more thorough analysis of the distribution of R(n) were desired, various other tests are available
that may be more appropriate. The Anderson-Darling or Cramer-von Mises tests can provide more
information on the distribution comparisons, but are unnecessary in this case.
To calculate the estimated CDF of the residue function, it is first necessary to approximate its
probability mass function (PMF). The PMF is approximated using a histogram approach. This
gives a nice data-distribution visual representation, with particular interest shown to the tail ends
of the distribution where high-power swings occur. PMF visual inspection should lead to several
observations about the data.
4 Residue Histogram
x 10
PV Residue
Smooth Residue
2
1.5
Number of Instances
0.5
0
100 50 0 50 100
Distance from Mean [kW]
Figure 15: Probability mass function.
First, note the mean. This value is calculated in Equation 14, but a quick inspection of the PMF
should provide visual verification that most of the data falls in the zero (or near-zero) bin. The
PMF should have no skewness, as the removal of the moving average should result in a data set
that is no more likely to have a value higher than the temporal mean than it is to have a value
lower.
Also, note that the tail ends of the distributions compare (i.e., are the tails pushed toward the
zero-mean in the smoothed PMF compared to the unsmoothed PMF). Because the energy-storage
systems desired function is smoothing, the smoothed PMF should not have as high a probability
of containing large power deviations as the untreated R(n) PMF does.
55
The PDF of R(n) is then summed cumulatively to produce the CDF. The KS statistic in equation
15 is applied to the CDF against a set of equal-length data simulated by a Monte-Carlo simulation.
As stated previously, the logical first assumption is that the data R(n) is distributed in a Gaussian
shape. While this is a good starting point as it is symmetrically distributed, easily visualized, and
a likely distribution for such a signal, other distributions appear to fit this data slightly better,
according to the KS statistic.
The Laplacian distribution, described in Equation 16 and found in Kay (2013), is a noticeably
better fit for R(n) than a Gaussian distribution (Kay (1998) p.382).
r !
1 2
p R(n) = exp |R(n)| ; < n < + (16)
2 2 2
where 2 = variance of the random process. It appears quite similar to a Gaussian distribution,
but with a steeper probability concentration at the mean. It is symmetric about the mean, and can
be described by its standard deviation much like a Gaussian distribution. More information on the
Laplacian distribution and why it fits R(n) is provided in the Results section.
v
u N 1
u 1 X
= t (R(n) E[R(n)])2 (17)
N 1 n=0
where n = 0, 1, ..., N-1; E[R(n)] is the expected value of R[n]; and N = length of R(n). A proven
minimum variance unbiased estimator for the WSS signals variance is the sample variance estimator
given in equation 17. It is important to choose the standard deviation R(n) estimator appropriately
as the smoothing metric of choice will directly depend upon this calculation.
The standard-deviation calculation in equation 17 is the most important result of the statistical
analysis. This is because the standard deviation represents how much variation or dispersion from
the expected value exists in a particular data set. In power-system terms, it represents the amount
of power that can be expected to deviate from the projected power output at any given time. Due
to this, the standard deviation of both the unsmoothed and smoothed powers is of great importance
in quantifying the impact that the ESS device has, as shown in equation 18.

smooth
= 100 1 (18)
unsmooth
where smooth = standard deviation of smoothed PV residue Rsmooth (n), and unsmooth = standard
deviation of unsmoothed PV residue Runsmooth (n). Ultimately, the metric for quantifying the impact
that the ESS device has on smoothing the PV power output is the so-called percent reduction
in standard deviation of power (RSDP). This measurement shows the relationship between the
smoothness of the raw PV power and its battery smoothed counterpart. It captures the batterys
smoothing effects, and makes efficient use of all the data points. These are highly desirable qualities
of the estimator function, and they justify its use as the smoothness metric.
As mentioned in the Methods section, using ramp-rates to further quantify the ESSs ability
to smooth PV power as an additional metric is important for several reasons. First, it can be
56
used as a general check for consistency with the previous smoothing metric. Because a large PV
arrays power output can change very rapidly (on the order of killowatts per second), a smoothing
systems performance will need to dampen the rate of the power fluctuations appreciably in order
to be considered an effective smoothing device. This is acceptable because if the storage system is
capable of reducing the magnitude of power fluctuations by a certain amount, it should also cause
a similar effect on the rate at which those fluctuations occur. Comparing the two metrics for a
system should produce similar results for a given smoothing system.
Second, as alluded to above, rapid ramp-rates can be a nuisance to a power system. They can
cause undesirable voltage magnitude changes at the distribution level, flicker, and other unwanted
noise on the line. The rapid fluctuations also make PV generation more difficult to predict as a
generating asset, and may cause other equipment on the system to try to compensate for rapid
generation fluctuations (load tap changers, VAR compensation devices, etc.). In this context, the
phrase ramp-rate will be defined as follows:
X(n + k)X(n)
Y (n) = ; n = 0, 1, ..., (N 1) k (19)
K
where k = sample lag related to time interval of interest, and N = length of X[n]. As shown in
equation 19, the definition of ramp-rate will follow that of a discrete-time derivative. However, there
is a slight modification, in that k can take on a number of values. Note that k is chosen based upon
the desired time interval of interest. Choosing k for a particular sampling frequency and desired
time interval creates a definition of ramp-rate that varies.
For instance: If fs = 1 Hz (sample rate), and desired time interval is 1 second = t0 , then Y(n) is in
terms of W/s. If t0 =1 hr, then Y(n) is in terms of W/hr.
While this may seem quite obvious, it is important to note the units of Y(n). The resulting ramp-
rate signal is highly dependent upon the time interval chosen, and its units describe the utility of
the particular time interval chosen. For example, we are not interested in the second case above
where Y[n] is described in terms of [W/hour]. That calculation is far too slow for the purposes of
this metric. We are interested in ramp-rates on the order of [W/min], with the motive discussed
previously.
A very useful property of the derivative (in this case, discrete-time derivative) is that it has a
de-trending effect on data that is non-stationary in time. In the literature, a stationary series is
considered integrated of order one [or I(1)] if the series becomes stationary after a first differencing.
The data sets of interest in this discussion are known to be non-stationary in time. However, after
applying the ramp-rate calculation above, they have been effectively zero-meaned. This makes the
calculation incredibly powerful because the resulting Y(n) from 19 represents the ramp-rates of the
power with the bonus of being approximately WSS.
Y(n) from equation 19 can now be treated similarly to R(n) from the previous set of calculations.
That is, equations (3), (4), (5), and (6), as well as the distribution characterization, can be applied
to Y(n) with the final metric being the so-called percent reduction in RSDR. It should be apparent
that the metric put forth as the RSDR is closely related to the previously described RSDP. A close
relationship between these two is to be expected when applied to a system, but it is important to
note that they describe different parameters.
A third metric termed in Methods as max-min reduction is highly intuitive, but less insightful than
57
the two previously described methods. As was formerly mentioned, power fluctuations on large PV
arrays may be on the order of kilowatts/second, and these large power swings can be of concern to
a utility with a large PV system tied to it.
Pmax swing = max|R(n)| (20a)

!
Rsmooth (m)
swing = 100 1 (20b)
Pmax swing
where m = linear index of Pmax swing . By looking at the maximum of the above described residue
function R(n) of the unsmoothed power data set, we can determine the magnitude of the biggest
power swing of that particular day. The time index of that power swing is noted, and the value
of smoothed power at that particular time is gathered. The significance of this calculation is to
determine how well the maximum power swing of the day is mitigated. A high-performing smoothing
ESS will reduce these maximum power swings well.
58
C AGC Signal Assumptions for Smoothing Battery
An AGC signal from PJM (originally Pennsylvania-Jersey-Maryland, PJM is a Regional Trans-
mission Organization covering all or portions of 13 states and the District of Columbia) was used
to examine how the PNM Prosperity smoothing battery system might operate taking such a sig-
nal.
The PJM data used represents an AGC signal in four-second intervals over a 24-hour period. This
data is normalized to be between 1 and +1, and so it was multiplied by 250 kW to scale the
signal up to a magnitude that is reasonable for the 500 kW shifting battery. The resulting power
demanded signal is shown in Figure 16.
250
200
150
Power Demanded (kW)
100
50
50
100
150
200
250
2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
Figure 16: Power demanded (kW) over time by the sample AGC signal from PJM.
The signal in Figure 16 would require the smoothing-battery system to discharge 900 kWh over the
24-hour period. It would also require charging of the same amount. Note that the signal does not
take into account losses due to inefficiencies, which would require the battery to charge more than
it discharges. In order to prevent the SoC from drifting ever lower, the signal would need to be
modified to take round-trip losses into account. Such a modification was not made here.
The SoC resulting from this signal would keep the battery in a tight band, as shown by Figure
17 which illustrates the cumulative energy flows this signal would produce. Even though the com-
manded power reaches plus or minus 250 kW at times, the cumulative power flows never exceed
plus or minus 15 kWhmeaning that the state of charge never changes by more than about 2%.29
Therefore, following this AGC signal would not require a storage system with the useable storage
29
For a battery system of 1 MWh total capacity (the capacity of the PNM Prosperity smoothing battery), the
change in SoC can be found by 15 kWh/1000 kWh = 1.5%, which here is rounded to 2%.
59
capacity that the PNM Prosperity smoothing battery has (approximately 350 kWh)a battery with
a much smaller capacity would suffice.
20
15
Cumulative Energy Flows (kWh)
10
10
15
20
2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
Figure 17: Cumulative Energy Flows (kWh) resulting from the sample AGC signal from PJM
For this sample PJM signal, the maximum up ramp rate of is 13.8 kW/second, and maximum down
ramp rate is 30.6 kW/second.
While we realize that PJMs AGC signal differs from day to day, and that PNMs AGC signal
is likely to differ from PJMs signal, this analysis is intended to show that following an actual
AGC signal is a task that the PNM Prosperity smoothing battery could reasonably be expected to
handle.
60
DISTRIBUTION:
1 Dr. Imre Gyuk

U.S. Department of Energy
1000 Independence Ave SW
Washington, D.C. 20585-0001
1 MS 1033 Abraham Ellis, 6112

1 MS 1108 Sean Hearne, 6111
1 MS 1108 Georgianne Huff, 6111
1 MS 1108 Daniel Borneo, 6111
1 MS 1108 David Rose, 6111
1 MS 1140 Dhruv Bhatnagar, 6113
1 MS 1140 Ryan Elliott, 6113
1 MS 1140 David Schoenwald, 6113
1 MS 1140 Dakota Roberson, 6113
1 MS 1140 James Ellison, 6113
1 MS 1140 Ross Guttromson, 6113
1 MS 1140 Cesar Augusto Silva Monroy, 6113
1 MS 0576 Raymond Byrne, 5521
1 MS 0899 Technical Library, 9536 (electronic copy)
61

Sand2014 2883

Uploaded by

Copyright:

Available Formats

Sand2014 2883

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sand2014 2883

Uploaded by

Copyright:

Available Formats

What is the purpose of this study?

What is the purpose of this study?

What does the PNM Prosperity electricity storage project consist of?

What does the PNM Prosperity electricity storage project consist of?

SANDIA REPORT

Performance Assessment of the PNM

Dakota Roberson, James F. Ellison, Dhruv Bhatnagar, and David A. Schoenwald

Approved for public release; further dissemination unlimited.

Available to DOE and DOE contractors from

Telephone: (865) 576-8401

Available to the public from

Telephone: (800) 553-6847

PNM Prosperity Electricity Storage

Dakota Roberson, James F. Ellison, Dhruv Bhatnagar, and

Acronyms and Abbreviations 9

4 Recommendations for Battery and Photovoltaic Systems 37

A ARRA Energy Storage Projects 43

B Solar PV Smoothing Metric Derivation 51

C AGC Signal Assumptions for Smoothing Battery 59

Table 1: Shifting battery efficiency.

Measurement Includes BoP Losses? Round-Trip Efficiency Annual Efficiency

1.1 ARRA Energy-Storage Demonstration Projects

Project Name Location Project Description Funding ($M) Project

Table 3: ARRA electricity storage research and development projects.

Project Name Location Project Description Funding ($M) Project

Figure 1: PNM Prosperity energy-storage project.

A 0.5 MW/0.35 MWh power-smoothing battery system

Figure 2: Battery storage-system one-line diagram.

Figure 3: System one-line diagram.

1.3 Battery Description

Figure 4: UltraBattery configuration (Lam et al., 2006).

2.1 PV Smoothing Effectiveness

Pmax swing = max|R(n)| (3)

A detailed discussion of these metrics can be found in Appendix B.

2.2 Round-Trip Efficiency

Figure 5: PV shifting, 26-Jun-2013 data.

2.3 Balance-of-Plant Load

The Balance-of-Plant (BoP) load is an important consideration in calculations of storage system

2.5 Estimated Remaining Battery Lifetime

3.1 PV Smoothing Effectiveness

measured solar insolation

length of measured irradiance plot

Figure 6: High PV variability, 15-Apr-2012 data.

Figure 7: Moderate PV variability, 22-Jun-2013 data.

Figure 8: Mild PV variability, 13-Apr-2012 data.

3.2 Shifting Battery Efficiency

Table 5: Shifting Battery Efficiency

Measurement Includes BoP Losses? Round-Trip Efficiency Annual Efficiency

Figure 9: Shifting-batttery action, 23-June-2013 data.

Table 6: Round-Trip Efficiency over Sampled Days

Year: 2012 2013

Figure 10: Shifting-battery efficiency, by month.

Year: 2012 2013

The average efficiency including BoP load for a round-trip is 83%.

Table 8: AC-to-AC Round-Trip Efficiency of Shifting-Battery System

Year: 2012 2013

3.3 Balance-of-Plant Load Characterization

Figure 11: Availability (source of data for calculations: Ecoult)