Professional Documents
Culture Documents
Udegbe MS Thesis
Udegbe MS Thesis
Udegbe MS Thesis
A Thesis in
by
Egbadon Udegbe
Master of Science
August, 2014
The thesis of Egbadon Udegbe was reviewed and approved* by the following:
Seth Blumsack
Associate Professor of Energy Policy and Economics
Thesis Co-Adviser
David Riley
Associate Professor of Architectural Engineering
Thesis Co-Adviser
Mort Webster
Associate Professor of Energy and Mineral Engineering
Luis F. Ayala H.
Associate Professor of Petroleum and Natural Gas Engineering
Associate Department Head for Graduate Education
ii
ABSTRACT
traditional backup infrastructure aimed at promoting power quality and surety for business
operations fail to maximize cost and operational efficiency, while offering limited ancillary
benefits related to revenue generation, environmental footprint, and grid stability. These
challenges have the potential to be addressed through the integration of grid-interactive storage
and solar photovoltaic (PV) on-site generation with the building electrical systems.
This research focuses on modeling the energy management system (EMS) for a commercial
building microgrid capable of performing peak-shaving and providing backup reserve power,
while participating in the PJM frequency regulation market. This has been achieved by defining
the operational framework and control strategy for the EMS, by utilizing system data and
information from the ongoing Building 7R microgrid development at the GridSTAR Center in
the Philadelphia Navy Yard. Based on these, a MATLAB model has been developed to simulate
annual energy transfers between the building, the local utility and PJM.
Using the EMS model simulation results, a methodology has been developed to assess the
economical tradeoffs between various operational modes, and also technological choice between
lithium-ion and lead-acid batteries. These have been analyzed by considering market revenues
and backup cost mitigation provided by the building-integrated microgrid. Within the proposed
framework, it has been determined that lead-acid systems provide more peak-shaving revenues
and building demand curtailment, while a comparable lithium-ion system yields more annual
revenues. Overall, allocation of battery capacity away from frequency regulation operation has
iii
Table of Contents
List of Figures…………………………………………………………………………………………………………………………………………vi
List of Tables………………………………………………………………………………………………………………………………………….vii
Acknowledgements……………………………………………………………………………………………………………………………….viii
iv
5.2 Sensitivity of yearly revenues to capacity allocation between peak-shaving and regulation .......... 73
5.3 Sensitivity of yearly revenues to overall capacity allocation ............................................................ 76
5.4 Sensitivity of yearly revenues to overall capacity allocation and blackout costs ............................. 80
Chapter 6: Conclusions ......................................................................................................................... 87
6.1 Summary of work .............................................................................................................................. 87
6.2 Key findings ....................................................................................................................................... 87
6.3 Recommendations for future study .................................................................................................. 89
Appendix A: 7R eQuest Model Parameters and Assumptions ................................................................ 90
Appendix B: Lithium-ion battery specifications ...................................................................................... 98
Appendix C: MATLAB code for lead-acid battery sizing optimization ................................................... 100
Appendix D: TRYNSYS PV generation model ......................................................................................... 101
Appendix E: MATLAB code for peak-shaving optimization and EMS model......................................... 105
Bibliography…………………………………………………………………………………………………………………………………………120
v
List of Figures
Figure 1-1: U.S. electricity retail sales by end-use sector [2] ........................................................................ 1
Figure 1-2: CO2 emissions from energy consumption in the commercial [5] ............................................... 3
Figure 1-3: Estimated interruption costs by customer class (2002 CPI-weighted dollars) [8]...................... 5
Figure 1-4: Classification of prevalent emergency standby systems ............................................................ 7
Figure 1-5: NOx and PM10 emission factors from diesel generators relative to other sources [12] ............. 9
Figure 2-1: Annual trends in U.S. PV installation capacity and average system price [15] ........................ 13
Figure 2-2: Key features of microgrid systems [22] .................................................................................... 17
Figure 2-3: Example of microgrid topologies according to scale [22]......................................................... 19
Figure 2-4: Storage system operational uses [30] ...................................................................................... 21
Figure 2-5: Technoeconomic comparison between storage technologies [33] ......................................... 23
Figure 2-6: Hierarchical control framework of microgrid EMS [34]............................................................ 24
Figure 3-1: 7R microgrid system electrical characteristics ......................................................................... 28
Figure 3-2: Functionality of Building 7R microgrid EMS (based on [47]). ................................................... 31
Figure 3-3: Battery capacity allocation approach ....................................................................................... 33
Figure 3-4: Summary of EMS operational framework ................................................................................ 34
Figure 3-5: Overall EMS control strategy. (1) Peak-shaving optimization; (2) PV recycling; (3) Operational
mode decision ............................................................................................................................................. 35
Figure 3-6: EMS model overview ................................................................................................................ 38
Figure 4-1: Typical building energy demand profiles for different day types duing representative days in
Winter, Spring, Summer, and Spring .......................................................................................................... 49
Figure 4-2: Average, minimum and maximum possible building demand (based on Table 4-2(a)) ........... 51
Figure 4-3: Average energy price and average price differential ............................................................... 52
Figure 4-4: Dispatch-to-contract ratio characterization ............................................................................. 55
Figure 4-5: Illustration of lead-acid array sizing problem ........................................................................... 58
Figure 4-6: TRNSYS PV simulation model.................................................................................................... 60
Figure 5-1: Example simulated battery energy flows for one-day period, using lithium-ion ..................... 65
Figure 5-2: Simulation results for days representing Spring, Summer and Fall seasons (li-ion) ................ 68
Figure 5-3: EMS simulation results for select days (lead-acid) ................................................................... 72
Figure 5-4: Sensitivity of revenues to battery capacity allocation by battery technology ......................... 74
Figure 5-5: Comparison between lead-acid and lithium-ion revenues ...................................................... 76
Figure 5-6: Market revenues from allocating lithium-ion system in various ratios.................................... 78
Figure 5-7:Market revenues from allocating lead-acid system capacity in various ratios ......................... 79
Figure 5-8: Total system revenues on a low-demand day (lithium-ion system)......................................... 81
Figure 5-9: Total system revenues on an average- to high-demand day (lithium-ion system) .................. 82
Figure 5-10: Total system revenues on a low-demand day (lead-acid system).......................................... 83
Figure 5-11: Total system revenues on an average-demand day (lead-acid system) ................................ 84
Figure 5-12: Total system revenues on a high-demand day (lead-acid system) ........................................ 85
vi
List of Tables
Table 1-1: Types of commercial buildings and their electricity consumption profiles [4]............................ 2
Table 1-2: Estimated cost-per-event of power interruptions by duration and consideration [9] ................ 5
Table 1-3: Auxiliary power systems for representative high-sensitivity commercial classes [10] ............... 8
Table 2-1: Benefits of PV systems in buildings [18] .................................................................................... 14
Table 2-2: Major microgrid system components [25] ................................................................................ 18
Table 2-3: Comparison between major cell battery chemistries [31] ........................................................ 22
Table 4-1: Modeled occupancy and usage schedules within Building 7R .................................................. 47
Table 4-2: Summary of Building 7R energy demand................................................................................... 50
Table 4-3: Price data summary ................................................................................................................... 52
Table 4-4: Qualitative comparison between lead-acid and lithium-ion systems [31] [32] [56] ................. 56
Table 4-5: EMS model battery specifications ............................................................................................. 59
Table 4-6: Key technical characteristics of PV modules in 7R microgrid (STC) [59].................................... 60
Table 4-7: Avoided outage cost for varying backup reserve allocation on different day types ................. 62
Table 5-1: Example energy flows for one-day simulation, using li-ion battery (Orange=battery charge;
Purple=battery discharge) .......................................................................................................................... 64
Table 5-2: Example simulation results using lead-acid battery (Orange=battery charge; Purple=battery
discharge) .................................................................................................................................................... 70
Table 5-3: Total revenues for various ratios of peak-shaving (PS) and frequency regulation (FR) ............ 73
Table 5-4: Percent change in system revenues resulting from utilization of lead-acid over lithium-ion
system ......................................................................................................................................................... 75
vii
Acknowledgements
Firstly, I would like to thank Dr. Riley for his support throughout this project, and for
surrounding me with the resources to immerse myself in this research area. Thank you also to Dr.
Blumsack for providing invaluable insight, and for helping me sculpt my ideas into useful work.
Secondly, I extend my gratitude to the GridSTAR industry partners and facilitators, who
shared their expertise and provided information that went into this work. Thank you to Dan
Finally, I would like to extend my heartfelt gratitude to my family for their unwavering
support throughout my educational career. Specifically, I thank my parents, Joe and Bola
Udegbe, for providing a support structure and motivating me to push the limits. I am eternally
grateful.
I would also like to acknowledge Luke Witmer for helping me with the PV simulation
Thank you to many others who directly or indirectly contributed to this study.
viii
Chapter 1: Introduction
Commercial buildings play a significant role in the U.S. energy market. The commercial sector
obtains 79% of its overall energy consumption from electricity [1]. This represents a drastic
increase over time, with the commercial sector deriving only 54% of its energy from electricity
40 years ago [1]. Specifically, the 2013 commercial sector electricity consumption amounted to
approximately 1.34 million kilowatt-hours (kWh), which made up 36.3% of national electricity
retail sales and $137.8 trillion in revenues [2]. Figure 1-1 below displays the share of the
There currently exists a wide variety of commercial buildings, based on application and principal
business activities. According to the U.S. Energy Information Agency (EIA), the classification
encompasses “all buildings in which at least half of the floorspace is not used for residential,
1
industrial, or agricultural purposes…,” including “building types that might not traditionally be
considered ‘commercial,’ such as schools, correctional institutions, and buildings used for
religious worship” [3]. Based on the most recent EIA Commercial Buildings Energy
Consumption Survey (CBECS), different categories of commercial buildings and their respective
Table 1-1: Types of commercial buildings and their electricity consumption profiles [4]
As is apparent in the above information, office buildings represent the majority of commercial
electricity consumption, with computers, dedicated servers, printers and photocopiers accounting
for most of this usage. In fact, the onset of these technologies from 1979 is concurrent with the
2
The commercial sector’s increasing dependence on electricity as an energy source means that it
has a strong bearing on the U.S. aggregate emissions output, especially since 40% of electricity
is generated from coal and 29% comes from natural gas [1]. Figure 1-2 below displays the
Figure 1-2: CO2 emissions from energy consumption in the commercial [5]
The EIA predicts a 19% increase in commercial electricity consumption, compared to 13% and
17% in the residential and industrial sectors. Based on the 2009 U.S. Environmental Protection
Agency (EPA) estimate of 1,216 lb of CO2 per megawatt-hour (MWh) of electricity generated,
and a corresponding social cost of $35 per ton on CO2, an annual social cost of $28 billion is
imposed by the U.S. commercial sector [1]. These statistics signal the necessity for alternate
means to meet the energy requirements of commercial buildings, especially since the EPA has
established a goal of 25% reduction in greenhouse gases by 2020 from Scope 1 and Scope 2
3
The Pennsylvania State University Sustainability Institute has thus proposed the development of
the Building 7R commercial building microgrid at the GridSTAR Center in the Philadelphia
Navy Yard, with the aim of fostering research and improved understanding of energy
consumption and cost-saving mechanisms in the commercial building sector. The 7R system is a
25,600 square-foot classroom building structure, which is intended to feature photovoltaic (PV)
solar and battery storage integration, coupled with tactical utilization of multiport inverter
topologies and control mechanisms to optimize system efficiency and provide backup power to
critical loads. System information and usage data from this building will form the framework for
The frequency, voltage stability, and reliability of electrical power demanded by the commercial
sector may not always be guaranteed in practice, due to power supply disturbances beyond the
control of the utility supplier. Weather factors such as lightening, wind, rain, ice storms, storms,
hurricanes, and floods, are common causes of these power failures. In addition, factors related to
supply power system operation and reliability may result in supply disturbances. For example, in
lines experiencing excessive current, supply interruptions are created when overcurrent and
short-circuit current protective devices (such as fuses and circuit-breakers) are removed to clear
faults [7].
Power interruptions and power quality events represent significant costs to the commercial
sector, and various studies have been conducted to quantify these costs in monetary and energy
terms. On the basis of duration, frequency and customer class, [8] consolidates data and
quantifies the economic cost of reliability events in the U.S in 2001. Momentary interruptions are
found to constitute 67% of overall costs, relative to 33% incurred from sustained interruptions.
4
As displayed in Figure 1-3 below, the commercial sector bears the lion share of these outage
Figure 1-3: Estimated interruption costs by customer class (2002 CPI-weighted dollars) [8]
Based on 2002 survey data collected by the MidAmerican Energy Company and focused on
commercial, industrial, and organizational/institutional customers, the cost of power outages for
a variety of durations have been estimated [9]. This survey takes into account frequency,
duration and cost or inconvenience factors associated with outages. Table 1-2 below summarizes
the results of this survey in terms of average costs per event, per annual kWh and per demand
Duration of Average cost per Average costs per annual Average cost per kW
interruption interruption event ($) kWh interruption ($) demand ($)
1 min 379 0.00206 9.03
20 min 744 0.00705 30.87
1h 1,002 0.00875 37.52
4h 2,299 0.02766 121.15
8h 4,188 0.05146 225.41
Table 1-2: Estimated cost-per-event of power interruptions by duration and consideration [9]
5
To varying levels, commercial buildings are unable to tolerate power interruptions and unreliable
power quality, and therefore require supplementary power supply infrastructure. The type and
size of these systems depends of the level and nature of need, as determined by the business
operations performed in the buildings. These supplementary power systems may take the form of
standby and/or emergency backup power. Standby systems provide power after the detection of a
power outage or disturbance, with the possibility of short interruptions to load supply. On the
other hand, emergency systems provide more seamless supply to sensitive loads during loss of
ordinary power. They are typically intended to provide power for shorter durations, in order to
buffer supply before ordinary power returns or the standby system begins operation [10].
The level of need for emergency and standby power systems also differs among commercial
Mandatory installations to meet federal, state, county, and municipal regulations. These
regulations are stipulated by the National Electrical Code (NEC), the National Fire Protection
Associate (NFPA) and the American National Standards Institute (ANSI) [7].
Based on nature of need, commercial buildings typically employ emergency and backup power
system for two major reasons: (1) to preserve health and personal safety, and (2) to avoid
interruptions in critical processes, which cost time and money to restore [10].
There are different types of emergency and standby supply solutions typically deployed in
commercial buildings. In [7], these systems are classified into generator and stored energy
6
systems. The chart in Figure 1-4 below represents the major classes of prevalent emergency
standby systems.
Emergency
and Backup
Power Systems
Mechanically
Engine-driven Turbine-driven Battery
stored energy
generators generators systems
systems
The above factors contribute to an elevated demand for auxiliary power systems in some
particular classes of commercial buildings. The case-study performed in [10] compares the needs
and requirements for four main sensitive commercial customer classes: generating stations and
substations, hospitals, voice and data communication centers, and data centers. Table 1-3 below
displays representative installations for each class, including the critical loads and implemented
backup solutions.
7
System Backup System Power
Installation Critical loads
voltage Emergency Standby rating
Generator station 100 V dc 12 hr - 5 kW Lighting, motors, PLC,
and substation battery protective relays
Hospital 10 and 0.4 kV 0.5 hr UPS Diesel 1.9 kW Life safety and life
ac generator support loads
Voice and data -48V dc and 4 hr battery Diesel 150 kW Switchboards and
communication 230 V ac generator computer loads
center
Data center 400 V ac 0.7 hr UPS Diesel 160 kW Server, fire alarms and
generator for protection
critical loads
Table 1-3: Auxiliary power systems for representative high-sensitivity commercial classes [10]
Diesel generators have traditionally served as the prevailing option for backup generator
applications [7] [9] [11] [12] [13]. This is also visible in the representative commercial building
installations shown in Table 1-3. In 2001, there were an estimated 350,000 stationary diesel
generators in the U.S., totaling more than 125 gigawatts (GW) [12]. The key factors that make
these systems favorable are technological maturity, relatively low cost, availability and
The predominance of diesel generators in commercial emergency and backup applications has
brought about concerns regarding pollutants given off as byproducts of their combustion process.
In particular, they are documented to emit significant levels of particulate matter (PM) and
nitrogen oxides (NOx), as well as carbon-monoxide (CO) and hydrocarbons (HC) [11] [12] [13],
which are hazardous to health and/or environmentally harmful. As displayed in Figure 1-5
below, the PM and NOx and emissions from diesel generators are higher than all other
generation sources, with NOx rates up to 200 times higher than combined cycle natural gas
turbines [12]. This problem is exacerbated by the fact that most commercial-scale
implementations are located closer to populated areas [11] [12] and their emissions are usually
8
Figure 1-5: NOx and PM10 emission factors from diesel generators relative to other sources [12]
Due to these high emission rates, the U.S. Environmental Protection Agency (EPA) introduced
regulations to reduce emissions from diesel-powered non-road equipment. This increases the
design complexity of the systems [13] [14], and thus has the potential to increase their cost.
9
1.3 Problem definition and research objectives
through investments in traditional backup infrastructure, which typically take the form of
standby generators with battery UPS backup systems for critical loads. These systems fail to
maximize cost and operational efficiency, while offering limited ancillary benefits related to
These challenges have the potential to be addressed through the implementation of solar
Backup battery reserve for critical loads, in the event of power outages
service markets
Despite these advantages, the diverse nature of commercial building applications leads to
varying microgrid operational objectives and consequent disparities in their design and
configuration. Thus, there is currently limited information on the technological and economic
tradeoffs inherent in the use of various battery chemistries in different operational modes. This
has led to ambiguity in assessing the operational and economic viability of these systems as
Therefore, the goal of this research is to characterize the economic and technological tradeoffs
associated with the utilization of a PV-battery commercial building microgrid for selected
operational objectives and functional modes, as well as the impact of battery chemistry on
10
system revenues. This will be achieved by defining the control framework and developing a
management system, based on input data from ongoing microgrid development projects at
This chapter introduces the pertinent energy problem in commercial buildings summarizes the
objectives of this thesis. Chapter 2 will include literature review of the current microgrid
landscape, including challenges, current work and research opportunities. Chapter 3 will provide
details of the proposed model formulation. Chapter 4 will describe the data inputs and
components featuring in the proposed model. Chapter 5 will display and discuss the model
simulation results. Finally, Chapter 6 will summarize the conclusions from this research.
11
Chapter 2: Literature Review
The global solar photovoltaic (PV) market has experienced prodigious growth in recent years, in
attempt to meet growing electricity demand and curb emissions from fossil fuel consumption.
According to the Solar Energy Industries Association (SEIA), 4,751 MW of PV capacity was
installed in the U.S. in 2013, representing a 41% increase from 2012 and 29% of all new
electricity generation capacity [15]. As at the end of 2013, the cumulative solar electricity
capacity is 13,000 MW, with about 6,000 MW of PV forecasted to come online in 2014 [15].
This increase in solar implementation has been accompanied by a steady decline in system cost,
with an average PV installed system price of $2.59/W at the end of 2013 [15]. This reflects a
15% drop from 2012 and 60% since 2011 [15]. The annual trends in PV installed capacity and
12
Figure 2-1: Annual trends in U.S. PV installation capacity and average system price [15]
In particular, PV systems are increasingly being implemented in the U.S. building sector. As
visible in Figure 2-1, non-residential systems sector (commercial, government, school and non-
profit installations) accounted for 1,112 MW (23.4%) of installed capacity in 2013, representing
a 33.8% increase from 2011. These PV systems are either grid-connected or off-grid, and
typically take the form of (1) ground-mounted; (2) roof-mounted; (3) shade structure; or (4)
PV system integration with commercial buildings is important in the context of the current
energy market, which is characterized by high dependency on electricity and increasing concerns
over greenhouse gas emissions. In general, PV systems have the advantages of zero greenhouse
13
or toxic gases, reclamation of degraded land, reduction of transmission lines from electric grids,
diversification and security of energy supply, and acceleration of rural electrification [17].
Specifically considering buildings, PV systems are utilized to perform the following functions
[18]:
Operation of “active” (electrochromic) windows, air movement systems (fans) and fluid
Table 2-1 below summarizes some benefits for various types of PV systems in buildings.
14
Despite these advantages, the penetration rate experienced by PV systems has been hindered by a
(2) Production costs are non-competitive with fossil fuels. According to the U.S. Department
of Energy (DOE), solar electricity lifetime prices must drop to $0.06/kWh in order to be
(4) Intermittencies associated with solar radiation leads to unstable production and difficulty
On a global scale, government legislature has been directed at offsetting the cost of PV and other
renewable systems, with the main objective of limiting the environmental impact of the energy
sector, reducing fossil fuel dependence and fostering industrial development [17]. In particular,
the U.S. attempts to lower these system costs through a set of subsidies and incentives.
Renewable portfolio standards (RPS) have been set by U.S. states, prescribing a proportion of
energy production to be derived from renewable sources, such as solar, wind, geothermal and
biomass. As of 2011, 28 states have adopted this mechanism, with the majority employing a
system of renewable energy credits (RECs) and solar renewable energy credits (SRECs) to
promote utility compliance [17] [21]. Federal investment tax credits (ITC) offer 30% subsidies
and accelerated depreciation schedules for commercial installations [21]. In addition, states and
Third-party ownership is the predominant business model for solar installations in the non-
residential and utility sectors, in which a power purchase agreement (PPA) stipulates that the
15
third-party owns the system and the consumer makes regular payments to this owner. The PV
system is located on the consumer’s facility, where they utilize the generated power. This
arrangement is aimed at deferring the sizable upfront cost associated with PV system deployment
[21].
The frequency and magnitude of recent blackouts in the U.S. highlighted the vulnerability of the
grid to failure resulting from natural disasters and unpredictable phenomena. The need for
needs, additional stress resulting from liberalized electricity markets, and increased dependence
Distributed energy resources (DERs) have emerged as a means to improve grid reliability, curb
emissions and provide service differentiation [22] [23]. DERs take the form of small medium
and low voltage (MV and LV) generation sources both in utility networks and downstream of the
meter, such as reciprocating generators, gas turbines, microturbines, fuel cells, combined heat
and power (CHP) systems, electricity storage, as well as renewables like PV, wind turbines and
The degree to which DER interconnection can provide the above benefits may however be
limited. Firstly, the intermittency of renewable resources often leads to power fluctuations with
resulting negative impacts on grid power quality [22]. Secondly, in the event of grid shutdown,
DERs would be unable to deliver power while their output is required [22]. Thirdly, operation
and control of a huge number of DERs may prove a significant challenge [23]. In addition, the
16
prevailing contingency management mechanism involves load-shedding, and disconnection of
Microgrid systems offer potential solutions to the above drawbacks of DERs. IEEE Std 1547.4-
2011 defines microgrids as electric power systems (EPS) that: (1) have distributed resources and
load, (2) have the ability to disconnect from and parallel with the area EPS, (3) include the local
EPS and may include portions of the area EPS, and (4) are intentionally planned [24].
Technically speaking, they may also be defined as low-voltage distribution networks located
consist of distributed generators, distributed energy storage and controllable loads, and are
capable of operating in both grid-connected and islanded (autonomous) modes [25]. Based on
[22], the following chart summarizes the key features of microgrid systems.
Microgrid System
The above features contribute to provide the following benefits, as characterized by [24]:
Improve reliability by providing power to the islanded portion of the EPS during an
17
Relieve area EPS overload by allowing intentional islanding of individual systems
Isolate individual system from area EPS power quality issues such as voltage distortion,
Resolve power quality issues by reducing total harmonic distortions at the loads
Although microgrids vary in technology and application, they are typically made up of a
microgrid systems are (1) distributed generators (DG), (2) distributed energy storage (DES), (3)
controllable loads, (4) critical loads, and (5) point of common coupling (PCC) [25]. Table 2-2
summarizes key functionalities and examples of the above major microgrid components.
Microgrid architectures may be classified by application, ownership structure and type of loads
[22]. On the basis of grid-integration, microgrids may be grid-connected or act as an isolated grid
in remote microgrid architectures. This thesis research focuses on grid-connected systems, which
can be further categorized according to scale: (1) utility microgrids (2) commercial/industrial
18
microgrids (3) residential microgrids. Figure 2-3 below displays an example of microgrid
Utility microgrids are located on a portion or the entirety of distribution substation feeders
managed by distributed network operators (DNO), and are aimed at meeting load growth and
order to limit the influence of the main grid on power quality and reliability. As shown in Figure
2-3, these microgrids may be deployed to serve single or multiple facilities, each with load
classification and service differentiation mechanisms aimed at defining power quality and
19
reliability requirements of the loads. These systems may experience unplanned islanding during
microgrid systems will form the focus of this thesis research. Specifically, the term building-
integrated microgrid (BIMG) has been used to refer to buildings with decentralized and
cooperative architectures for local power flow balance between loads and renewable power
Residential microgrids may also be deployed to serve small residential customers. They offer
convenient and reliable supply which is customized to the customer’s demand [22].
There have been increased efforts to develop and study microgrid systems on a global scale.
Overviews of microgrid research and testbeds in various economies are presented in [23], [27]
and [28]. These development efforts are mostly concentrated in the European Union (EU)
countries, North America and Asia. Representative testbeds in the EU include, the “Microgrids
Project” in Kythnos Island, Greece, the ISET microgrid in Germany, and the “More Microgrids
Reliability Technology Solutions (CERTS) is the most-quoted microgrid testbed. In Asia, Japan
leads the research and development efforts, with most projects funded by the New Energy and
these development efforts are highlighted in [27]. North American countries are generally
focused on exploring the power supply reliability benefits of microgrid systems, with a
prevalence of autonomous control for maintaining voltage and frequency. On the other hand,
European and Asian efforts display a greater emphasis on renewable energy utilization, in
addition to power quality and reliability. While Asian systems are predominantly centrally
20
2.3 Battery systems for microgrid applications
Energy storage systems are essential in microgrid applications, where they are utilized to
mitigate intermittency in renewable energy sources, improve microgrid stability and perform
ancillary services to the grid [29]. Figure 2-4 below summarizes operational uses for storage
Battery storage systems, categorized either as cell or flow batteries, are the most frequently
utilized storage applications [31]. Flow batteries, which are typically more desirable in large-
scale projects, store their electrolytes in separate tanks and moved into electrochemical cells
using pumps, while cell batteries store their electrolytes in the same location as the chemical
reaction [31]. This thesis research focuses on the application of cell batteries in small- to
medium-scale commercial microgrid systems. Other types of energy storage systems in power
21
system applications include compressed air energy storage (CAES), pumped hydro storage
(PHS), hydrogen fuel cells, flywheels, supercapacitors, and superconducting magnetic energy
There are various cell battery technologies with varying levels of suitability for microgrid
applications. The leading cell battery chemistries on the market are lead acid, lithium ion (Li-
ion), sodium sulfur (NaS), and nickel cadmium (Ni-Cd) [31] [32]. Table 2-3 below summarizes
The pros and cons of these technologies for power system applications are compared in [32].
Among the above battery chemistries, lead acid batteries are the oldest and most mature
technologies, while Li-ion, NaS and Ni-Cd are favorable due to their high energy density. Li-ion
is considered to hold the greatest potential for future development and optimization, since they
possess the highest energy density and storage efficiency close to 100%, despite their key
drawbacks of high cost and detrimental effects of deep discharging. On the other hand, lead-acid
and Ni-Cd batteries can supply good pulsed power at lower costs, although they are large,
contain toxic heavy metals, and suffer from severe self-discharge. While NaS batteries are
relatively small and light, they require constant heat input to maintain molten states of
22
electrolytes. Figure 2-5 below displays comparisons between these cell batteries and other
23
2.4 Microgrid energy management systems and modeling
The role energy management system (EMS) is to allocate power output from distributed
generation, economically serve the loads, and orchestrate transitions between interconnected and
islanded operational modes [25]. This operation typically utilizes three hierarchical levels of
control: (1) the distribution network (DNO) and market operator (MO), (2) the microgrid central
controller (MGCC), and (3) the local controllers (LC) [25] [28] [34]. The MGCC acts as an
interface between the greater grid (DNO and MO) and the microgrid LCs, which control the
individual components of the microgrid (such as DG, storage, loads and protection equipment)
Figure 2-6: Hierarchical control framework of microgrid EMS (PCC-point of common coupling;
MGCC-microgrid central controller; SD-separation device; CB-circuit breaker; GC-DG
controller; ES-storage device; LC-local controller) [34]
24
Based on the EMS decision-making process, system control may be considered centralized or
decentralized. In the centralized case, the MGCC manages the system with emphasis on market
prices, while decentralized control responds to the revenue- and performance- maximizing
Widespread microgrid research efforts have been directed at EMS modeling with the aim of
implementation. The roles performed by the EMS necessarily involve economic and technical
optimization, in order to ensure that desired system integrity and performance are achieved with
Certain tools have been built specifically for microgrid EMS modeling. The Distributed Energy
Resources Customer Adoption Model (DER-CAM) has been developed at Berkeley Lab since
2000. DER-CAM takes in load profile, price, system cost, and system technical data, and
minimizes the cost of operating DG and CHP systems. Model outputs include optimal capacities
and running schedules of DG and CHP systems, and total cost of supplying electric and heat
loads [35]. As an example, this modeling tool is utilized in [36], which is concerned with the
solar thermal collector, electrical storage, and thermal storage system to meet both electrical and
thermal loads of a California hotel, for four separate system operational objectives. The results of
this optimization are the best combination of these technologies, in terms of operating capacity
25
Existing microgrid EMS research has focused on modeling various aspects of the microgrid
system. Certain building-integrated microgrid EMS models take into account all building system
loads. For example, [37] explores the ability of a building microgrid to promote energy
gas costs in a building with electric, thermal, HVAC, as well as electric vehicle (EV) loads, in
the presence of utility, CHP, fuel cell, PV, wind, and battery inputs. On the other hand, some
studies are focused on modeling the performance of specific generation and storage components
of the system. In [38], a microgrid EMS model is proposed for optimal dispatch of PV energy
and controllable loads, based on forecasted PV production and load in a system with embedded
storage units and a microturbine system. [39] presents equivalent electrical and models for wind-
battery integrated system concerned with output smoothing. [40] analyzes the economic value
associated with load levelling, control power and peak-shaving applications for four different
battery types. In addition, [41] applies a battery degradation model is in assessing the economic
There are a number of different modeling objectives implemented in different microgrid EMS
studies. Cost minimization in serving loads is a common feature of these EMS models, as earlier
discussed in [36]. The dual objective of profit maximization has also been frequently explored.
These profits may be derived from sales to the energy market as in [42], which develops a day-
ahead optimization for a microgrid consisting of wind, microturbine, PV, fuel, electrolyzer, H2
storage, reformer, boiler, and electrical and thermal load systems. On the other hand, [43]
26
Another branch of microgrid EMS research focuses on optimal size of the individual system
components, as earlier discussed in [36] where the optimal capacity and combination of various
DG sources is determined for different operational objectives. In [44] also, the optimal energy
storage system size is determined for a microgrid with PV, wind and microturbine DGs in a
mixed linear integer unit commitment formulation for both islanded and grid-tied mode, which is
solved using AMPL. Aside this, some studies are primarily concerned with maximizing the
performance of microgrid subsystems. For example, [45] and [46] both propose models to
The above studies have failed to emphasize and concisely depict the technological and economic
particularly as sources of revenue from portfolios of ancillary service and energy markets. This
issue is exacerbated by the presence of multiple system objectives, components, and parameters
across various implementations. In order to make a more compelling case for these systems,
further studies are required to quantify these tradeoffs, as well as their relationship to the degree
27
Chapter 3: Model Formulation
The purpose of this research is to characterize the economic tradeoffs associated with the use of a
Based on system data and specifications from the Building 7R commercial microgrid
development at the GridSTAR Center in the Philadelphia Navy Yard, an operational strategy has
28
The system architecture of the microgrid to be modelled is motivated by the Building 7R
development at GridSTAR Center in the Philadelphia Navy Yard, Philadelphia, PA. Figure 3-1
(1) DC inputs: The 7R system accepts DC inputs from the PV distributed generation
The DC input section consists of load-break DC contactors to separate the battery from
the DC/DC converter as well as a DC pre-charge circuit to protect against power surges
(2) Bidirectional inverter: The system features a bidirectional insulated gate bipolar
The output of the filter is then put through an AC contactor to separate the filter and
inverter from the grid. In the event that the inverter needs to be bypassed (grid to load)
The scope of this thesis does not include a characterization of the inverter power
electronics. Thus, the developed model assumes lossless inverter operation and perfect
(3) AC Output: This portion of the system contains a 480V AC 3-phase grid interconnection
The grid interconnection serves as the PCC with the greater distribution system, through
which the microgrid interacts with the energy and ancillary service markets.
29
The AC load connection serves all the building electrical loads, including special
designation for the data servers, which make up the critical loads.
Also included are voltage sensing potential transformers (PT) and current sensing current
transformers (CT), which are used for voltage, current, power, and power factor controls
and monitoring. These sensors also detect grid restart and are used to synchronize the
system to the grid for reconnection. This section also contains an isolating transformer as
(4) DC Loads: The DC load section consists of three DC outputs with rated nominal voltage
of 500-700VDC and maximum voltage of 460-715VDC. The loads served include a rapid
charge EV charging unit, a DC grid classroom facility, and a DC mini split air
conditioner.
(5) Energy Management System: The microgrid EMS encompasses the site control, power
electronics control and battery management system, as well as the building automation
system (BAS). The EMS is responsible for controlling transitions between grid-
connected and islanding operation, grid communications for ancillary services, battery
cycling operation, and BAS controllable loads such as lighting and temperature systems.
As indicated in Figure 3-1, the EMS model in this study is primarily concerned the
functions related to orchestrating energy flows between the PV system, batteries, utility,
regional transmission organization (RTO) and the building loads. This functionality is
The EMS model does not include electrical circuit-level characterizations of the
30
Figure 3-2: Functionality of Building 7R microgrid EMS (based on [47]).
As shown in Figure 3-2, based on the framework developed in [47], the building has two
contracts A and B. Contract A reflects the agreement for electricity demand and net metering
payments with PECO, which is the utility local to the Philadelphia Navy Yard. Contract B
reflects the agreement with the PJM RTO for ancillary services. In orchestrating energy flow
between the utility, RTO, battery, PV system and loads, the central objective of the EMS is to
maximize revenues from building energy demand curtailment, while providing ancillary
market participation and backup reserve capacity, subject to battery performance limitations.
31
(1) Minimization of building energy demand purchases
meeting the building load demands. Any excess PV is recycled based on control logic
energy bills by successive charging and discharging of the battery during the lowest cost
(off-peak) and highest cost (on-peak) periods of the day, respectively. This involves the
formulation of an optimization problem for the battery cycling operation on a daily basis,
It is assumed that the 7R microgrid is located within the PJM RTO region, in which
frequency regulation has been determined to be the most profitable A/S [48]. The EMS
meets this second objective by means of the frequency regulation operational mode.
specific resource within the appropriate telecommunications, control and response capability
The EMS offers the battery capacity to the PJM market as a dynamic regulation (Reg-D)
resource, which is a designation for fast-cycling resources [50]. This battery capacity
accrues revenue when it is utilized by PJM as a power source or sink in the event of
32
instantaneous grid power undersupply or oversupply, during the bid period. The bidding
The EMS aims to fulfil both of the above operational objective subject to user-defined
minimum stored battery backup capacity, battery performance limitations, and RTO
regulations. To this effect, a battery capacity allocation approach has been adopted
towards providing the above-defined functionalities. In other words, the usable battery
capacity is partitioned between (1) peak-shaving operational mode, (2) frequency regulation
operational mode, and (3) backup battery reserve. This is depicted in Figure 3-3 below.
The usable battery capacity is defined by the maximum and minimum state of charge limits,
as determined by the implemented battery technology. Figure 3-4 below summarizes the
framework employed by the EMS in meeting the central system operational objective.
33
Maximize Local PV
Minimize Building Generation
Energy Demand
Purchases Peak-Shaving
Operational Mode
This section presents the proposed control strategy for the Building 7R EMS model, based on the
input information to be discussed in the following chapter. As earlier stated, the operational logic
presented in this research focuses on the functions of the microgrid EMS relating to the transfer
of energy between the PV system, batteries, utility, RTO and the building loads. The model does
not include circuit-level characterizations of the distribution systems, protection devices and
power electronics. The thermal and electrochemical behavior of the battery systems are not
considered in this analysis, with the assumptions of limited impact of disparities in performance
between cells on the larger array, as well as periodic equalization charging and uniform
34
Figure 3-5: Overall EMS control strategy. (1) Peak-shaving optimization; (2) PV recycling; (3) Operational mode decision
35
Figure 3-5 above displays the overall control strategy implemented by the EMS model. At a high
Based on the entailments of the peak-shaving operational mode, in which the EMS attempts to
offset building electricity costs by capitalizing on hourly fluctuations in energy price, the control
framework is evaluated in 24-hour intervals. Thus, as shown in Figure 3-5 below, the first and
last portions of the control algorithm define the conditions for resetting EMS control operations
on a daily basis.
The data inputs discussed in section 3.3 and the control framework in Figure 3-5 above are based
on a deterministic approach, which means the EMS makes decisions with the assumption of
perfect knowledge of pertinent variables. Thus, at the start of each day, building demand,
electricity prices, PV production and frequency regulation data are compiled for the 24-hour
duration. With the aid of this information, the major decisions carried out by the EMS are as
follows:
(1) Peak-shaving optimization: The first priority of the EMS is to determine the optimal
hours each day to allocate battery capacity towards peak-shaving operation. This decision
is made based on predefined battery capacity allocation, battery operational limits, hourly
36
(2) PV energy recycling: In line with the objective of maximally utilizing local PV
generation, the EMS directs all PV output towards the building loads. On an hourly basis,
any excess PV production over demand is sent to the battery, after which the balance is
This operation assumes the presence of a net-metering agreement with the utility, in
which a market exists for excess PV at prices equal to the hourly LMP.
(3) Determination of operational mode: With information on optimal peak-shaving hours and
PV energy flows available, the EMS proceeds to make decisions on the operational mode
in effect each hour, which impacts the load-serving mechanism employed and nature of
grid interaction. This functionality is depicted in the loop 3 in Figure 3-5 above.
As shown, the EMS charges or discharges the battery if the pertinent hourly period has
been optimally designated for peak-shaving. After discharging the battery to the loads,
Note that within the battery capacity allocation framework discussed in section 3.2, it has
been assumed that the battery does not simultaneously provide peak-shaving and
necessary in order to guarantee conformance with the predefined allocations and battery
operational limits. Thus, in hours where the battery does not provide peak-shaving, the
EMS bids the predefined battery allocation for frequency regulation into the PJM market.
When the system participates in regulation services, the building energy demand is
37
3.4 EMS simulation model
This section presents the simulation model developed to carry out the functions described in
this chapter. It accepts hourly inputs of building demand, electricity prices, regulation market
data, PV production, user settings, and battery parameters, and outputs hourly demand utility
purchases, battery energy in-flow and out-flow, and market revenues. This analysis is
performed on a daily basis for the duration of one year. Figure 3-6 below depicts an overview
As discussed in section 3.4, the decision-making performed by the system involves three key
mode. MATLAB has thus been selected as the microgrid EMS simulation tool due to the
38
implementation of the outlined logical decision-making process, in addition to compatibility
features with Excel spreadsheets for easy import and export of relevant data sets.
The peak-shaving optimization problem has been formulated as a linear maximization of net
revenues obtained by the system from mitigated energy cost, ancillary market participation and
excess PV sales, in the midst of building demand purchases. As shown in Figure 3-5 above, this
optimization is calculated by the EMS on a daily basis, for the duration of one operational year.
INPUT PARAMETERS
39
DECISION VARIABLES
OBJECTIVE FUNCTION
The first two terms in the objective function describe the outflow and inflow of energy from the
battery in response to hourly price fluctuations. On a daily basis, the optimization aims to
maximize the difference between mitigated on-peak costs and purchased off-peak energy. The
second term represents purchases to serve building demand, which is minimized by the function.
The third term shows the revenue from excess PV recycling, which is made up of the avoided
cost from charging the battery with solar and excess net metering PV sales.
(1) ∑ ( )
40
This ensures that the daily quantity stored in the battery towards peak-shaving, which
includes utility purchases and excess PV, equals the total capacity allocated towards
peak-shaving. This is scaled by the battery state-of-charge range and maximum constant
charge rate.
(2) ∑ ( )
This sets the total battery intake from PV and utility purchases equal to the total energy
discharged during peak hours, scaled by the battery efficiency. Therefore, the peak-
shaving allocation is always satisfied. It also includes an embedded assumption that the
(3)
This guarantees that the building demand is served at every given hour, either through
(4) ( )
This ensures that the hourly capacity charged to the battery does not exceed its
maximum charge rate. Thus, the maximum energy stored during off-peak periods is the
total peak-shaving allocation, scaled by these battery operational limits, with a minimum
(5) ( )
This performs the same function as constraint (4) above in the case of battery discharge.
The maximum battery discharge in on-peak hours is given by the total peak-shaving
allocation, scaled by the battery state-of-charge, charge rate, and efficiency limits. The
41
(6)
This ensures that the system never purchases more than the building electricity demand
in any given hour. The variable also cannot attain a value less than zero.
(7)
This sets an upper limit on the amount of excess PV transferred to the battery, based on
(8) ∑ ( ) ∑
This ensures that the sum of battery charge from solar (scaled by efficiency) and sold
(9)
In line with the control framework discussed in section 3.3, this imposes the condition
that hourly sold PV energy is not more than excess PV in the system. The
(10) ∑ ∑
This ensures that the total quantity discharged from the battery from the start of the day
to any hour does not exceed the stored energy available in the battery from all
sources.
The above linear optimization problem is solved in 24-hour increments for the duration of a year,
using the linprog function in MATLAB. The simulation code developed to consolidate input and
output data, solve the optimization problem, implement the operational logic described in section
42
Note that this optimization formulation does not include a state-of-charge control variable within
the objective function. This is because such a variable would lead to inconsistency in the
objective function, since maintaining all components of in terms of revenue would entail the
introduction of a degradation cost factor, which is beyond the scope of this research. However,
due to programming limitations imposed by the linprog function, the hourly battery state-of-
charge cannot be regulated by the optimization solver, since it does not feature in the objective
function. Therefore, the following three-pronged approach has been adopted in order to ensure
compliance of battery state-of-charge with predefined allocation (see Figure 3-3) and operation
limits:
Daily peak-shaving allocation net-zero: The battery energy allocation towards peak-
Regulated discharge time: The system can only discharge what is available in the battery
Weekday peak-shaving operation: Hourly price fluctuations associated with on- and off-
peak demand periods in PJM typically do not occur on weekends and holidays [49]. Also,
significantly less than weekdays. For these reasons, it has been assumed that peak-
shaving operation takes place only on weekdays. Therefore on weekends, the system
participates strictly on the regulation market and sells excess PV to the grid.
Also note that regulation revenue does not feature in the objective function because market
prices and required regulation quantity are considered external to the optimization problem, since
43
they are based on regulation market information, predefined regulation allocation and dispatch-
The above optimization problem yields hourly quantities of the following EMS energy flows:
Conditional programming has been utilized in determining when the system participates on the
frequency regulation market, as discussed in section 3.4 (see Appendix E). Based on all of the
above information, the EMS model yields outputs based on the following calculations:
This represents the total building electricity cost mitigation derived from dedicating
utility purchases from on-peak discharge as well as PV charging, and the cost of off-peak
charging:
∑ [ ( ) ]
amount of purchased off-peak energy required to meet the allocation. It also reduces the
44
amount of effective charge from solar and on-peak discharge, both of which have been
As discussed in section 3.5, the EMS bids the battery’s capacity into the regulation
market based on the calculated optimal periods for peak-shaving and corresponding
energy flows. During each of these hours, the system receives a capacity payment based
Note that this model assumes an hourly net-zero of battery charge and discharge in
response to the signal from the RTO. In PJM training documents on ancillary services
[52], it is stated that fast regulation (Reg-D) resources (which include battery systems)
are “short term balanced to a zero state” by the system operator. This notion is also
supported by [41] and [48]. Assuming that the regulation energy flows are adjusted for
each given frequency regulation battery allocation, the EMS model leaves half of the
capacity open to allow for energy intake. For example, if the battery is partitioned in a
45
(3) Battery state-of-charge (SOC)
The value of percent SOC ( ) has been calculated hourly as the ratio of the battery’s
net energy transfer to its energy capacity, both of which are measured in kWh, i.e.:
( ) ( ( ) )
The SOC at the start of the year has been initialized to the sum of the minimum
( ) ( )
( ) ( )
As earlier noted, half of the frequency allocation is left open to allow for energy intake.
The bottom term is included in this equation to account for the possibility that the first
46
Chapter 4: EMS Model Input Data and Component Sizing
This chapter describes the input data and sizing criteria that have featured in the characterization
of the Building 7R microgrid EMS operation, based on the framework describe in Chapter 3
above.
The EMS model will utilize a simulated demand profile, as modeled by the Building 7R
architects, Kieran Timberlake. The model is developed using the DOE eQUEST analysis tool
and DOE Typical Meteorological Year 2 (TMY2) weather information for Philadelphia. The
total building area is assumed to be a two-story 25,600 square-feet (ft2) expanse. The output of
the model is the total electrical end-use demand (kWh) on an hourly basis for the duration of one
year, assuming day types and holidays from the year 2013.
The system is modeled as a classroom building with four occupancy and usage schedules. These
Due to these usage patterns, the daily peak demands vary according to day type. These variations
are displayed for representative days in each of the four seasons in Figure 4-1(a)-(c) below.
47
(a) Building electricity demand on Mondays, Wednesdays and Fridays
48
(c) Building electricity demand on weekends and holidays
Figure 4-1: Typical building energy demand profiles for different day types duing representative
days in Winter, Spring, Summer, and Spring
On Mondays, Wednesdays and Fridays (MWF), these peaks typically occur around mid-
morning, while they occur in the afternoon on Tuesdays and Thursdays (TR). In both cases,
demand is highest in July (summer). Weekends and holidays are characterized by low demand
and no identifiable peak demand patterns. The full details of the assumptions and data inputs
The hourly output building energy demand data has been catalogued for the whole year, with the
maximum, minimum and average demand has calculated for each hour. These results are shown
49
End of Average Min Max
hour (kWh) (kWh) (kWh)
1 11.24 5.15 28.58
2 11.14 5.15 28.55 Annual Demand Statistics
3 11.08 5.15 28.41 Min (kWh) 3.27
4 10.94 5.15 28.41
5 10.95 5.15 28.39 Max (kWh) 119.28
6 10.50 5.15 28.39 Average (kWh) 26.70
7 25.18 4.40 63.72
Total (kWh) 233921.58
8 40.99 4.21 105.25
9 48.71 3.27 115.89
(b) Annual building demand statistics
10 49.24 3.43 113.57
11 47.93 3.43 114.26
12 41.14 3.43 105.97
13 41.56 3.43 110.34
14 45.24 3.43 116.61
15 45.82 3.43 118.01
16 46.05 3.43 119.28
17 31.60 3.43 81.46
18 21.04 3.99 55.05
19 18.01 4.21 37.52
20 18.23 4.21 37.81
21 18.77 4.58 38.28
22 13.19 5.14 39.17
23 11.24 5.19 20.35
24 11.10 5.19 21.14
Total 640.88 102.69 1584.41
(a) Hourly building demand statistics
Based on these results, the building would demand a total of 640.9 kWh on the average day, with
minimum and maximum possible demands of 102.7 kWh and 1584.4 kWh respectively. Figure
4-2 displays the hourly demand on these days. The relatively flat demand on the low demand day
is due to the fact that these periods of minimum demand correspond to weekends and holidays.
50
Figure 4-2: Average, minimum and maximum possible building demand (based on Table 4-2(a))
The average hourly demand is 26.7 kWh, while the minimum and maximum energy required on
any hour during the year are 3.3 kWh and 119.3 kWh respectively. The total building demand in
The EMS model derives hourly energy price information from 2013 PJM day-ahead locational
marginal pricing (LMP) data for the PECO region, in which the GridSTAR Center is located
[53]. This represents the hourly market clearing marginal price for this region, based on day-
The use of this data is based on the assumption that the 7R system faces the hourly LMP, instead
of fixed rates available from electricity retailers. In this way, the EMS is able to take full
51
advantage of hourly price fluctuations and optimally utilize battery capacity towards energy cost
mitigation. Table 4-3 and Figure 4-3 below display monthly energy input price characteristics.
Avg
Avg Price
Month Price Diff
(c/kWh)
(c/kWh)
Jan 3.67 2.48
Feb 3.45 1.77
Mar 3.89 2.29
Apr 3.92 2.52
May 3.83 2.64
Jun 3.80 3.12
Jul 4.88 6.75
Aug 3.55 3.28
Sep 3.56 3.14
Oct 3.47 2.15
Nov 3.63 2.79
Dec 4.02 2.94
52
The first column in Table 4-3 displays the average hourly cost of electricity in cents per kilowatt-
hours (c/kWh). The second column shows the monthly average price differential, which captures
the daily price fluctuations as the average of the difference between the daily maximum and
( )
energy cost savings. As visible in Figure 4-3, the highest price differentials occur during the
summer months of June to September, with a peak value of about 6.75 c/kWh in July, which also
In the PJM ancillary services market, settlements from frequency regulation resources come in
(1) Regulation capability: This is a capacity payment that represents the resource owner’s
(2) Regulation performance: This portion captures the resource owner’s price to provide
regulation movement, either in the form of energy inflow or outflow from the battery.
This payment represents the absolute sum of energy inflow and outflow experienced by
the system.
PJM receives bids from resource owners in the day-ahead market, with the option of updating
resource availability up to one hour ahead. These bids are used in the calculating the hourly
regulation market capability clearing price (RMCCP) and performance clearing price (RMPCP)
53
for economical dispatch, both in dollars per megawatt-hour ($/MWh). Historical regulation
The EMS model utilizes 2013 hourly RMCCP and RMPCP data in determining frequency
regulation revenue. For various ratios of frequency regulation battery capacity allocation,
capacity payments have been evaluated using RMCCP. However, subject to the results of
economic dispatch, the actual battery capacity called upon by the system operator (i.e. mileage)
The SolarGridStorage frequency regulation battery system at the GridSTAR Center has
providing mileage data for this analysis. The system features a 250kW/125kWh battery system
with full capacity dedicated to frequency regulation services. Data from this system has been
utilized in characterizing the dispatch-to-contract ratio, which describes the ratio of hourly
battery mileage incurred to capacity made available. This data provides the net sum of energy
flow at 15-minute intervals for the whole of March, 2014 (with the exception of 47 unavailable
data points between March 7 and March 9, representing 1.58% of the total data set). It has been
assumed that the hourly mileage is represented by the data field referring to the net sum of grid
demand at each interval. This total hourly mileage has been divided by the system capacity
Note that the dispatch-to-contract ratios for unavailable data points have been assumed to be the
mean of the remaining data set. The histogram in Figure 4-4 below represents the resulting
54
distributed around the mean. The average value is a ratio of 25.1% of dispatched capacity to bid
capacity, with minimum and maximum values of 2.8% and 42.4% respectively. The standard
Min: 0.028
Max: 0.424
100 Mean: 0.251
SD: 0.049
80
Number of hours
60
40
20
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Dispatch-to-contract ratio
Due to unavailability of complete additional monthly data at the time of this analysis, the
dispatch-to-contract ratio characterization derived from March, 2014 has been applied for all
other months in the EMS model, with the assumption of negligible month-to-month variations in
the framework for regulation market bidding and clearing. Also, the performance score assigned
by PJM to regulation resources based on their degree of responsiveness to the regulation signal is
55
4.4 Battery technology sizing and specifications
The battery storage component of the building 7R microgrid model will play an essential role in
fulfilling the system objective of revenue maximization through energy cost mitigation and
ancillary service market participation. With this in mind, performance of lead-acid and lithium-
ion battery chemistries will be compared within the framework of this analysis. As shown in
Table 4-4, these technologies provide reasonable contrasts in maturity, performance, physical
characteristics and cost. Lead-acid systems represent the established technology for hybrid
renewable power systems due to their low cost and technological maturity [31] [32] [56], with a
ion systems on the other hand represent the emerging technology in the industry [32] [56], with
superior performance capabilities and physical characteristics, in spite of high cost and relative
immaturity. Table 4-4 below qualitatively compares the pros and cons of both battery
chemistries.
Pros Cons
Lead-acid Most mature technology. Relatively low energy density,
Low cost. efficiency depth of discharge, and
Established manufacturing base. cycle life. High toxicity and
temperature sensitivity.
Lithium-ion High efficiency and energy density, High cost.
and cycle life. Small weight and size. Relatively immature, with high
Low toxicity and maintenance manufacturing complexity.
requirements.
Table 4-4: Qualitative comparison between lead-acid and lithium-ion systems [31] [32] [56]
In sizing the battery systems for the EMS model, PJM regulations for ancillary market
participation served as the primary consideration. According to [54], generation and demand
56
resources must be able to provide a minimum of 0.1 MW (i.e. 100kW) of regulation capability
for market eligibility. Based on this, both battery systems in the model have been sized for
In the lithium-ion case, available pre-packaged specifications for 100 kW/107.74 kWh system
have been utilized in the EMS model. The array contains two rows and five columns of Kokam
KBM255 2P lithium polymer battery modules. The relevant specifications for this system are
listed in Appendix B.
In the case of the lead-acid system, a suitable battery configuration must be determined to meet
the 100 kW criteria, based on manufacturer model catalogs. Performance specification catalogs
for the Enersys PowerSafe SBS valve regulated lead acid (VRLA) battery series provide nominal
capacity, nominal voltage and constant power discharge rates for a range of battery models (see
[57]). The SBS 190F model has been selected, since it displays the highest nominal capacity and
The sizing exercise for this lead acid system involves selecting the minimum number of modules
configuration, with aggregate nominal voltage within acceptable range of typical inverter DC
inputs. For the purpose of this analysis, a DC input range of 280-600V has been selected, based
on the specifications for a 100kW Princeton Power DRI-100 inverter [58]. Figure 4-5 below
57
Figure 4-5: Illustration of lead-acid array sizing problem
In order to ascertain the minimum number of cells and array size to achieve the desired output, a
simple mixed-integer nonlinear genetic algorithm optimization has been developed in MATLAB,
Where
58
The results of this optimization model are integer numbers of battery module rows and columns
which minimize the number of cells required to achieve an output voltage between 280 and 600
V, as well as a constant discharge rates no greater than 0.5% above 100 kW. This analysis yields
an optimal array size of 37 columns, 5 rows, 444 V and 100.34 kW. This yields a nominal
capacity of 950 Ah, and thus an energy capacity of approximately (444 V)*(950 Ah) = 421.8
kWh1, which is about four times larger than the lithium-ion case with similar constant discharge
rate.
The key specifications to feature in the EMS model for this battery system, as well as the
lithium-ion system, are displayed in Table 4-5 below. Note that the expected efficiency at the
specified C-rates are not typically included in manufacturer datasheets. Thus these values are
obtained from [56], with the assumption that the battery systems under study display similar
efficiency at these C-rates. Also, C-rate of C/4 has been used in the lead-acid case due unfeasibly
low efficiencies associated with typical systems at the 1C rate used in the lithium-ion case.
1
This derivation of kWh energy capacity as the product of Ah capacity and nominal output voltage assumes that
the nominal voltage stays relatively constant throughout discharge
2
In the lead-acid case, it is assumed that the maximum charge and discharge rates are equal
59
4.5 PV system production
monocrystalline PV array. For the purpose of this analysis, this PV system is assumed to have
been sized externally based on the building loads. Table 4-6 below, summarizes the key technical
Rated Power Open circuit Short circuit Max power Max power Efficiency
voltage current point voltage point current
260 W 38.1 V 8.98 A 31.1 V 8.36 A 15.7%
The EMS model makes use of simulated PV output based on the technical specifications of the
PV system in Table 4-6 above, as well as Typical Meteorological Year Type 2 (TMY2) data
available from the DOE, which also served as weather input for the eQUEST load profile model
in 3.3.1. These are fed into the TRNSYS model displayed in Figure 4-6 below, which outputs
expected PV production on an hourly basis for the duration of the year. The model details and
60
4.6 Outage cost characterization
In order to attach a monetary value to the functionality of the microgrid relating to battery
backup reserve, a characterization of the degree of power outage costs has been developed, based
In [8], a comprehensive end-use framework for quantifying the cost of power interruptions and
power quality events to U.S. electricity consumers has been developed, which takes into account
number of customers by class and region, duration and frequency of annual reliability events,
costs and vulnerability of customers to these events. This analysis has derived SAIDI (System
Average Interruption Duration Index) and SAIFI (System Average Interruption Frequency
Index) values of 106 minutes (54 minute standard deviation) and 1.2 (0.5 standard deviation).
This means that on average, U.S. customers will face 1.2 106-minute reliability events every
year. Based on this information, the EMS model in this thesis research has assumed equal-
Table 4-2(a) above displays the building energy profile for low, average and high demand days
during the year. Considering this data, the 2.12-hour periods of highest demand have been
determined to be about 11.0 kWh, 103.1 kWh, and 250.1 kWh on low, average and high demand
days respectively. These values represent potential energy losses experienced by the building in
Thirdly, [60] presents data on interruption costs for residential and commercial/industrial (C&I)
customer groups, based on twenty-four studies conducted by eight electric utilities between 1989
and 2002. This report has determined the cost per peak kW for small to medium C&I customers
61
to be $40 for 1-hour3 durations. Adjusted to 2013 CPI (in uniformity with other data inputs) and
from when the report was released in 2003, the EMS model in this thesis study assumes a cost
The above information has been combined with the energy capacity and efficiency of each
battery system discussed earlier in this chapter. Altogether, Table 4-7 below represents costs
avoided by the building in the event of power outages by allocating varying proportions of the
Table 4-7: Avoided outage cost for varying backup reserve allocation on different day types
During average- to high-demand days, the lead acid system has higher potential for blackout cost
mitigation, due to higher energy capacity required to produce the same maximum power
3
1-hour duration has been assumed, since the next highest duration available in the [60] data is 4 hours. Also, this
value provides a conservative estimate to the SAIDI value of 106 minutes.
62
Chapter 5: Results and Analysis
This section presents the results of the Building 7R simulation, based on the model discussed in
Chapter 3. Daily battery energy flows, utility purchases, and revenues from peak shaving are
presented. Also presented are sensitivity analyses with respect to relative allocation percentages
between peak-shaving, frequency regulation and backup power, as they impact overall system
revenue.
In order to evaluate the daily building energy consumption profile, the battery backup allocation
has momentarily been constrained to be zero. This means that the EMS can dedicate the entire
usable capacity toward peak-shaving and frequency regulation operation. Also, for this initial set
of simulations, the battery has been allocated in a 50%:50% ratio between backup, peak-shaving
and regulation respectively. In the lithium-ion case, this corresponds to a SOC ratio of 0%: 40%:
40% (assuming 80% depth of discharge) an energy ratio of 0 kWh: 43.10 kWh: 43.10 kWh.
Table 5-1 below represents the one-day EMS simulation for November 18, using the lithium-ion
battery specifications. Columns to the left of the indicated demarcation are model inputs, while
those to the right display model results and output calculations. The highlighted cells show hours
in which the battery is charged from off-peak utility purchases and excess PV (orange) or
discharged to the loads (purple). During all other periods of the day, the battery is bidding into
the frequency regulation market, as reflected by the corresponding non-zero regulation revenue
values.
63
4
(c/kWh) (c/kWh) (c/kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (%) ($) ($)
1 2.40 0.50 0.18 11.17 5.43 0.00 0.00 0.00 5.43 0.00 0.00 0.00 30.00 0.00 0.22
2 2.23 0.32 0.43 9.87 5.43 0.00 0.00 0.00 5.43 0.00 0.00 0.00 30.00 0.00 0.17
3 2.11 0.31 0.11 0.00 5.43 0.00 42.95 0.00 5.43 0.00 0.00 0.00 69.87 -0.99 0.00
4 2.12 1.21 0.11 9.00 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 69.87 0.00 0.49
5 2.27 1.79 0.09 9.37 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 69.87 0.00 0.72
6 2.65 0.08 0.05 11.48 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 69.87 0.00 0.04
7 3.57 0.61 0.37 10.57 5.02 0.00 0.00 0.00 5.02 0.00 0.00 0.00 69.87 0.00 0.28
8 3.31 2.11 0.22 9.98 8.23 1.28 0.00 0.00 6.95 0.00 0.00 0.00 69.87 0.00 0.87
9 3.36 1.65 0.31 8.34 11.18 2.80 0.00 0.00 8.39 0.00 0.00 0.00 69.87 0.00 0.68
10 3.45 1.86 0.30 7.46 10.42 6.18 0.00 0.00 4.24 0.00 0.00 0.00 69.87 0.00 0.77
11 3.58 1.41 0.27 9.98 10.26 8.76 0.00 0.00 1.50 0.00 0.00 0.00 69.87 0.00 0.59
12 3.53 1.91 0.44 12.31 10.42 9.65 0.00 0.00 0.77 0.00 0.00 0.00 69.87 0.00 0.82
13 3.48 2.25 0.36 0.00 10.89 11.05 0.00 0.00 0.00 0.14 0.00 0.15 70.00 0.00 0.00
14 3.40 2.15 0.40 9.53 10.59 9.42 0.00 0.00 1.16 0.00 0.00 0.00 70.00 0.00 0.90
15 3.26 3.58 0.26 10.96 10.28 6.81 0.00 0.00 3.47 0.00 0.00 0.00 70.00 0.00 1.46
16 3.23 1.15 0.26 11.24 10.28 4.41 0.00 0.00 5.87 0.00 0.00 0.00 70.00 0.00 0.49
17 3.81 1.11 0.44 0.00 10.43 2.17 0.00 6.28 1.98 0.00 0.00 0.00 63.67 0.24 0.00
18 5.29 1.22 0.25 0.00 11.45 0.97 0.00 10.47 0.00 0.00 0.00 0.00 53.10 0.55 0.00
19 4.54 1.46 0.27 0.00 7.77 0.00 0.00 7.77 0.00 0.00 0.00 0.00 45.26 0.35 0.00
20 4.44 2.47 0.30 0.00 7.46 0.00 0.00 7.46 0.00 0.00 0.00 0.00 37.74 0.33 0.00
21 4.30 1.90 0.19 0.00 7.67 0.00 0.00 7.67 0.00 0.00 0.00 0.00 30.00 0.33 0.00
22 3.58 1.83 0.45 8.54 8.23 0.00 0.00 0.00 8.23 0.00 0.00 0.00 30.00 0.00 0.77
23 2.40 0.50 0.18 14.76 8.07 0.00 0.00 0.00 8.07 0.00 0.00 0.00 30.00 0.00 0.91
24 2.23 0.32 0.43 9.59 8.07 0.00 0.00 0.00 8.07 0.00 0.00 0.00 30.00 0.00 0.74
TOTAL 199.26 63.49 42.95 39.65 96.28 0.14 0.00 0.15 0.83 10.91
Table 5-1: Example energy flows for one-day simulation, using li-ion battery (Orange=battery charge; Purple=battery discharge)
represents the hourly absolute sum of regulation- up and down battery usage, and is given by ( )
4
64
Figure 5-1: Example simulated battery energy flows for one-day period, using lithium-ion
Figure 5-1 above depicts the energy purchases and resultant battery cycling operation for the
output in Table 5-1 above. As shown, the optimal off-peak charging hour is between 2am and
3am, when electricity price is at its lowest point during the day. On this particular day, the EMS
meets 99.6% of its peak-shaving allocation from utility purchases and the small remainder from
excess PV. This predominance of off-peak utility purchases over excess PV charging is
representative of results from most other days throughout the year. This is due to the size of the
PV system relative to the building loads, as well as efficiency losses associated with battery
charging.
65
On the other hand, the battery is discharged to the loads during periods of peak price. From
Table 5-1, the highest cost period of this day is between 5pm and 6pm. However, since the
concurrent building loads are not sufficient to fully utilize the stored energy, the rest of this peak-
shaving capacity is successively distributed between the next highest cost periods, until the
allocation has been completely exhausted. Note that the amount discharged (39.36 kWh) is less
than the available capacity from off-peak and PV charging (43.10 kWh), again as a result of
The battery SOC begins and ends at 30%, which reflects the sum of the minimum allowable
SOC (10%) and 20% for frequency regulation services (with half of the capacity left open for
“regulation-up”). As discussed in Chapter 3, this daily cycling behavior is due to the control
strategies involving complete utilization of peak-shaving allocation and hourly net-zero from
frequency regulation participation. The battery SOC peaks out at 70%, which reflects the sum of
the 10% minimum SOC, 20% “regulation-down” allowance and 40% peak-shaving allocation.
This battery SOC operating range is repeated each day throughout the year, since the relative
The revenues from peak-shaving and frequency regulation have been calculated for this day, as
discussed in section 3.4.2. As shown, the net peak-shaving revenue for the given capacity
allocation of the lithium-ion battery under study significantly less than regulation revenue, since
it accounts for 7.6% of total system revenue on this day (neglecting avoided costs from PV
charging). This is proportion is typical of most other days, as will be discussed later in this
chapter.
66
Overall, in line with the microgrid operational objective of minimizing building energy demand
purchases, the combination of peak-shaving and local PV utilization has led to a 28.3% reduction
Figure 5-2(a)-(c) below show the simulation results for three other days, with each set
representing the other three seasons. The first point of note is that the demand on each of these
days is significantly higher than in the previous example. This means that while the optimal off-
peak charging periods are relatively similar, higher hourly demands in the on-peak discharging
periods utilize more of the stored capacity. Therefore, discharging occurs over a shorter period,
leaving more time for the system to bid into the regulation market. For this reason, the regulation
(a)
67
(b)
(c)
Figure 5-2: Simulation results for days representing Spring, Summer and Fall seasons (li-ion)
68
Apart from this, the major similarities between all the days are the uniform cycling range
The same battery allocation ratios and system configurations have been simulated using
specifications for the lead-acid battery system, as shown in Table 5-2 below. Also, Figure 5-3(a)-
(c) below shows the simulation results for this battery system for the same days discussed above
in Figure 5-2. From the discussion in section 3.3.4, the lead-acid system chosen for this analysis
requires about 3.9 times the lithium-ion energy storage capacity to produce roughly the same
power output. This means that for the same model inputs, the lead-acid battery buys more off-
peak energy and dissipates it over a longer period. This greatly reduces the proportion of the day
in which the battery can participate in frequency regulation, leading to diminished regulation
revenues in three of the four days under study. However, the peak-shaving revenue increases on
days of higher demand in June and September, when the demand is high enough to justify the
battery capacity of the lead-acid battery. Nevertheless, in the no-backup case considered in this
initial analysis, the use of a lithium-ion system generally leads to higher daily system revenue
than a lead-acid system with approximately equal power discharge capabilities (neglecting
Note also that in the lead-acid case, the battery cycles over a smaller range (42.5-67.5%), which
reflects the earlier discussed low depth of discharge associated with similar systems.
69
(c/kWh) (c/kWh) (c/kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (kWh) (%) ($) ($)
1 2.40 0.50 0.18 7.00 5.43 0.00 0.00 0.00 5.43 0.00 0.00 0.00 42.50 0.00 0.14
2 2.23 0.32 0.43 6.19 5.43 0.00 0.00 0.00 5.43 0.00 0.00 0.00 42.50 0.00 0.11
3 2.11 0.31 0.11 0.00 5.43 0.00 105.33 0.00 137.09 0.00 0.00 0.00 67.47 -2.78 0.00
4 2.12 1.21 0.11 5.64 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 67.47 0.00 0.31
5 2.27 1.79 0.09 5.88 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 67.47 0.00 0.45
6 2.65 0.08 0.05 7.20 5.42 0.00 0.00 0.00 5.42 0.00 0.00 0.00 67.47 0.00 0.02
7 3.57 0.61 0.37 0.00 5.02 0.00 0.00 5.02 0.00 0.00 0.00 0.00 65.98 0.18 0.00
8 3.31 2.11 0.22 0.00 8.23 1.28 0.00 6.95 0.00 0.00 0.00 0.00 63.92 0.23 0.00
9 3.36 1.65 0.31 0.00 11.18 2.80 0.00 8.39 0.00 0.00 0.00 0.00 61.44 0.28 0.00
10 3.45 1.86 0.30 0.00 10.42 6.18 0.00 4.24 0.00 0.00 0.00 0.00 60.18 0.15 0.00
11 3.58 1.41 0.27 0.00 10.26 8.76 0.00 1.50 0.00 0.00 0.00 0.00 59.74 0.05 0.00
12 3.53 1.91 0.44 0.00 10.42 9.65 0.00 0.77 0.00 0.00 0.00 0.00 59.51 0.03 0.00
13 3.48 2.25 0.36 0.00 10.89 11.05 0.00 0.00 0.00 0.14 0.00 0.15 59.54 0.00 0.00
14 3.40 2.15 0.40 0.00 10.59 9.42 0.00 1.16 0.00 0.00 0.00 0.00 59.19 0.04 0.00
15 3.26 3.58 0.26 0.00 10.28 6.81 0.00 3.47 0.00 0.00 0.00 0.00 58.16 0.11 0.00
16 3.23 1.15 0.26 0.00 10.28 4.41 0.00 2.99 2.87 0.00 0.00 0.00 57.28 0.10 0.00
17 3.81 1.11 0.44 0.00 10.43 2.17 0.00 8.26 0.00 0.00 0.00 0.00 54.83 0.31 0.00
18 5.29 1.22 0.25 0.00 11.45 0.97 0.00 10.47 0.00 0.00 0.00 0.00 51.72 0.55 0.00
19 4.54 1.46 0.27 0.00 7.77 0.00 0.00 7.77 0.00 0.00 0.00 0.00 49.42 0.35 0.00
20 4.44 2.47 0.30 0.00 7.46 0.00 0.00 7.46 0.00 0.00 0.00 0.00 47.21 0.33 0.00
21 4.30 1.90 0.19 0.00 7.67 0.00 0.00 7.67 0.00 0.00 0.00 0.00 44.94 0.33 0.00
22 3.58 1.83 0.45 0.00 8.23 0.00 0.00 8.23 0.00 0.00 0.00 0.00 42.50 0.29 0.00
23 3.01 0.50 0.18 9.25 8.07 0.00 0.00 0.00 8.07 0.00 0.00 0.00 42.50 0.00 0.57
24 2.85 0.32 0.43 6.01 8.07 0.00 0.00 0.00 8.07 0.00 0.00 0.00 42.50 0.00 0.46
TOTAL 199.26 63.49 105.33 84.36 183.23 0.14 0.00 0.15 0.57 2.06
Table 5-2: Example simulation results using lead-acid battery (Orange=battery charge; Purple=battery discharge)
70
(a)
(b)
71
(c)
(d)
Figure 5-3: EMS simulation results for select days (lead-acid)
72
5.2 Sensitivity of yearly revenues to capacity allocation between peak-shaving and
regulation
This section presents the results of EMS simulation for aggregate revenues obtained utilization
of the 7R microgrid in peak-shaving and frequency regulation operational modes, for the
duration of one year. This has been achieved by again assuming zero allocation towards backup
The sensitivity of the total system revenue to the relative allocation ratio between peak-shaving
and frequency regulation has been studied for one year of operation. This has been achieved by
varying the ratio of usable battery capacity between these two operational modes and calculating
resulting individual and total revenues, using MATLAB (see Appendix E). These results are
displayed and compared between lithium-ion and lead-acid battery systems in Table 5-3 and
Lithium-ion Lead-acid
PS Rev FR Rev Total Rev PS Rev FR Rev Total Rev
PS% FR %
($) ($) ($) ($) ($) ($)
100 0 514.02 0.00 514.02 524.81 0.00 524.81
90 10 468.09 1496.80 1964.89 493.74 848.88 1342.62
80 20 421.18 3022.50 3443.68 462.63 1724.40 2187.03
70 30 373.12 4575.50 4948.62 421.28 2663.40 3084.68
60 40 324.35 6163.70 6488.05 375.32 3625.70 4001.02
50 50 273.92 7828.40 8102.32 324.65 4620.30 4944.95
40 60 222.22 9448.20 9670.42 270.67 5661.90 5932.57
30 70 168.79 11178.00 11346.79 212.19 6727.50 6939.69
20 80 113.76 12919.00 13032.76 148.20 7895.70 8043.90
10 90 57.34 14668.00 14725.34 77.19 9109.90 9187.09
8 92 45.88 14996.00 15041.88 62.26 9331.60 9393.86
6 94 34.41 15324.00 15358.41 47.00 9600.70 9647.70
4 96 22.94 15643.00 15665.94 31.34 9820.50 9851.84
2 98 11.48 15978.00 15989.48 15.68 10028.00 10043.68
0 100 0.00 17560.00 17560.00 0.00 11013.00 11013.00
Table 5-3: Total revenues for various ratios of peak-shaving (PS) and frequency regulation (FR)
73
(a)
(b)
Figure 5-4: Sensitivity of revenues to battery capacity allocation by battery technology
74
The above results both show that peak-shaving operation contributes relatively little to total
yearly system revenues obtained by both types of battery systems. The lead-acid system displays
slightly higher peak-shaving revenues than the lithium-ion system, due to the former’s higher
energy capacity for comparable power output. However, while the use of the lead-acid system
regulation revenue, and consequently overall system revenues. Table 5-4 and Figure 5-5 below
display the percent changes in revenue resulting from the use of the lead-acid over the lithium-
ion system. As shown, the lead-acid system provides more peak-saving revenue for each
allocation ratio, although with negative percent-changes to regulation and total system revenues
Table 5-4: Percent change in system revenues resulting from utilization of lead-acid over
lithium-ion system
75
Figure 5-5: Comparison between lead-acid and lithium-ion revenues
This section presents the simulated revenues obtained by the 7R microgrid system based on the
proportion of battery capacity allocated toward emergency backup reserve, in addition to peak-
shaving and frequency regulation operation. Both of the previous analyses in this chapter have
held backup reserve capacity allocation to zero, thus representing the highest market revenues 5
These results have been obtained by carrying out a three-variable sensitivity analysis, which has
been achieved by successively increasing battery backup allocation and determining revenues
from partitioning the remaining battery capacity in varying ratios between peak-shaving and
5
These market revenues are assumed to neglect avoided costs from possible blackout events, as a function of
backup reserve allocation.
76
frequency regulation, using MATLAB (see code in Appendix E). This has been evaluated for
Figure 5-6(a)-(b) and Figure 5-7(a)-(b) display the results of this sensitivity analysis for each
battery system. In 5-6(a) and 5-7(a), the results have been presented as contour maps of revenues
of yearly system revenues, with each line representing relative allocation ratios that yield the
same revenues between peak-shaving, frequency regulation and backup reserve. The horizontal
axes show the proportion of battery capacity dedicated to backup reserve, while the vertical axis
80
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
6
The proportion of battery capacity used for peak-shaving operation is therefore the 100% minus each value on
the vertical axis
77
4
x 10
2
Total Market Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
80
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
78
4
x 10
2
Total Market Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
(b) Market revenues distribution based on battery capacity allocation
Figure 5-7:Market revenues from allocating lead-acid system capacity in various ratios
For example, based on Figure 5-6(a), allocating 20% of the lithium-ion system’s capacity toward
backup reserve and sharing the remainder in a 40-60% ratio between frequency regulation and
peak-shaving would yield approximately the same revenue as 60% backup, 80% regulation and
20% peak-shaving.
characterizations. Notice that the outer edge of each chart (zero-backup scenario) resembles the
Overall, the curves shown in Figure 5-6 and Figure 5-7 are similar in shape and display
the lithium-ion system is able to obtain higher market revenues for the same ratios of lead-acid
allocation.
79
5.4 Sensitivity of yearly revenues to overall capacity allocation and blackout costs
The results discussed so far in this chapter only account for market revenues in the assessing the
monetary performance of the system under various conditions. In this section, the above revenue
characterization has been expanded to include avoided costs from potential blackouts and power
reliability events. This has been achieved by adding above derived market revenues to the values
of avoided costs for low, average and high-demand days, as determined in section 4.6. The
impact of these avoided costs on overall system revenues are displayed in Figure 5-8 to Figure 5-
12 below.
80
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
(a) Lithium-ion iso-revenue characterization for different battery capacity allocation ratios
80
4
x 10
2
Total System Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
81
Lithium-ion Total Revenues (Average/high demand cases)
100
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
4
x 10
2
Total System Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
82
Lead-acid Total Revenues (Low demand case)
100
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
4
x 10
Total System Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
(b) Total revenues distribution based on battery capacity allocation
Figure 5-10: Total system revenues on a low-demand day (lead-acid system)
83
Lead-acid Total Revenues (Average demand case)
100
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
4
x 10
Total System Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
(b) Total revenues distribution based on battery capacity allocation
Figure 5-11: Total system revenues on an average-demand day (lead-acid system)
84
Lead-acid Total Revenues (High demand case)
100
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
Battery Backup (%)
4
x 10
Total System Revenue ($)
1.5
0.5
0
100
80 100
60 80
40 60
40
20 20
0 0
Rel. Regulation(%) Battery Backup (%)
.
(b) Total revenues distribution based on battery capacity allocation
Figure 5-12: Total system revenues on a high-demand day (lead-acid system)
85
In the case of the lithium-ion system, the inclusion of avoided blackout costs has limited effects
on overall system revenues on low-demand days, as shown in Figure 5-7. On average to high-
demand days7, while system revenues are relatively higher, the revenue distribution in Figure 5-8
still shows increasing system revenues with less backup reserve and more frequency regulation
allocation.
On the other hand, due to much greater energy capacity for the same level of power output as the
lithium-ion system as well as lower frequency regulation revenues, the lead-acid system is more
significantly impacted by the inclusion of avoided blackout costs in the revenue calculation,
especially during average and high-demand days. As shown in Figure 5-10, in the event of a
power reliability event on a day of average demand during the year, total annual system revenues
increase as both regulation and backup allocation increase, until a backup allocation of about 60-
70%. In the high-demand day shown in Figure 5-11, total annual system revenues continue to
increase past 60% in the directions of increasing backup allocation, although the highest
revenues still occur when the battery is completely dedicated toward frequency regulation.
Notice that in all lithium-ion and lead-acid cases, 100% peak-shaving allocation leads to the least
amount of revenues. This is because representing backup revenue as avoided blackout costs
means that 100% peak-shaving yields less revenues than 100% backup battery allocation. This is
based on the assumption that the kWh battery backup allocation has a linear relationship to
avoided blackout costs, i.e. each additional unit of allocated backup capacity yields the same
7
Note that the total usable lithium-ion battery capacity is less than the potential energy losses calculated for both
average and high-demand days in section 3.3.5. Avoided cost for the lithium ion system is the same during average
and low demand days.
86
Chapter 6: Conclusions
The purpose of this research was to characterize the technological and economic tradeoffs
associated with the operation of a PV-battery commercial building microgrid system, based on
system specifications, electricity demand information, weather data, as well as energy and
ancillary market prices faced by the 7R Building microgrid at the Philadelphia Navy Yard. This
has been achieved by defining the control strategy required for the microgrid EMS to meet the
set objective of maximizing net revenues from building energy demand purchases and ancillary
market participation, while storing emergency backup capacity. Thus, the system has been
constrained to only provide peak-shaving, frequency regulation and backup reserve functionality.
These operational modes have been consolidated by means of the battery capacity allocation
approach, in which the usable battery capacity is partitioned between the modes in predefined
ratios. On this basis, the overall system revenues have been simulated and compared between a
The results of thesis research have yielded information on the degree to which battery capacity
allocation between peak-shaving, frequency regulation and emergency backup impact overall
revenues accrued by the microgrid system over the course of one year. This is has been
determined with reference to the influence of technology selection between lithium-ion and lead-
acid systems.
87
Firstly, for a fixed ratio of battery capacity allocation considered on a daily basis, peak-shaving
revenues have been found to be significantly less than frequency regulation revenues for the
demand profile and market prices faced by the microgrid system under study. Also lithium-ion
yields significantly higher daily system revenues than the lead-acid system with comparable
continuous discharge power capabilities. Extending to one year of operation, the lithium-ion
system has been determined to yield more market revenues than the lead-acid system, for the
same fixed allocation ratio. While lead-acid provides more peak-shaving revenue and demand
purchase reduction than lithium-ion due to larger energy capacity, this increase is accompanied
by a greater decrease in frequency regulation revenues, and thus overall market revenue.
Secondly, the revenues obtained by the building microgrid have been determined for varying
capacity allocation ratios between peak-shaving, frequency regulation and emergency backup.
Without considering avoided costs from potential blackout mitigation, the highest revenues
achievable by the system occur in the case of 100% allocation towards frequency regulation for
both lithium-ion and lead-acid systems, with revenues dropping steeply in all directions away
from this condition. The introduction of avoided blackout cost has little impact on this outcome
in the lithium-ion case, with the optimum allocation clearly pointing to 100% frequency
regulation in all scenarios of building demand magnitude. This suggests that under the proposed
operational framework, the potential energy and revenue losses from blackouts do not justify the
allocation of the lithium-ion system for backup and peak-shaving functionality. On the other
hand, the revenue map of the lead-acid system changes drastically with the inclusion of avoided
blackout costs. Specifically, results suggest that overall system revenues occurring on average to
high demand days increase both in the direction of increasing backup and frequency regulation
allocations (with a limit of 60-70% on an average day). However, the maximum overall revenues
88
still occur at maximum frequency regulation allocation in the lead-acid case. Therefore, under
the proposed framework, the objective of overall system revenue maximization means that
Based on the foundation of this research, opportunities exist for future research to foster
improved understanding of the economic and technological tradeoffs involved in the design and
load classes
Proof and analysis of proposed EMS model using the Building 7R microgrid in the
89
Appendix A: 7R eQuest Model Parameters and Assumptions
90
Opaque Building Envelope Constructions Baseline Case Proposed Case Baseline
Proposed
Space- Assembly Assembly Roof
Model Input New / Roof
Conditioning U-factor/ U-factor/ Reflectivity
Parameter Existing Description Description Reflectivity
Category C-factor/ C-factor/ Modeled as
Modeled
F-factor F-factor 0.3?
Green Roof assembly with Soil Layer
(6 – 8")
TPO roof membrane
Roof R-20 insulation entirely above
New Nonresidential U-0.048 Glas-mat gypsum coverboard U-0.025 Yes 0.30
Constructions deck with a U-factor of 0.048
Tapered polyisocyanurate Insulation
(R-30 min)
Steel Deck & roof Structure
Brick veneer
2" air space
steel-framed with R-13 cavity and 4" rigid XPS insulation board (R-20)
New Nonresidential R-7.5 continuous insulation with a U-0.064 Air barrier U-0.05
U-factor of 0.064 Sheathing board
Above-Grade Cold-formed metal framing
Exterior Wall Gypsum board finish
Constructions Metal panel cladding on metal cont. Z-
furring
steel-framed with R-13 cavity and
4" XPS rigid insulation board (R-20)
New Nonresidential R-7.5 continuous insulation with a U-0.064 U-0.05
Sheathing board
U-factor of 0.064
Cold-formed metal framing
Gypsum board finish
6" Concrete slab (Unheated)
ASHRAE 90.1-2007 Appendix G
Slab-On-Grade 2" perimeter rigid XPS insulation
New Nonresidential Slab on grade construction F-0.73 F-0.58
Floors board under slabs extending 4 ft.
Unheated slab
horizontally in from perimeter.
N/A
Orientation
(%) (ft2)
North 5,884
740
South
11,307
3,076
91
The table below lists maximum occupancy, the average peak connected power density (Watts/ft²) of the ambient
lighting (including task lights), decorative lighting, and the average peak equipment load (Watts/ft²) for all zones
throughout the building.
Total 25,050 1.76 Total Power Modeled Using Space-by-Space Method (kW): 44.2
92
The modeling assumptions for the maximum occupancy and area per person in the building were estimated based on
the email communication with the design team in Construction Administration (CA) Phase. Peak lighting and equipment
loads were diversified by estimating an electric usage pattern for the building based on scheduling assumptions. The
tables below list occupancy, lighting and equipment schedules assumed for the individual usage type for the building.
The owner and the design team should review the schedules and provide feedback based on their expected use of the
building.
Annual Schedule:
Two semester schedule with Thanksgiving and winter breaks.
Unoccupied weeks:
Thanksgiving Recess: November 17 thru November 25
Winter Recess: December 22 thru January 8
93
Hour Occupancy Lighting Misc. Equipment
From To Weekday Sunday Weekday Sunday Weekday Sunday
12:00 AM 1:00 AM 0% 0% 2% 2% 6% 6%
1:00 AM 2:00 AM 0% 0% 2% 2% 6% 6%
2:00 AM 3:00 AM 0% 0% 2% 2% 6% 6%
3:00 AM 4:00 AM 0% 0% 2% 2% 6% 6%
4:00 AM 5:00 AM 0% 0% 2% 2% 6% 6%
5:00 AM 6:00 AM 0% 0% 2% 2% 6% 6%
6:00 AM 7:00 AM 0% 0% 2% 2% 6% 6%
7:00 AM 8:00 AM 25% 0% 25% 2% 25% 6%
8:00 AM 9:00 AM 90% 0% 90% 2% 45% 6%
9:00 AM 10:00 AM 88% 0% 88% 2% 45% 6%
10:00 AM 11:00 AM 75% 0% 75% 2% 45% 6%
11:00 AM 12:00 PM 54% 0% 54% 2% 45% 6%
12:00 PM 1:00 PM 54% 0% 54% 2% 45% 6%
1:00 PM 2:00 PM 75% 0% 75% 2% 45% 6%
2:00 PM 3:00 PM 88% 0% 88% 2% 45% 6%
3:00 PM 4:00 PM 90% 0% 90% 2% 45% 6%
4:00 PM 5:00 PM 89% 0% 89% 2% 45% 6%
5:00 PM 6:00 PM 5% 0% 5% 2% 6% 6%
6:00 PM 7:00 PM 0% 0% 2% 2% 6% 6%
7:00 PM 8:00 PM 0% 0% 2% 2% 6% 6%
8:00 PM 9:00 PM 0% 0% 2% 2% 6% 6%
9:00 PM 10:00 PM 0% 0% 2% 2% 6% 6%
10:00 PM 11:00 PM 0% 0% 2% 2% 6% 6%
11:00 PM 12:00 AM 0% 0% 2% 2% 6% 6%
From To Weekday Weekday Weekend Weekday Weekday Weeken Weekday Weekday Weekend
MWF TR MWF TR d MWF TR
12:00 AM 1:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
1:00 AM 2:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
2:00 AM 3:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
3:00 AM 4:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
4:00 AM 5:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
5:00 AM 6:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
6:00 AM 7:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
7:00 AM 8:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
8:00 AM 9:00 AM 90% 0% 0% 90% 2% 2% 50% 6% 6%
9:00 AM 10:00 AM 90% 0% 0% 90% 2% 2% 50% 6% 6%
10:00 AM 11:00 AM 90% 0% 0% 90% 2% 2% 50% 6% 6%
11:00 AM 12:00 PM 90% 0% 0% 90% 2% 2% 50% 6% 6%
12:00 PM 1:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
1:00 PM 2:00 PM 0% 90% 0% 2% 90% 2% 6% 50% 6%
2:00 PM 3:00 PM 0% 90% 0% 2% 90% 2% 6% 50% 6%
3:00 PM 4:00 PM 0% 90% 0% 2% 90% 2% 6% 50% 6%
4:00 PM 5:00 PM 0% 90% 0% 2% 90% 2% 6% 50% 6%
5:00 PM 6:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
6:00 PM 7:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
7:00 PM 8:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
8:00 PM 9:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
9:00 PM 10:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
10:00 PM 11:00 PM 0% 0% 0% 2% 2% 2% 6% 6% 6%
11:00 PM 12:00 AM 0% 0% 0% 2% 2% 2% 6% 6% 6%
94
Basic Zone Parameters
Minimum Flow Ratio: 30% (Minimum allowable supply air flow rate for VAV system)
Minimum Design Flow: 0.4 cfm/ft²
Design Conditions
Space (AHU) Winter temp °F Summer temp °F High RH% Low RH%
Occupied: 70 Occupied: 70
General Areas Unoccupied Wk day: 60 Unoccupied Wk day: 60 55% n/a
Holiday: 50 Holiday: 50
Occupied: 67 Occupied: 78
Corridor / Stairs Unoccupied Wk Day: 55 Unoccupied Wk Day: 80 55% n/a
Holiday: 55 Holiday: 80
Primary HVAC System System 6 - Packaged VAV with PFP Boxes All spaces except server room
Other HVAC System(s) System 4 - PSZ-HP Server room Server room
System 10 - Heating and Ventilation Heating only spaces Mechanical rooms
95
Model Input Parameter Baseline Case Units Proposed Case Units
HHW HT °F 10 °F
96
Service Water Heaters
Model Input Parameter Baseline Case Proposed Case
System Type & Fuel Electric hot water boiler Electric hot water boiler
Input Rating (kW, MBH, etc.) 10 MBH 10 MBH
Efficiency (EF, SL, %, etc.) 1.02 EIR 1.02 EIR
Storage Volume (gal) 20 20
Storage Temperature (°F) 120 120
Peak Hot Water Demand (gpm) 0.75 gpm 0.75 gpm
Condenser heat recovery N/A N/A
*Hot water perimeter heating provided by electric heat pumps is modeled as electrical perimeter heat due to eQUEST’s inability to
couple WSHPs to baseboard radiators.
Electricity (PECO)
97
Appendix B: Lithium-ion battery specifications
1. Product Information
SLPB Cell
Model Unit SLPB90255255H
Capacity Ah 52
Voltage Range V .0 ~ 4.2
98
G
SLPB
G 52Ah HIGH POWER
SUPERIOR
LITHIUM POLYMER BATTERY
Electrical Characteristics
Items Specification Remarks
Advantages of SLPB Rated Capacity 52Ah
Charge@0.2C, 23±3ଇ
Discharge@0.2C, 23±3ଇ
%The purpose of this program is to find the optimal array size: row*col of
%a lead acid battery bank capable of putting out 100kW, based on the
%Enersys PowerSafe SB190 battery specs:
%Max C/4 power per cell, for discharge to 1.75V= 90.4W
%Number of cells/module=6
%Voltage per battery pack=12V
%The program makes use of a genetic mixed-integer optimization, as follows:
% min x1*x2 where x1=row, x2=col
% s.t. 100k <= x1*x2*90.4*6 <= 100.5k
% 280 <= x2*12 <= 600 to ensure the voltage is in reasonable
% dc input range of typical inverter
lb=[2,280/12];
ub=[10,600/12];
[x,fval]=ga(@objfcn,2,[],[],[],[],lb,ub,@confun,[1 2]);
% @objfun
%This program defines the objective function for the lead-acid ga
%optimization.
%Note: This should be saved under a separate m-file and run through the
%main optimization program
function f= objfcn(x)
f= x(1)*x(2);
end
% @confun
%This program defines the conditional function for the lead-acid ga
%optimization
%Note: This should be saved under a separate m-file and run through the
%main optimization program
100
Appendix D: TRYNSYS PV generation model
TRNSYS Simulation Studio Screenshot
Parameters
Sky model (Perez model) Citation: Perez, R., Stewart, R., Seals, R., Guertin, T., "The Development
and Verification of The Perez Diffuse Radiation Model", Sandia Report SAND88-7030, (Sandia
National Laboratories, Albuquerque, New Mexico, 87185, USA) October, 1988.
Tracking Mode (Fixed surface)
Inputs
101
Outputs
102
Type 94a – PV Component
Parameters
103
Outputs
104
Appendix E: MATLAB code for peak-shaving optimization and
EMS model
%battoptimize.m
%This program runs the monthly baseline peak-shaving optimization.
%It receives the following inputs:
%res_ratio--> Battery capacity reserve allocation
%reg_ratio--> Battery capacity frequency regulation allocation
%ps_ratio--> Battery capacity peak-shaving allocation
%month_name--> String referring to the name of the month
%total_days--> Total number of days in the month
%first_cell--> First cell corresponding to the month in the spreadsheet
%containing all input data
%results_file-->String name of file in which the results will be stored
%Note: This should be saved under a separate m-file and run through the
%main optimization program (i.e. test_run.m below)
function battoptimize(res_ratio,reg_ratio,ps_ratio,month_name,...
total_days,first_cell,results_file)
h=1:24';
p_utility=month_data((day-1)*24+1:(day-1)*24+24,2);
D=month_data((day-1)*24+1:(day-1)*24+24,3);
q_pv=month_data((day-1)*24+1:(day-1)*24+24,4);
p_rmc=month_data((day-1)*24+1:(day-1)*24+24,5);
p_rmp=month_data((day-1)*24+1:(day-1)*24+24,6);
day_type=month_data((day-1)*24+1:(day-1)*24+24,7);
dc=month_data((day-1)*24+1:(day-1)*24+24,8);
%Battery specifications
%Lithium-ion specs
105
% eff=0.92; %Battery round-trip efficiency @ 1C
% soc_max=0.8; %Max state-of-charge
% soc_min=0.1; %Min state-of-charge
% Q_batt=107.74; %Total useable battery capacity
% B=100; %Maximum battery power output
% r_dis=1*Q_batt; %C-rate of 1C is assumed, thus max charge rate
%is 100%/hour
for i=1:24
if D(i)-q_pv(i)>0
D_unmet(i,1)= D(i)-q_pv(i);
else
D_unmet(i,1)=0;
end
end
first=(day-1)*24+1;
last=(day-1)*24+24;
if day_type < 6 %During weekdays
%% Optimization model
106
%% Objective function
%% Equality constraints
%sum(q_psc+q_pvc)= Q_batt*PS
%sum(eff*q_psc-q_psd+eff*q_pvc)=0
Aeq(2,:)=[eff*ones(1,24) -1*ones(1,24) zeros(1,24) eff*ones(1,24)...
zeros(1,24)];
beq(2,1)=0;
for i=2:23
Aeq(i+2,:)=[zeros(1,24) zeros(1,i-1) 1 zeros(1,24-i) zeros(1,i-1)...
1 zeros(1,24-i) zeros(1,24) zeros(1,24)];
end
for i=1:24
beq(i+2,1)=D_unmet(i);
end
for i=2:23
Aeq(i+26,:)=[zeros(1,24) zeros(1,24) zeros(1,24) zeros(1,i-1)...
1/eff zeros(1,24-i) zeros(1,i-1) 1 zeros(1,24-i)];
end
for i=1:24
beq(i+26,1)=q_excess(i);
end
107
%% Inequality constraints
b=zeros(24,1);
%% Linprog
x=linprog(f,A,b,Aeq,beq,lb,ub);
%% Battery level
%In kWh
for i=[2:24]
q_batt(i,1)=q_batt(i-1)+x(i)+ x(i+72)-x(i+24)/eff;
end
for i=[2:24]
soc(i,1)=soc(i-1)+(100/Q_batt)*(+x(i)+ x(i+72)-x(i+24)/eff);
end
%% Catalog results
results(first:last,1)=h;
results(first:last,2)=p_utility;
results(first:last,3)=D;
results(first:last,4)=q_pv;
108
for i=1:5
for j=1:24
x_matrix(j,i)=x(j+(i-1)*24,1);
end
end
results(first:last,5)=x_matrix(:,1);
results(first:last,6)=x_matrix(:,2);
results(first:last,7)=x_matrix(:,3);
results(first:last,8)=x_matrix(:,4);
results(first:last,9)=x_matrix(:,5);
results(first:last,10)=q_excess;
results(first:last,11)=q_batt;
results(first:last,12)=soc;
for i=1:24
if results(first+i-1,5)<1e-5 && results(first+i-1,6)<1e-5...
&& results(first+i-1,8)<1e-5
q_reg(i,:)= dc(i)*FR*B;
reg_rev(i,:)= FR*B*p_rmc(i)+ dc(i)*FR*B*p_rmp(i);
else
q_reg(i,:)=0;
reg_rev(i,:)=0;
end
end
results(first:last,13)=p_rmc;
results(first:last,14)=p_rmp;
results(first:last,15)=q_reg;
%Peak-shaving revenue
for i=1:24
ps_rev(i,:)=x(24+i)*p_utility(i)+x(72+i)*p_utility(i)...
-x(i)*p_utility(i)*(1/eff);
end
results(first:last,16)=ps_rev;
results(first:last,17)=reg_rev;
%Warning flag
for i=1:24
if results(i,5)>=0.00001 && results(i,6)>=0.00001 ||...
results(i,6)>=0.00001 && results(i,8)>=0.00001
day_text=sprintf('%d',day);
disp(['WARNING!!! MUTUTAL CHARGE/DISCHARGE-', ...
month_name, day_text]);
end
end
109
results(first:last,1)=h;
results(first:last,2)=p_utility;
results(first:last,3)=D;
results(first:last,4)=q_pv;
results(first:last,5)=zeros(24,1);
results(first:last,6)=zeros(24,1);
results(first:last,7)=D_unmet;
results(first:last,8)=zeros(24,1);
results(first:last,9)=q_excess;
results(first:last,10)=q_excess;
results(first:last,11)=(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
results(first:last,12)=(100/Q_batt)*(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
for i=1:24
q_reg(i,:)= dc(i)*FR*B;
reg_rev(i,:)= FR*B*p_rmc(i)+ dc(i)*FR*B*p_rmp(i);
end
results(first:last,13)=p_rmc;
results(first:last,14)=p_rmp;
results(first:last,15)=q_reg;
%Peak-shaving revenue
for i=1:24
ps_rev(i,:)=0;
end
results(first:last,16)=ps_rev;
results(first:last,17)=reg_rev;
end
end
csvwrite(results_file,results,0,0);
end
110
%battoptimize_ps100.m
%This program runs the monthly peak-shaving optimization with special
%instructions for the case of 100% peak-shaving allocation
%It receives the following inputs:
%res_ratio--> Battery capacity reserve allocation
%reg_ratio--> Battery capacity frequency regulation allocation (i.e. 0)
%ps_ratio--> Battery capacity peak-shaving allocation
%month_name--> String referring to the name of the month
%total_days--> Total number of days in the month
%first_cell--> First cell corresponding to the month in the spreadsheet
%containing all input data
%results_file-->String name of file in which the results will be stored
%Note: This should be saved under a separate m-file and run through the
%main optimization program (i.e. test_run.m below)
function battoptimize_ps100(res_ratio,reg_ratio,ps_ratio,month_name,...
total_days,first_cell,results_file)
h=1:24';
p_utility=month_data((day-1)*24+1:(day-1)*24+24,2);
D=month_data((day-1)*24+1:(day-1)*24+24,3);
q_pv=month_data((day-1)*24+1:(day-1)*24+24,4);
p_rmc=month_data((day-1)*24+1:(day-1)*24+24,5);
p_rmp=month_data((day-1)*24+1:(day-1)*24+24,6);
day_type=month_data((day-1)*24+1:(day-1)*24+24,7);
dc=month_data((day-1)*24+1:(day-1)*24+24,8);
%Battery specifications
%Lithium-ion specs
111
% soc_min=0.1; %Min state-of-charge
% Q_batt=107.74; %Total useable battery capacity
% B=100; %Maximum battery power output
% r_dis=1*Q_batt; %C-rate of 1C is assumed, thus max charge rate
%is 100%/hour
for i=1:24
if D(i)-q_pv(i)>0
D_unmet(i,1)= D(i)-q_pv(i);
else
D_unmet(i,1)=0;
end
end
first=(day-1)*24+1;
last=(day-1)*24+24;
if day_type < 6 %During weekdays
%% Optimization model
%% Objective function
112
f= [p_utility(1:24); -1*p_utility(1:24); p_utility(1:24);
-1*p_utility(1:24); -1*p_utility(1:24)];
%% Equality constraints
%sum(q_psc+q_pvc)= Q_batt*PS
%sum(eff*q_psc-q_psd+eff*q_pvc)=0
Aeq(2,:)=[eff*ones(1,24) -1*ones(1,24) zeros(1,24) eff*ones(1,24)...
zeros(1,24)];
beq(2,1)=0;
for i=2:23
Aeq(i+2,:)=[zeros(1,24) zeros(1,i-1) 1 zeros(1,24-i) zeros(1,i-1)...
1 zeros(1,24-i) zeros(1,24) zeros(1,24)];
end
for i=1:24
beq(i+2,1)=D_unmet(i);
end
for i=2:23
Aeq(i+26,:)=[zeros(1,24) zeros(1,24) zeros(1,24) zeros(1,i-1)...
1/eff zeros(1,24-i) zeros(1,i-1) 1 zeros(1,24-i)];
end
for i=1:24
beq(i+26,1)=q_excess(i);
end
%% Inequality constraints
113
zeros(1,24)];
A(24,:)=[-1*ones(1,24) ones(1,24) zeros(1,24) -1*ones(1,24)...
zeros(1,24)];
b=zeros(24,1);
%% Linprog
x=linprog(f,A,b,Aeq,beq,lb,ub);
%% Battery level
%In kWh
for i=[2:24]
q_batt(i,1)=q_batt(i-1)+x(i)+ x(i+72)-x(i+24)/eff;
end
for i=[2:24]
soc(i,1)=soc(i-1)+(100/Q_batt)*(+x(i)+ x(i+72)-x(i+24)/eff);
end
%% Catalog results
results(first:last,1)=h;
results(first:last,2)=p_utility;
results(first:last,3)=D;
results(first:last,4)=q_pv;
for i=1:5
for j=1:24
x_matrix(j,i)=x(j+(i-1)*24,1);
114
end
end
results(first:last,5)=x_matrix(:,1);
results(first:last,6)=x_matrix(:,2);
results(first:last,7)=x_matrix(:,3);
results(first:last,8)=x_matrix(:,4);
results(first:last,9)=x_matrix(:,5);
results(first:last,10)=q_excess;
results(first:last,11)=q_batt;
results(first:last,12)=soc;
for i=1:24
q_reg(i,:)=0;
reg_rev(i,:)=0;
end
results(first:last,13)=p_rmc;
results(first:last,14)=p_rmp;
results(first:last,15)=q_reg;
%Peak-shaving revenue
for i=1:24
ps_rev(i,:)=x(24+i)*p_utility(i)+x(72+i)*p_utility(i)...
-x(i)*p_utility(i)*(1/eff);
end
results(first:last,16)=ps_rev;
results(first:last,17)=reg_rev;
%Warning flag
for i=1:24
if results(i,5)>=0.00001 && results(i,6)>=0.00001 ||...
results(i,6)>=0.00001 && results(i,8)>=0.00001
day_text=sprintf('%d',day);
disp(['WARNING!!! MUTUTAL CHARGE/DISCHARGE-',...
month_name, day_text]);
end
end
results(first:last,1)=h;
results(first:last,2)=p_utility;
results(first:last,3)=D;
results(first:last,4)=q_pv;
results(first:last,5)=zeros(24,1);
results(first:last,6)=zeros(24,1);
results(first:last,7)=D_unmet;
results(first:last,8)=zeros(24,1);
115
results(first:last,9)=q_excess;
results(first:last,10)=q_excess;
results(first:last,11)=(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
results(first:last,12)=(100/Q_batt)*(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
for i=1:24
q_reg(i,:)= 0;
reg_rev(i,:)= 0;
end
results(first:last,13)=p_rmc;
results(first:last,14)=p_rmp;
results(first:last,15)=q_reg;
%Peak-shaving revenue
for i=1:24
ps_rev(i,:)=0;
end
results(first:last,16)=ps_rev;
results(first:last,17)=reg_rev;
end
end
csvwrite(results_file,results,0,0);
end
116
%battoptimize_fr100.m
%This program runs the monthly peak-shaving optimization with special
%instructions for the case of 100% frequency regulation allocation
%It receives the following inputs:
%res_ratio--> Battery capacity reserve allocation
%reg_ratio--> Battery capacity frequency regulation allocation
%ps_ratio--> Battery capacity peak-shaving allocation (i.e. 0)
%month_name--> String referring to the name of the month
%total_days--> Total number of days in the month
%first_cell--> First cell corresponding to the month in the spreadsheet
%containing all input data
%results_file-->String name of file in which the results will be stored
%Note: This should be saved under a separate m-file and run through the
%main optimization program (i.e. test_run.m below)
function battoptimize_fr100(res_ratio,reg_ratio,ps_ratio,month_name,...
total_days,first_cell,results_file)
h=1:24';
p_utility=month_data((day-1)*24+1:(day-1)*24+24,2);
D=month_data((day-1)*24+1:(day-1)*24+24,3);
q_pv=month_data((day-1)*24+1:(day-1)*24+24,4);
p_rmc=month_data((day-1)*24+1:(day-1)*24+24,5);
p_rmp=month_data((day-1)*24+1:(day-1)*24+24,6);
day_type=month_data((day-1)*24+1:(day-1)*24+24,7);
dc=month_data((day-1)*24+1:(day-1)*24+24,8);
%Battery specifications
%Lithium-ion specs
117
% soc_min=0.1; %Min state-of-charge
% Q_batt=107.74; %Total useable battery capacity
% B=100; %Maximum battery power output
% r_dis=1*Q_batt; %C-rate of 1C is assumed, thus max charge rate
%is 100%/hour
for i=1:24
if D(i)-q_pv(i)>0
D_unmet(i,1)= D(i)-q_pv(i);
else
D_unmet(i,1)=0;
end
end
first=(day-1)*24+1;
last=(day-1)*24+24;
results(first:last,1)=h;
results(first:last,2)=p_utility;
results(first:last,3)=D;
results(first:last,4)=q_pv;
results(first:last,5)=zeros(24,1);
results(first:last,6)=zeros(24,1);
results(first:last,7)=D_unmet;
results(first:last,8)=zeros(24,1);
results(first:last,9)=q_excess;
results(first:last,10)=q_excess;
results(first:last,11)=(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
results(first:last,12)=(100/Q_batt)*(soc_min+R+0.5*FR)*Q_batt*ones(24,1);
118
%Regulation quantity and revenue
for i=1:24
q_reg(i,:)= dc(i)*FR*B;
reg_rev(i,:)= FR*B*p_rmc(i)+ dc(i)*FR*B*p_rmp(i);
end
results(first:last,13)=p_rmc;
results(first:last,14)=p_rmp;
results(first:last,15)=q_reg;
%Peak-shaving revenue
for i=1:24
ps_rev(i,:)=0;
end
results(first:last,16)=ps_rev;
results(first:last,17)=reg_rev;
end
csvwrite(results_file,results,0,0);
end
119
Bibliography
[1] M. E. Kahn, N. Kok and J. M. Quigley, "Carbon emissions from the commercial building sector: The
role of climate, quality, and incentives," Jornal of Public Economics, vol. 113, pp. 1-12, 2014.
[3] "Use of Energy in the United States Explained: Energy Use in Commercial Buildings," U.S. Energy
Information Administration, 27 July 2011. [Online]. Available:
http://www.eia.gov/energyexplained/index.cfm?page=us_energy_commercial#tab1.
[4] U.S. Energy Information Administration, "Commercial Buildings Energy Consumption Survey," 2003.
[6] U.S. Envinronmental Protection Agency, "EPA's Greenhouse Gas Emission Reductions," 5 November
2012. [Online]. Available: http://www.epa.gov/oaintrnt/ghg/index.htm.
[7] IEEE Recommended Practice for Emergency and Standby Power Systems for Industrial and
Commercial Applications, IEEE Std 446-1995, 1995.
[8] K. H. LaCommare and J. H. Eto, "Understanding the Cost of Power Interruptions to U.S. Electricity
Consumers," Ernest Orlando Lawrence Berkeley National Laboratory, Berkeley, 2004.
[9] IEEE Recommended Practice for the Design of Reliable Industrial and Commercial Power Systems,
IEEE Std 493-2007, 2007.
[10] D. Salomonsson and L. Soder, "Comparison of Different Solutions for Emergency and Standby
Power Systems for Commercial Consumers," in Telecommunications Energy Conference, 2006.
INTELEC '06. 28th Annual International, 2006.
[11] S. D. Shah, D. R. Cocker III, K. C. Johnson, J. M. Lee, B. L. Soriano and J. W. Miller, "Emissions of
regulated pollutants from in-use diesel back-up generators," Atmospheric Environment, vol. 40, no.
22, pp. 4199-4209, 2006.
[12] Northeast States for Coordinated Air Use Management (NESCAUM), "Stationary Diesel Engines in
the Northeast: An Initial Assessment of the Regional Population, Control Technology Options and
Air Quality Policy Issues," 2003.
[13] J. W. Herzog, "Current and near-term emission control strategies for diesel powered generator
120
sets," in INTELEC. 24th Annual International, 2002.
[14] S. K. Siu and J. Lopopolo , "Compatibility, Sizing, and Design Considerations for Generators and UPSs
in Tiers I, II, III, and IV Topologies," Industry Applications, IEEE Transactions on, vol. 47, no. 6, pp.
2324-2329, 2011.
[15] Solar Energy Industries Association, "Solar Market Insight Report 2013 Year in Review," 2014.
[Online]. Available: http://www.seia.org/research-resources/solar-market-insight-report-2013-
year-review.
[16] R. Wills, J. Milke, S. Royle and K. Steranka, "Commercial Roof-Mounted Photovoltaic System
Installation Best Practices Review and All Hazard Assessment," The Fire Protection Research
Foundation, 2014.
[17] K. H. Solangi, M. R. Islam, N. A. Rahim and H. Fayaz, "A review on global solar energy policy,"
Renewable and Sustainable Energy Reviews, vol. 15, no. 4, p. Pages 2149–2163, 2011.
[18] S. J. Hayter and R. L. Martin, "Photovoltaics for Buildings – Cutting-Edge PV," National Renewable
Energy Laboratory, 1998.
[19] F. Cucchiella and I. D'Adamo, "Estimation of the energetic and environmental impacts of a roof-
mounted building-integrated photovoltaic systems," Renewable and Sustainable Energy Reviews,
vol. 16, no. 7, p. 5245–5259, September 2012.
[20] Office of Energy Efficiency & Renewable Energy, "SunShot Initiative: About," U.S. Department of
Energy, [Online]. Available: http://energy.gov/eere/sunshot/about.
[21] L. Sherwood, "U.S. Solar Market Trends 2012," Interstate Renewable Energy Council, Inc. , 2013.
[22] J. Driesen and F. Katiraei, "Design for distributed energy resources," Power and Energy Magazine,
IEEE, vol. 6, no. 3, pp. 30-40, May-June 2008.
[23] N. Hatziargyriou, H. Asano, R. Iravani and C. Marnay, "Microgrids," Power and Energy Magazine,
IEEE , vol. 5, no. 4, pp. 78-94, 2007.
[24] IEEE Guide for Design, Operation, and Integration of Distributed Resource Island Systems with
Electric Power Systems, IEEE Std 1547.4-2011, 2011.
[25] W. Su and J. Wang, "Energy Management Systems in Microgrid Operations," The Electricity Journal,
vol. 25, no. 8, pp. 45-60, 2012.
[26] M. Sechilariu, B. Wang and F. Locment, "Building-integrated microgrid: Advanced local energy
management for forthcoming smart power grid communication," Energy and Buildings, vol. 59, p.
121
236–243, 2013.
[27] S. Bhaskara and B. Chowdhury, "Microgrids—A review of modeling, control, protection, simulation
and future potential," in Power and Energy Society General Meeting, 2012 IEEE, 2012.
[28] T. Ustun, C. Ozansoy and A. Zayegh, "Recent developments in microgrids and example cases around
the world—A review," Renewable and Sustainable Energy Reviews, vol. 15, no. 8, pp. 4030-4041,
2011.
[29] D. Tran and A. Khambadkone, "Energy Management for Lifetime Extension of Energy Storage
System in Micro-Grid Applications," Power Systems, IEEE Transactions on, vol. 4, no. 3, pp. 1289-
1296, 2013.
[30] J. Rittershausen and M. McDonagh, "Moving Energy Storage from Concept to Reality: Southern
California Edison’s Approach to Evaluating Energy Storage," Southern California Edison, 2011.
[32] K. C. Divya and J. Østergaard, "Battery energy storage technology for power systems—An
overview," Electric Power Systems Research, vol. 79, no. 4, pp. 511-520, 2009.
[37] X. Guan, Z. Xu and Q. Jia, "Energy-Efficient Buildings Facilitated by Microgrid," Smart Grid, IEEE
Transactions on, vol. 1, no. 3, pp. 243-252, 2010.
[38] H. Kanchev, D. Lu, F. Colas, V. Lazarov and B. Francois, "Energy Management and Operational
Planning of a Microgrid With a PV-Based Active Generator for Smart Grid Applications," Industrial
Electronics, IEEE Transactions on, vol. 58, no. 10, pp. 4583-4592, 2011.
[39] S. Grillo, M. Marinelli, S. Massucco, F. Silvestro, D. Di Rosa, I. Fastelli and G. Gigliucci, "Generation
122
and battery modelling and integrated control strategies for a better acceptance of intermittent
renewable energy sources in the electric distribution system," in CIRED Workshop, Lyon, 2010.
[40] A. Oudalov, D. Chartouni, C. Ohler and G. Linhofer, "Value Analysis of Battery Energy Storage
Applications in Power Systems," in Power Systems Conference and Exposition, 2006. PSCE '06. 2006
IEEE PES, 2006.
[41] S. Han, S. Han and K. Sezaki, "Economic assessment on V2G frequency regulation regarding the
battery degradation," in Innovative Smart Grid Technologies (ISGT), 2012 IEEE PES, 2012.
[42] A. Bagherian and S. Tafreshi, "A developed energy management system for a microgrid in the
competitive electricity market," in PowerTech, 2009 IEEE Bucharest, 2009.
[43] E. Mashhour and S. M. Moghaddas-Tafreshi, "Bidding Strategy of Virtual Power Plant for
Participating in Energy and Spinning Reserve Markets—Part I: Problem Formulation," Power
Systems, IEEE Transactions on, vol. 26, no. 2, pp. 949-956, 2011.
[44] S. X. Chen , H. B. Gooi and M. Q. Wang, "Sizing of Energy Storage for Microgrids," Smart Grid, IEEE
Transactions on, vol. 3, no. 1, pp. 142-151, 2012.
[45] S. A. Pourmousavi, R. K. Sharma and B. Asghari, "A framework for real-time power management of
a grid-tied microgrid to extend battery lifetime and reduce cost of energy," in Innovative Smart Grid
Technologies (ISGT), 2012 IEEE PES, 2012.
[46] D. Tran and A. M. Khambadkon, "Energy Management for Lifetime Extension of Energy Storage
System in Micro-Grid Applications," Smart Grid, IEEE Transactions on, vol. 4, no. 3, pp. 1289-1296,
2013.
[47] S. Beer, T. Gomez, D. Dallinger, I. Momber, C. Marnay, M. Stadler and J. Lai, "An Economic Analysis
of Used Electric Vehicle Batteries Integrated Into Commercial Building Microgrids," Smart Grid, IEEE
Transactions on, vol. 3, no. 1, pp. 517-525, 2012.
[48] W. Kempton , V. Uto, K. Huber, K. Komara, S. Letendre, S. Baker, D. Brunner and N. Pearre, "A Test
of Vehicle-to-Grid (V2G) for Energy Storage and Frequency Regulation in the PJM System," Center
for Carbon-free Power Integration. University of Delaware, 2008.
[49] PJM, "PJM Manual 35: Definitions and Acronyms," 11 April 2014. [Online]. Available:
https://www.pjm.com/~/media/documents/manuals/m35.ashx.
[50] PJM, "PJM Manual 12: Balancing Operations," 1 December 2013. [Online]. Available:
http://www.pjm.com/~/media/documents/manuals/m12.ashx.
[51] S. Drouilhet and B. L. Johnson, "A Battery Life Prediction Method for Hybrid Power Applications,"
123
National Renewable Energy Laboratory, 1997.
[52] PJM State & Member Training Department, "Training Materials: Generation 301-Ancillary Services,"
June 2013. [Online]. Available: http://www.pjm.com/~/media/training/core-curriculum/ip-gen-
301/gen-301-ancillary-services.ashx.
[54] PJM, "PJM Manual 11: Energy & Ancillary Services Market Operation," 7 March 2014. [Online].
Available: http://www.pjm.com/~/media/documents/manuals/m11.ashx.
[55] PJM, "PJM Regulation Zone Preliminary Billing Data," [Online]. Available: http://pjm.com/markets-
and-operations/market-settlements/preliminary-billing-reports/pjm-reg-data.aspx.
[56] G. Albright, J. Edie and S. Al-Hallaj, "A Comparison of Lead Acid to Lithium-ion in Stationary Storage
Applications," AllCell Technologies LLC, 2012.
[57] EnerSys, "PowerSafe SBS Batteries: Performance Specs," 2014. [Online]. Available:
http://www.enersys.com/PowerSafe_SBS_Batteries.aspx?langType=1033. [Accessed 7 July 2014].
[58] "Demand Response Inverter (4 Port) — DRI-100," Princeton Power Systems, [Online]. Available:
http://www.princetonpower.com/DRI-100.html.
[60] L. Lawton, M. Sullivan, K. Van Liere, A. Katz and J. Eto, "A Framework and Review of Customer
Outage Costs: Integration and Analysis of Electric Utility Outage Cost Surveys," Ernest Orlando
Lawrence Berkeley National Laboratory, 2003.
[61] J. M. Gurrero, L. G. de Vicuna and J. Uceda, "Uninterruptible power supply systems provide
protection," Industrial Electronics Magazine, IEEE, vol. 1, no. 1, pp. 28-38, 2007.
[62] C. C. Grant, "Fire Fighter Safety and Emergency Response for Solar Power Systems," The Fire
Protection Research Foundation, 2010.
[63] "CPI Inflation Calculator," U.S. Bereau of Labor Statistics, [Online]. Available: http://data.bls.gov/cgi-
bin/cpicalc.pl.
124