SIZING METHODOLOGY AND LIFE IMPROVEMENT OF ENERGY STORAGE SYSTEMS IN MICROGRIDS
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University
By
Hussam Jihad Khasawneh
Graduate Program in Electrical and Computer Engineering
The Ohio State University
2015
Dissertation Committee:
Mahesh Illindala, Advisor
Jin Wang
Jiankang Wang
Donald Terndrup
Copyright by
Hussam Jihad Khasawneh
2015
ABSTRACT
The demand for electric power has been steadily increasing, and this trend is expected to continue over the coming decades. With the increased usage of fossil fuels, there has been a growing concern with the environmental impacts of electric power generation.
Therefore, penetration of renewable sources of energy in the modern electric grid has also been increasing. The intermittent nature of renewables introduces uncertainty in the electric grid, which has a negative impact on grid reliability. To address these challenges, renewables are supplemented with energy storage systems (ESS). This dissertation evaluates ESS technologies according to their applications. A generalized method is proposed for ESS sizing for microgrids. This method can be successfully applied to any load profile; it also takes into account operating temperature and aging factors.
In addition, this dissertation presents a variety of ESS life balancing solutions using the new framework of Flexible Distribution of EneRgy and Storage Resources (FDERS). It is based on an in-situ reconfiguration approach through ‘virtual’ reactance and/or ‘virtual’ inertia to change the ‘electrical’ position of each DER without physically displacing it in the microgrid system. Several approaches toward balancing the ESS utilization are proposed taking advantage of the flexibility offered by FDERS framework. It is shown
ii that the estimated ESS cycle life is dependent on factors such as cycling sequence, pattern, and occurrence.
Finally, this dissertation proposes a multi-agent based fleet vehicle-to-grid (V2G) control strategy that intelligently computes each vehicle’s load share based on its battery state-of- health (SoH). Unlike state-of-the-art V2G systems, which treat all the fleet electric vehicles equally with no regard to their diverse driving histories and unequal battery aging, the proposed control strategy employs a programmable on-board smart device to estimate the vehicle’s battery SoH. When multiple vehicles are connected to an isolated microgrid, the individual load shares are dependent on their latest SoH value.
iii
DEDICATION
This dissertation is dedicated to my mother and my father.
iv
ACKNOWLEDGMENTS
Many people favorably influenced the progression and completion of my PhD dissertation. My sincere gratitude goes to my supervisor Prof. Mahesh Illindala for his support and help during all stages of my PhD studies. Prof. Illindala has been unfailingly generous with his time, advice, assistance, and support. I thank the University of Jordan for sponsoring my studies. I thank also Sandia Laboratories for the collaboration in research that has become a part of this dissertation and some stimulating and very helpful discussions. My heartfelt thanks also go to the faculty members, staff, and students at The
Center for High Performance Power Electronics (CHPPE), whose discussions were highly inspiring, and to my friends, who offered their unwavering support and warmth during the progressions and regressions of my research. Finally, I thank my family for believing in me and for their unconditional love and support.
v
VITA
2008...... B.S. Mechatronics Engineering,
The University of Jordan
2011...... M.S. Mechanical and Aerospace
Engineering, The Ohio State University
2014...... M.S. Electrical and Computer Engineering,
The Ohio State University
2012 to present ...... PhD Student, Electrical and Computer
Engineering, The Ohio State University
PUBLICATIONS
H. J. Khasawneh, and M. S. Illindala, “Battery cycle life balancing in a microgrid through flexible distribution of energy and storage resources,” Journal of Power Sources, vol. 261, no. 1, pp. 378-388, 2014. H. Khasawneh, and M. Illindala, “Supercapacitor cycle life equalization in a microgrid through flexible distribution of energy and storage resources,” IEEE Transactions on Industry Applications, vol. 51, no. 3, 2015. M. S. Illindala, H. J. Khasawneh, and A. Renjit, “Flexible distribution of energy and storage resources,” IEEE Industry Applications Magazine, vol. 21, no. 5, 2015.
FIELDS OF STUDY
Major Field: Electrical and Computer Engineering.
vi
TABLE OF CONTENTS
Abstract ...... ii
Dedication ...... iv
Acknowledgments...... v
Vita ...... vi
Publications ...... vi
Fields of Study ...... vi
Table of Contents ...... vii
List of Tables ...... xii
List of Figures ...... xiv
Chapter 1 Introduction ...... 1
Chapter 2 Evaluation of Energy Storage Systems ...... 10
2.1 Applications of Energy Storage Systems (ESS) ...... 10
2.2 Benefits of Energy Storage Systems (ESS) ...... 12
2.2.1 Spinning Reserve ...... 13
2.2.2 Peak Shaving ...... 14
vii
2.3 Review on Energy Storage Systems (ESS) Technologies ...... 14
2.3.1 Batteries ...... 14
2.3.2 Supercapacitor (Ultracapacitor) ...... 20
2.3.3 Flywheel ...... 21
2.3.4 Other Technologies ...... 22
2.3.5 Summary ...... 23
2.4 Hybrid Energy Storage System ...... 24
2.5 ESS Aging Modeling ...... 26
2.5.1 Battery Aging Model ...... 26
2.5.2 Supercapacitor Aging Model ...... 30
2.6 Summary ...... 32
Chapter 3 Sizing Energy Storage Systems...... 34
3.1 Energy Storage Sizing for Long-term Applications ...... 35
3.1.1 Determine the Duty Cycle ...... 35
3.1.2 Calculate the Required Overall Energy ...... 36
3.1.3 Determine Required and Available Capacity ...... 37
3.1.4 Calculate the Battery Pack Size ...... 37
3.1.5 Estimate the Battery Cycle Life ...... 38
viii
3.1.6 Additional Considerations ...... 39
3.2 Energy Storage Sizing for Short-term Applications ...... 40
3.3 Case Study: Sizing ESS for Residential/Commercial/Industrial Loads ...... 41
3.4 Summary ...... 49
Chapter 4 Flexible Distribution of Energy and Storage Resources (FDERS) ...... 50
4.1 Inspiration for FDERS ...... 52
4.2 State-of-the-Art in Integration of Small-Scale Distributed Energy Resources
(DERs) ...... 54
4.3 Qualitative Evaluation ...... 55
4.4 Synthesis of Virtual Reactance and Virtual Inertia...... 57
4.5 Benefits of FDERS ...... 59
Chapter 5 Cycle Life Improvement in Energy Storage Systems ...... 60
5.1 System Description ...... 60
5.2 Problem Identification ...... 65
5.3 Strategies for Balancing the Battery Cycle Life ...... 69
5.3.1 Periodic Cycling – Approach A ...... 69
5.3.2 Power Rating-Weighted Cycling – Approach B ...... 72
5.3.3 Power Rating-Levelized Cycling – Approach C ...... 74
ix
5.3.4 Adaptive Cycling – Approach D...... 76
5.4 Industrial Microgrid with Equal-Rated Distributed Energy Resources ...... 81
Chapter 6 Discussion on Practical Considerations ...... 85
6.1 Different Levels of Fluctuations in Load ...... 85
6.2 Different Values of DER Interface Reactances (Xko) ...... 86
6.3 System of Different Types of DERs ...... 87
6.4 Balancing the Supercapacitor Cycle Life ...... 95
Chapter 7 Sharing Strategy for Fleet Vehicle-To-Grid (V2G) Systems ...... 105
7.1 Motivation ...... 105
7.2 State-of-the-art strategies in V2G systems ...... 106
7.3 Fleet Vehicle-to-Grid (V2G) System and Adopted Models ...... 107
7.4 Load Sharing Strategy and its Benefits ...... 113
7.5 Results ...... 117
7.6 Discussion ...... 119
Chapter 8 Conclusions and Future Work ...... 121
8.1 Conclusions ...... 121
8.2 Contributions...... 125
8.3 Future Work ...... 128
x
References ...... 130
xi
LIST OF TABLES
Table 2.1: Comparison between EES Technologies ...... 24
Table 2.2: Battery aging model constants ...... 30
Table 2.3: Supercapacitor aging model constants...... 32
Table 3.1: Target Cycle Life with the Corresponding Klife ...... 39
Table 3.2: Specifications of Lead-acid Battery ...... 43
Table 3.3. Specifications of Supercapacitor ...... 43
Table 3.4: Battery Sizes (in number of cells) for Peak-shaving of Residential,
Commercial, and Industrial Applications ...... 44
Table 3.5: ESS Sizes (in number of cells) for frequency regulation/spinning reserve
Applications ...... 48
Table 4.1: Qualitative evaluation of FDERS design parameters ...... 56
Table 4.2: Relationships between “What’s” and “How’s” ...... 57
Table 5.1. Motor loading profiles for a 3-phase, 480-V crusher-conveyor ...... 62
Table 5.2: DER ratings and design parameters ...... 65
Table 5.3. Settings of parameters for cycling ...... 71
Table 5.4. Settings of parameters for cycling – Approach C ...... 76
Table 5.5: Parameter settings for adaptive cycling – Li-ion batteries ...... 78
Table 5.6: Comparison of estimated life of batteries (in cycles) – Unequal ratings ...... 81
xii
Table 5.7: DER ratings and design parameters – Equal ratings ...... 82
Table 5.8: Comparison of estimated life of batteries (in cycles) – Equal ratings ...... 84
Table 6.1. Design parameters for 2x75-kW DERS system shown in Figure 6.3 ...... 88
Table 6.2: Battery sizing particulars for the 2x75-kW DERS system ...... 91
Table 6.3: Design parameters for obtaining results shown in Figure 6.6 ...... 93
Table 6.4: Supercapacitor Specifications [128] ...... 97
Table 6.5: DER ratings and design parameters for microgrid in Figure 6.3 ...... 98
Table 6.6: Parameter settings for adaptive cycling – Supercapacitors ...... 104
Table 7.1: Comparison of Battery Life (in years) for the Proposed Solution against
Baseline Case ...... 119
Table 7.2: Comparison of Battery Life (in years) for the Proposed Solution against
Baseline Case (Dissimilar Initial SoH) ...... 120
xiii
LIST OF FIGURES
Figure 1.1: Annual U.S. Solar PV Installations, 2000-2014 [3] ...... 2
Figure 1.2: Energy Storage Deployment in the U.S.A. during 2013-2015 [16] ...... 5
Figure 1.3: Microgrid with distributed energy and storage resources [29] ...... 6
Figure 2.1: Voltage regulated DER featuring Surge Module ...... 11
Figure 2.2: Randle equivalent electrical circuit model ...... 17
Figure 2.3: Block diagram of control-oriented Li-ion battery model ...... 19
Figure 2.4: Electrical equivalent circuit model of the supercapacitor ...... 21
Figure 2.5: A Ragone plot showing the specific energy density versus power density for various energy storage devices ...... 25
Figure 2.6: Factors affecting battery aging (a) SoC average, (b) SoC standard deviation,
(c) Temperature ...... 28
Figure 2.7: Sample state-of-health (SoH) curves ...... 30
Figure 2.8: Sample capacitance retention curves ...... 32
Figure 3.1: Schematic diagram of a microgrid test bed for peak shaving application showing the various distributed energy resources and energy storage options ...... 42
Figure 3.2: Typical daily power output for Photovoltaic ...... 42
Figure 3.3: Typical daily residential loadsupply and demand curves with required energy for battery (hatched) ...... 45
xiv
Figure 3.4: Typical daily commercial load—supply and demand curves with required energy for battery (hatched) ...... 45
Figure 3.5: Typical daily industrial load—supply and demand curves with required energy for battery (hatched) ...... 46
Figure 3.6: Frequency of microgrid when (a) 80kW step load is applied and (b) 90kW step load is applied causing stalling of generator ...... 48
Figure 3.7: Frequency of microgrid with supercapacitor is present as the spinning reserve when 90kW step load is applied ...... 48
Figure 4.1: Transient dynamic response of a fuel cell-battery hybrid system ...... 52
Figure 4.2: Energy saving formations ...... 53
Figure 4.3: DER outer loop power controller block diagram (k = 1, 2, … , n) ...... 55
Figure 4.4: Synthesis of variable reactance in the DER’s controller ...... 58
Figure 5.1: Single-line diagram of a microgrid consisting of four fuel cell-battery hybrid
DERs ...... 61
Figure 5.2: Basic hydrogen fuel cell ...... 63
Figure 5.3: Power flow and control structure of each fuel cell-battery hybrid DER ...... 63
Figure 5.4. Block diagram of simulation strategy for microgrid operation and battery life estimation ...... 67
Figure 5.5: Ideal case (a) power profile, and (b) battery state of health (SoH) ...... 67
Figure 5.6: Power response characteristic ...... 68
xv
Figure 5.7: Baseline case (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 70
Figure 5.8: Power profiles of (a) Order I, (b) Order II, (c) Order III, and (d) Order IV ... 72
Figure 5.9: Approach A (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 73
Figure 5.10: Approach B (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 75
Figure 5.11: Approach C (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 77
Figure 5.12: Block diagram for Approach D (of FDERS)...... 77
Figure 5.13: Approach D (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 79
Figure 5.14: Equal Ratings Baseline (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 82
Figure 5.15: Equal Ratings Approach D (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH ...... 83
Figure 6.1: Analysis for different levels of load ...... 86
Figure 6.2: Analysis for different values of DER interface reactances ...... 87
Figure 6.3: Single line diagram of a 2-DERS system supplying a crusher-conveyor load
...... 88
Figure 6.4: Battery life analysis for 2-DERS microgrid in Figure 6.3 ...... 90
xvi
Figure 6.5: Block diagram illustrating division of power flow in each DERS ...... 91
Figure 6.6: Response of the 2-DERS system with the implementation of FDERS ...... 94
Figure 6.7: Battery aging in a 2-DERS system illustrating the benefits of FDERS ...... 94
Figure 6.8: Comparison between the battery replacement times for the 2-DERS system –
(a) conventional (without FDERS) and (b) with FDERS ...... 95
Figure 6.9: Single-line diagram of a microgrid consisting of four fuel cell-supercapacitor hybrid DERs supplying a crusher-conveyor ...... 97
Figure 6.10: Block diagram of modeling strategy for supercapacitor life estimation ...... 98
Figure 6.11: DER power profiles for baseline case ...... 99
Figure 6.12: (a) Voltage and (b) current of the supercapacitor module for one cycle .... 100
Figure 6.13: Block diagram for the adaptive cycling approach to equalize the lifetimes of supercapacitors in the microgrid ...... 101
Figure 6.14: Equal ratings - baseline case (a) supercapacitor SoC, (b) temperature and
(c) capacitance retention ...... 102
Figure 6.15: Equal ratings – FDERS solution (a) supercapacitor SoC, (b) temperature and
(c) capacitance retention ...... 103
Figure 7.1: Supply and demand of the microgrid ...... 108
Figure 7.2: A microgrid for testing the fleet V2G systems consisting of PV, FC, charging stations for the fleet vehicles at a commercial office building ...... 109
Figure 7.3: Driving cycles – (a) UDDS (b) HWFET ...... 110
Figure 7.4: Batteries parameters for different driving cycles ...... 112
xvii
Figure 7.5: Batteries (a) SoC when connected to the microgrid, (b) life aging parameter, and (c) overall SoH for baseline scenario...... 115
Figure 7.6: Block diagram of the BEV agent...... 116
Figure 7.7: Block diagram of charging station (CS) agent ...... 116
Figure 7.8: Batteries (a) SoC when connected to the microgrid, (b) life aging parameter, and (c) overall SoH after applying the proposed strategy...... 118
Figure 7.9: Batteries SoH when starting at dissimilar SoH values ...... 120
xviii
Chapter 1
INTRODUCTION
The world’s electricity delivery system is experiencing a continuously escalating energy demand. In addition, there are growing concerns about the environmental impacts of the electricity generation from fossil fuels. Newer regulations and policies are being enacted to limit the CO2 and other greenhouse gas emissions as per the recommendations of the
Intergovernmental Panel on Climate Change (IPCC) and International Energy Agency
(IEA). This is important because fossil fuel electricity generation accounts for around
40% of overall energy related CO2 emissions, or one quarter of total greenhouse gas emissions [1]. For this reason, the penetration of renewable energy sources in the modern electric grid has been rapidly increasing and the same trend is expected to continue in the future. In 2014, the photovoltaic (PV) installations reached 6,200MW in the U.S. This number is 30% above the 2013 figures and around 12 times the amount installed in 2009, see Figure 1.1. Furthermore, wind energy annual global market in 2014 was 12.5% higher as compared to 2013. The growing trend for renewables is driven mainly by the decreasing cost of renewable technologies, smart metering policies, and renewable portfolio standards (RPS) [2].
1
6000
5000
4000
3000
2000
1000
Photovoltaic Photovoltaic PV Installations (MW) 0 2000 2002 2004 2006 2008 2010 2012 2014 Year Figure 1.1: Annual U.S. Solar PV Installations, 2000-2014 [3]
The variable nature of the renewable sources of energy introduces another source of uncertainty in the operation and planning of electric power systems. Therefore, the modern electric grid with higher penetration of renewables faces major challenges.
Adding intermittency in renewable sources to the changing usage patterns is leading to greater variability and uncertainty in the electricity delivery; this may also have a negative impact on the grid reliability. Such an impact may manifest itself in the form of electricity shortages, power quality problems, and blackouts. In general, these problems result from two main issues with the current electricity delivery system: first, there is not always enough power generation available to meet peak load demand; second, the existing transmission lines in certain areas cannot carry all of the electricity needed by consumers.
2
Generally, the Independent System Operator (ISO) directs the thermal, controllable power plant units to increase or decrease with the instantaneous or variable demand.
While the demand has always been variable, the system operators are required to match these variations with controllable resources (mostly dispatchable thermal generation).
However, studies indicate that with high levels of non-dispatchable renewable sources of energy, matching generation and demand in the grid can be quite challenging [4].
Simultaneous matching of variable demand and variable supply needs tighter controls.
Renewable sources of energy typically exhibit very low marginal costs and hence are most often operated at their maximum available output. Such an operation makes it necessary to treat the renewables as non-dispatchable energy sources in the sense that operators cannot control their output power. To maintain the reliability, system operators must continuously match the demand for electricity with supply on a second-by-second basis, and this can be very challenging.
To manage the variability of renewables, several solutions are possible. For instance, constructing large power plants and expanding transmission allows higher penetration of renewables by providing greater flexibility. But this is not cost-efficient due to the magnitude of energy required to make them profitable [5]. A second possible solution is improving the forecasting of renewable sources like solar radiation and wind speed to reduce dispatch errors in the system. However, it does not give full economic opportunity to the renewables sources [6]. Another possible solution is the increase of dispatchable back-up power generation to improve the system’s ability to cope with dispatch errors.
Yet this may be at the cost of greenhouse gas emissions, since these dispatchable units
3 are generally using fossil fuels for power generation [7-9]. Although hydro power can also respond and absorb the fluctuations from renewables, it is not a viable option at many places [10]. Decoupling renewable sources of energy from the grid is a feasible solution because it reduces the power-quality problems in the grid. However, this will cut off the clean energy sources from feeding the power grid [11, 12].
None of the above solutions can mitigate all the challenges resulting from integration of renewables in the electric grid. The energy storage systems (ESS) can adequately resolve various renewable integration issues and challenges [13-15]. In 2014, the U.S. has installed ESS of 61.9 MW, which is about 40% more compared to 2013 according to the
U.S. Energy Storage Monitor, a quarterly publication issued by a joint research partnership between Greentech Media Inc. Research and the Energy Storage Association
(ESA) [16]. The same report also predicts the year 2015 to be the biggest in energy storage market’s history with deployment of 220 MW, which approximately doubles the capacity installed in 2013 and 2014 combined, see Figure 1.2.
An issue that modern electric grids must address is the infrastructure needed to accommodate renewable sources and ESS. The distributed energy resources (DERs) are being increasingly preferred to meet the growing needs of high energy consumers with minimal changes to the existing power grid. DERs refer to a variety of small, modular power-generating technologies that can be combined with energy management and ESS to improve the operation of the electric grid [17]. Hence, the DER technologies can play an increasingly important role in the coming decades [18, 19]. They typically employ small electricity prime-movers and energy storage technology to augment the electricity
4 supplied by a large, central-station power plant. The different kinds of prime-mover technologies include internal combustion engines, wind generators, solar/photovoltaic
(PV), microturbines, and fuel cells [20, 21].
250
200
150
100
50
Energy Storage Energy Deployments (MW) 0 2013 2014 2015E Year Figure 1.2: Energy Storage Deployment in the U.S.A. during 2013-2015 [16]
At higher renewable penetration levels, the available power output may exceed the system load demand. This makes limiting the renewables’ output inevitable unless there is flexibility available in the electric grid. Such flexibility can be provided by energy storage systems (ESS). Conceptually, ESS stores the excess energy in the system for later use when other means of supplying power in the grid are unavailable, uneconomical, or when unforeseen additional power is demanded [22]. The storage device used as ESS in an electric grid can be charged not only by electricity generated from a renewables but also by main grid or by non-renewables like fuel-cells and microturbines [23].
5
A cluster of interconnected DERs, loads, and ESS that co-operate with each other to be collectively treated by the grid as a controllable load or generator is called a microgrid [21, 24]. The key objectives of a microgrid are to facilitate the high penetration of DERs without causing power quality problems to the distribution network, and to provide reliable energy delivery to sensitive loads [25, 26]. The components that constitute a microgrid may be physically close to each other or distributed geographically
[27]. Figure 1.3 depicts a typical microgrid. It is noteworthy that the interface of DER with the system is either a rotating machine generator or a power electronic inverter. This was not shown in Figure 1.3 since it is considered an inherent part of the DER [26, 28].
Figure 1.3: Microgrid with distributed energy and storage resources [29]
6
The microgrid can operate in grid-connected or islanded mode of operation. In the grid- connected mode, the microgrid either draws power from or supplies it to the main grid, depending on the generation and load mix and implemented market policies. However, it should not tightly regulate the voltage at the point of common coupling (PCC) [30]. It must be able to separate or island itself from the main grid when an abnormal condition or severe power quality event occurs [24, 31]. During this mode of operation, the primary function of the microgrid is to satisfy all of its load requirements and contractual obligations with the grid. Operation under this mode may cause multiple problems a key problem is the voltage and frequency management when the main grid is absent. If the voltage and frequency are not well regulated, it can lead the microgrid to a voltage and/or frequency collapse. ESS can play an important role to maintain voltage and frequency as a spinning reserve1 [32]. This is particularly important when the load is continuously changing and DERs in the microgrid have slow response time, for e.g., fuel- cells and microturbines [21] . On the other hand, the ESS also ensures the supply-demand balance in case the demand does not match the available supply at any time. Under islanded mode of operation, communication between DERs and ESS is vital to maintain the integrity of the microgrid.
This dissertation proposes a novel sizing methodology and operating strategies to improve cycle life of energy storage systems (ESS) in microgrids. Chapter 2 includes a comprehensive review of the different applications and benefits of ESS in microgrids. It
1 Spinning reserve can be defined as generation capacity that a system holds in reserve to prevent interruption in the service, when some part of the operating electric supply resources becomes unavailable.
7 also presents the various technologies available for ESS as batteries, supercapacitors, and flywheels. This chapter also reviews the aging models adopted in the later chapters.
The new contributions made in this dissertation are presented in the remaining chapters. Chapter 3 proposes a generalized method for EES sizing for microgrids. The proposed method uses energy balance concept rather than current levels. This method uses the load profile of the targeted industry, and calculates the battery size by taking into account operating temperature and aging factors. Chapter 3 ends with a case study where a battery pack and a supercapacitor module are sized for a residential, commercial, and industrial microgrid.
Chapter 4 presents the new framework known as Flexible Distribution of EneRgy and
Storage Resources (FDERS). This framework was proposed for a reliable supply of the fast varying loads from a network of multiple smaller-rated DERs, especially when power from the main grid is not available [29]. FDERS was inspired by the cooperative
V-shape formation of flocks of birds and peloton/echelon formation of cycling racing teams for extending their endurance limits. This chapter illustrates the analogy of such cooperative formations found in nature with the integration of DERs in a microgrid to achieve greater sustainability through the benefits of increased resource lifetime and optimal energy storage deployment.
In Chapter 5, a microgrid consisting of four fuel cell-battery hybrid DERs is devised for an industrial crusher-conveyor load. Each fuel cell was accompanied by a Li-ion battery to provide energy storage support under islanded condition of the microgrid. This is necessary because the fuel cells are characterized by a poor transient response to sudden
8 change in demand. This chapter presents several battery life balancing solutions making use of the FDERS framework.
Chapter 6 discusses the practical considerations of FDERS for achieving cycle life improvements of energy storage systems. Furthermore, it also discusses the application of
FDERS on various types of prime-movers across different ESS technologies.
Chapter 7 presents a novel control strategy for fleet vehicle-to-grid (V2G) systems to meet diverse customer demands. This proposed strategy intelligently computes each vehicle’s load share based on its battery state-of-health (SoH). Such a strategy ensures that a battery with higher SoH contributes more for the load as compared to a lower SoH battery. Finally, Chapter 8 presents the conclusion that summarizes the contributions expounded in this dissertation along with possible topics of future research.
9
Chapter 2
EVALUATION OF ENERGY STORAGE SYSTEMS
Energy storage systems (ESS) have become an important element of the microgrid. ESS and its associated power electronics have in fact seen major technological innovations in the past decade, heralding a new era of high penetration of cost-effective solutions [33].
The ESS in a microgrid allows peak shaving, i.e., the shifting of electrical power between the time the power is generated and the time it is used. It also provides a spinning reserve that is necessary when a prime mover has slow dynamic characteristics, when anomalies occur in the microgrid such as loss of generation, or when the price of electrical energy changes [34].
2.1 Applications of Energy Storage Systems (ESS)
Microgrids use ESS to fulfill two main purposes – (i) to supply energy for short-term requirements, called power applications [21], and (ii) to supply energy for long-term needs, called energy applications. The power applications require high power output for short time between a few seconds to a few minutes [33]. Storage technologies used for power applications could be supercapacitors. This is because they have high power density, low internal resistance, and a high cycle life. Recent advancements in
10 supercapacitor technology have established them as a practical and sustainable choice for short-term high-power applications [35]. Besides supercapacitors, flywheels are another candidate characterized by their high power density and their utility for unlimited number of cycles [36, 37].
A power electronics-based distributed energy resource (DER) includes prime mover/energy source and power electronics as the utility interface [38]. If such DERs are adequately rated to supply the steady state load demand in the microgrid, the short-term
ESS function is in meeting the mismatch between change in load demand and DER response [29]. Some of the short-term energy storage options are connected to the DC link of the distributed energy resource (DER), see Figure 2.1. This figure shows the power-electronic-interface of the DER that includes a storage (surge) module connected to the prime-mover through the DC-bus which allow the AC-DC inverter to draw its energy from either the prime-mover or the energy storage, as needed [39]. However, short-term ESS could be directly connected to the AC side of the microgrid, such as flywheels.
Figure 2.1: Voltage regulated DER featuring Surge Module
11
The second purpose of ESS in microgrids is to supply larger amount of energy for long- term requirements. These load demands are called energy applications. Such long-term applications need energy storage discharge from many minutes up to multiple hours.
Therefore, ESS should have a large energy storage capability. Storage technologies that are best suited to energy applications include pumped hydro, compressed-air, and most battery types [33]. Long-term ESS enhances the economic benefits of microgrid by adding features such as peak shaving and energy time-shift, which will be explained in
Section 2.2 [38]. Long-term storage can be connected at any point in the microgrid, because it is independent of the individual DER characteristics.
2.2 Benefits of Energy Storage Systems (ESS)
A major benefit of ESS in microgrids is that it supports high penetration of renewable energy resources. The renewables may also generate a significant percentage of electric energy during off-peak times, when there is a low financial value; for example, at night, on weekends, or during holidays. Long-term ESS can be used in conjunction with renewable energy resources to store their energy during off-peak hours. Later, the same stored energy may be utilized to offset other energy costs or to be sold when it is financially viable.
There are two main roles for a microgrid – (i) generating electric energy from prime movers, (ii) delivering this energy to the consumers. However, additional resources are needed to support microgrid operation. The Federal Energy Regulatory Commission
(FERC) defines ancillary services as those essential to support the delivery of electric
12 energy from prime movers to end-users while maintaining the integrity and reliability of the interconnected transmission system [40]. This dissertation covers the following ancillary services: spinning reserve, frequency regulation, and peak shaving.
2.2.1 Spinning Reserve
In order to maintain the supply-demand balance between dispatchable and non- dispatchable renewable DERs, and load demands, the microgrid needs to have sufficient reserve. This reserve needs to be adequate to compensate for any sudden change in the renewable DERs power output or load demand. Any shortage in the characteristics of this reserve (such as insufficient power capability or poor frequency response) may cause drop in microgrid’s frequency, transmission system violations, stability issues, and other reliability problems. Spinning reserve can be defined as the generation capacity that a system holds in reserve to prevent interruption in the service, when some part of the operating electric supply resources becomes unavailable. Spinning reserve must be connected to the grid and capable of responding within 10-15 minutes to compensate for generation or transmission outages. Furthermore, frequency-responsive spinning reserve responds in less than 10 seconds to maintain system frequency [33, 40]. Frequency regulation is associated with spinning reserve and defined as the process of holding the frequency of microgrid within strict tolerance limits. In an islanded microgrid, due to its low inertial response, the frequency deviation is the main indicator of inadequate power generation (high load demand). A solution to such problem could be an added power generation from an ESS or generator inertia, or load shedding (smart loads) [41].
13
Frequency-responsive spinning reserve is considered as a short-term (power) application of ESS.
2.2.2 Peak Shaving
Peak shaving refers to supplying the load during on-peak interval by discharging the ESS and to planning the recharge during off-peak hours [42, 43]. Peak shaving has been accomplished for many years by operating on-site gas turbines and diesel generators [44].
However, nowadays users can install long-term ESS that can discharge during the peak periods and charge during the low demand periods, thus decreasing the peak demand
[45]. Peak shaving is considered a long-term (energy) application of ESS.
Other benefits of ESS include transmission support, voltage regulation, power quality and reliability, and customer load management [46]. Peak shaving and frequency regulation are the main subjects of this chapter. The next section reviews various technologies available for ESS.
2.3 Review on Energy Storage Systems (ESS) Technologies
There exists a wide array of technologies to implement the energy storage systems (ESS) in a microgrid. Among them, the major ones include batteries, supercapacitors, and flywheels.
2.3.1 Batteries
Batteries are the most common and successful energy storage technology. They are attractive due to their price, commercial availability, and ease of use. Their high energy
14 density2 reflects the large amount of energy that can be obtained from the electrochemical cells; their low power density3 reflects the limit on how fast this energy can be provided to the load. The limitation on power density is a result of the time required for the electro- chemical reaction to occur and the extremely high internal equivalent resistance of the cells.
A typical battery cell consists of negative electrode (or anode), electrolyte, separator, and positive electrode (or cathode); the two electrodes are separated by the separator, and electrodes are filled with electrolyte. The electrolyte, often a good ionic conductor but an electronic insulator, provides a transport-medium for ions to travel between the two electrodes and keeps electrons in the external circuit [47]. Among the different types of batteries, the most dominant types are covered in this section.
Lead-acid Battery
Lead-acid battery is the most developed secondary battery4 technology because of its high efficiency, high reliability, low cost, and high degree of maturity over years [46].
However, lead-acid battery has relatively short cycle life, poor low temperature performance, and low energy density due to the inherent high density of lead as a metal
[48]. Despite these disadvantages, the lead-acid battery maintains high power density, which implies the ability to supply large surge currents.
2 Energy Density is defined as is the amount of stored energy divided by the volume (Wh/L) or mass (Wh/kg). 3 Power density is defined as the rated output power divided by the volume (W/L) or mass (W/kg) of the EES element 4 Secondary battery is the rechargeable battery which its electrochemical reactions are electrically reversible
15
During discharge in lead-acid batteries, positive hydrogen ions migrate out of the anode to the cathode. In conjunction with the hydrogen ion transport, sulfuric acid is consumed in reactions at both the anode and cathode. When a source is applied to the battery to recharge it, the reverse reactions occur. The reversible reactions for the cathode and anode are given in (2.1) and (2.2), respectively.
− + − PbO2 + HSO4 + 3H + 2e ⇔ PbSO4 + 2H2O (2.1)
− + − Pb + HSO4 ⇔ PbSO4 + H + 2e (2.2)
Lithium-ion (Li-ion) Battery
The Li-ion battery is another popular secondary battery. The rising popularity of Li-ion battery is due to its excellent energy density, power density, efficiency, cycle life, fast- charging capability, and low self-discharge rate. However, the initial investment necessary for Li-ion batteries is still considered high when compared to the lead-acid batteries [37]. The economic problems are exacerbated by the limited lithium reserves in the world [49]. Similar to the lead-acid battery, during their charging, the lithium ions migrate out of the cathode, travel through the electrolyte, and collect in the anode. This builds up a potential across the two electrodes and when the cell is disconnected from a charger and connected across a load, the reverse processes occur. The reversible reactions for the cathode and anode are described in (2.3) and (2.4), respectively.
+ − LiCoO2 ⇔ Li0.5CoO2 + 0.5Li + 0.5e (2.3)
+ − 0.5Li + 6C + 0.5e ⇔ Li0.5C6 (2.4)
16
In this dissertation, the Li-ion battery is modeled on the basis of the Randle equivalent circuit phenomenological model [50]. This model, shown in Figure 2.2, consists of a voltage source (Voc) and an internal resistance R0 connected in series to n-times parallel
RC circuits representing diffusion within a cell. All the circuit parameters are dependent on the battery SoC and temperature.
Figure 2.2: Randle equivalent electrical circuit model
The battery terminal voltage is computed as [51, 52]:
n
Vbatt = Voc − R0I − ∑ Vi (2.5) i=1 where Vbatt is the battery terminal voltage, Voc is the open circuit voltage, and Vi is the voltage across the ith parallel RC circuit. The number of RC circuits defines the order of the model. Of course, a higher order model provides an improved frequency response, leading to a better time domain approximation, yet increasing the system's complexity.
Therefore, a reasonable order model has been employed. The simple structure and low computation effort of the low-order equivalent circuit models is advantageous for designing control algorithms. It is noteworthy that the control-oriented model for Li-ion battery was built on two platforms - thermal and electrical - as indicated in Figure 2.3.
17
As seen in Figure 2.3, the thermal model computes heat generated by the discharge current, and then determines the temperature of the battery, assuming that the battery temperature distribution is uniform in all directions at any instant of the transient heat transfer process. The battery temperature is determined by solving the differential equation [53]:
dT Q̇ − Ah(T − T ) = ∞ (2.6) dt m c where 퐴ℎ(푇 − 푇∞) is the dissipated heat by convection, m is the mass, c is the specific thermal capacity, and 푄̇ is the heat generated by Joule effect in the battery which is given by:
N 2 2 Q̇ = RI + ∑ Ri Ii (2.7) 1 After computing the temperature, the battery model estimates the state of charge (SoC) by solving:
t ηI(t) SoC(t) = SoC0 − ∫ dt (2.8) 0 Cn where SoC0 is the initial state of charge, Cn is the nominal cell capacity, and η is the
Coulombic efficiency factor that is assumed to be unity.
18
Figure 2.3: Block diagram of control-oriented Li-ion battery model
Nickel-Cadmium Battery
Nickel-cadmium (NiCd) batteries have a higher energy density and longer cycle life, comparable to the Li-ion batteries. This makes them good for uninterruptible power supply (UPS) and generator-starting applications [48]. However, NiCd battery sales suffered a decline due to increasing environmental controls for toxic cadmium. The 2006
European Union’s directive banned the NiCd batteries in September 2008. Nowadays, the newer battery technologies are being considered as viable alternatives to the NiCd batteries.
Sodium-Sulphur Battery
Sodium-Sulphur Batteries (NaS) batteries are an attractive emerging technology for ESS to integrate with renewable energy resources. They have higher power density, efficiency, and longer cycle life. Besides, the NaS batteries are environmentally friendly since there are no emissions during their operation. Furthermore, more than 99% of the battery pack material can be recycled. However, the sodium is still treated as a hazardous
19 material. On the other hand, NaS batteries are feasible only for applications of large ratings (around 1 MW) because of their high operating temperature (300oC-350oC) [48].
2.3.2 Supercapacitor (Ultracapacitor)
Supercapacitor, also known as ultracapacitor, is electric double-layer capacitor. It is known to have high power density and an exceptional cycling capability. Moreover, it has superior cycle efficiency and a longer cycle life [35, 54]. Although supercapacitors have lower energy density, they exhibit significantly higher power density in comparison to batteries. Therefore, they have been recognized as a practical and viable option for short-term high-power applications. A fundamental disadvantage of a supercapacitor is its large self-discharge which makes it lose more than 20% of its stored energy per day, even if no load is connected.
In this dissertation, a new approach is considered for supercapacitors modeling using impedance spectroscopy. The supercapacitor is modeled, as shown in Figure 2.2, by an inductor L, a series resistor Ri, and the complex pore impedance modeled by a capacitor C in series with parallel-connected RC circuits that define the order of the model. The more parallel RC circuits, the more accurate is the model; but this comes at the cost of complexity. The full model and the derivations of all the equations and parameters along with experimental validation can be found in [55].
20
Figure 2.4: Electrical equivalent circuit model of the supercapacitor
2.3.3 Flywheel
Flywheel is a common type of mechanical energy storage technology, in which the energy is stored as rotational momentum. The maximum kinetic energy that can be stored in a flywheel is directly dependent on its moment of inertia and the maximum achievable angular velocity. Rotational momentum is the product of the flywheel’s moment of inertia and its angular velocity. The physical geometry of a particular flywheel and its internal distribution of mass determine its moment of inertia. Energy is converted between electrical energy and rotational momentum using an electric machine. To store energy in the flywheel, the electric machine operates as a motor to convert the electrical energy into a mechanical rotation. On the other hand, to retrieve energy from the flywheel, the machine operates as a generator producing electric energy resulting in a reduced rotational momentum.
The energy and power densities of a flywheel look good in theory. However, they are limited by the practicalities of building an effective flywheel system. The necessary equipment includes an electric machine, which is typically sized for high speed capability over torque to enable high angular velocity. Energy density of a flywheel is limited only
21 by maximum possible angular velocity. Power density is directly influenced by the maximum torque the machine can handle at any given angular velocity. One great advantage of flywheels is that they have, theoretically, infinite cycle life. However, they need regular maintenance. The main disadvantage is that the cost of fabricating, assembling, and installing sufficient support structuring for flywheels is high.
2.3.4 Other Technologies
Other energy storage technologies in microgrids include pumped-hydroelectric (hydro) energy storage, compressed-air energy storage (CAES), and superconducting magnetic energy storage (SMES). These technologies will be briefly explained in this subsection.
Pumped-hydroelectric (hydro) storage (or PHS) stores potential energy from vertical height differences in water levels between two reservoirs. PHS has large power capacity
(1000 MW), relatively high efficiency (65-85%), large storage capacity (24 hours), and a long life (35-60 years). However, PHS has multiple drawbacks such as its high capital cost (600 - 2,000$/kW), the damages to environmental it causes such as flooding huge land to make reservoirs, and also the long project lead time (typically in 10s of years) [13,
48].
Compressed-air energy-storage (CAES) system stores gas in a tank when energy demand is low that is released during periods of higher demand. There are only two CAES systems in the world. The first one is in Huntorf, Germany with a capacity of 580 MWh.
It was initially installed to support a nuclear plant, but now is used for grid support [56].
This system has showed 90% availability and 99% reliability [57]. The second CAES system was built in McIntosh, Alabama, U.S.A. with a capacity of 2860 MWh. This
22 system was designed to support a coal power plant in Lohman, Alabama [48]. Neither of these two systems were used with renewable energy resources. A new CAES project was planned for commissioning in Des Moines, Iowa in 2015. It would have been the first
CAES plant to use wind-energy to store compressed air. However, this project was terminated after 8 years of development and planning because of site geological limitations [58].
Superconducting magnetic energy storage (SMES) system is a device that stores energy in magnetic field generated by direct current flowing through a superconducting coil [48].
SMES systems can only generate electricity at rated capacity for a few seconds, and are extremely expensive. Nevertheless, it was found that the supercapacitors and flywheels can be a viable alternative to the SMES [59, 60].
2.3.5 Summary
The key performance metrics of various EES technologies are summarized in Table 2.1.
A Ragone plot that compared the performance of these technologies is depicted in
Figure 2.5. From both the table and figure, it is apparent that supercapacitors are the best candidate for power applications because of their high power density, long cycle life, and lower internal resistance that leads to less ohmic losses and heating. The flywheels have relatively lower power density than supercapacitors. However, it has almost unlimited number of cycles, a quality that makes them a good choice for power applications. For energy applications, on the other hand, batteries are good candidates because of their large energy density. Among the batteries, the lead-acid type have been used widely for energy applications because of high efficiency and low cost. Li-ion batteries, however,
23 provide a realistic compromise between the energy density and power density. The relatively higher power density (with respect to Lead-acid batteries) makes them a good storage system for high-power applications. Furthermore, Li-ion batteries have a higher energy density as compared to supercapacitors, and this makes them suitable for portable applications.
Table 2.1: Comparison between EES Technologies Lead-acid EES Technology Li-ion battery Supercapacitor Flywheel battery Power density 200-1,000 1,500-3,000 5,000-10,000 1,000 (W/kg) Energy density 30-50 150-250 5-15 1-10 (W/kg) Specific energy 50-200 500-1,500 3,000-5,000 500 cost ($/kWh) Power efficiency 70-85 80-90 95-98 90-95 (%)
Cycle life 500-800 1,200 >50,000 >50,000
Internal 3-5 7-10 0.1-0.2 N/A resistance (mΩ) Self-discharge 0.1-0.3 0.1-0.3 20-40 100 per day (%)
2.4 Hybrid Energy Storage System
As shown in Section 2.3, there is no single energy storage technology with both high energy density and high power density at the same time. Therefore, the hybrid energy storage system (ESS) is a plausible approach for achieving performance improvement by utilizing the best combination of ESS technologies. Rather than relying on a single technology of EES, such a hybrid ESS presents distinct advantages of multiple EES technologies and elements and suppresses their drawbacks. For instance, an industrial
24 plant with fast varying load exhibits frequent charge and discharge cycles with a short period of time and a large amount of current. Under such severe conditions, it is difficult to maintain a high efficiency and longer cycle life if only lead-acid batteries are employed. On the other hand, the use of supercapacitors can considerably improve the efficiency and cycle life. However, supercapacitors have main drawbacks in terms of energy density and cost, which makes it expensive to replace the batteries with supercapacitors completely. However, using supercapacitors to complement the batteries reinforces the drawback of the batteries in the form of high power density, long cycle life, and high efficiency [61, 62], This introduces the battery-supercapacitor hybrid ESS. 1000
Li-ion NaS Battery Battery
100 NiCd
(Wh/kg)
Battery
Lead-acid Battery Super- 10 capacitor
Energy Density
Flywheel
1 100 1000 10000 Power Density (W/kg)
Figure 2.5: A Ragone plot showing the specific energy density versus power density for various energy storage devices
25
The battery-supercapacitor hybrid ESS enhances the ability of the battery to accept and deliver high levels of power with low levels of electrical resistance. Supercapacitors store surplus energy from the battery during low demand periods and then provide extra energy during peak load current demand period. Traditionally, batteries and supercapacitor are separate components relying on electronic controllers and complex algorithms to switch power between both units. Incidentally, East Penn Manufacturing, through its subsidiary
Ecoult, has designed and constructed UltraBattery® a hybrid device containing both a supercapacitor and a battery in a common electrolyte. This enables a great balance between an energy storage lead-acid battery with the high-power discharge capabilities and long cycle life of a capacitor [63]. The new challenge posed by the battery- supercapacitor hybrid ESS is the sizing it needs to be carefully designed to maximize the benefits of the hybrid ESS over the traditional energy storage technologies.
2.5 ESS Aging Modeling
Before presenting the sizing and design requirements for ESS, aging cycle life estimation are explained in this section. It is vital to understand the way ESS ages since these numbers will be part of the ESS sizing presented in Chapter 3. This section explains the adopted aging models for battery and supercapacitor.
2.5.1 Battery Aging Model
Batteries are often classified into three categories primary, secondary, and reserve
[64]. While the primary batteries are made of chemical elements that cannot be
26 electrically recharged, the secondary batteries can be repetitively discharged and recharged. On the other extreme, the reserve batteries are designed for long-term storage.
In particular, this dissertation focuses on the secondary batteries to meet the power and energy demands of microgrids.
A battery, in general, has three main usage phases charging, discharging, and resting.
A discharge followed by a charge is called a cycle. The deliverable capacity of a Li-ion battery tends to decrease due to the cycling. This phenomenon is called battery aging
[65-67]. Battery aging depends on factors including the maximum charging voltage, and the discharge current level and rate [64, 68]. Discharging the cell at higher currents and to deeper depths of discharge (DoD) also results in shorter cell life. A descriptive measure of the discharge current level and rate is the average and the standard deviation of state of charge (SoC), respectively [69]. Batteries with higher SoC are expected live longer than batteries with lower SoC, see Figure 2.6(a). Those with higher standard deviation (lower charging/discharging rate) have longer cycle life as compared to ones with lower standard deviation, as shown in Figure 2.6(b). In addition, the battery temperature has an influence on its aging. The increase in temperature results in faster consumption of the battery reactants causing them to be consumed quickly, and hence shortening the cell life [70], as depicted in Figure 2.6(c).
27
100 100 98 Batt1 95 96 Batt2 94 Batt1 Batt2 90 92 SoC1 SoC1 50 Batt1 48 Batt2 C) o 46 44 42 40 Temperature ( Temperature SoC1 0 20 40 60 80 100 Time (s) (c) Figure 2.6: Factors affecting battery aging (a) SoC average, (b) SoC standard deviation, (c) Temperature The battery aging model must account for all the aging factors discussed above to insure accuracy. In this dissertation, the battery aging model adopted in [69] was used. This model has been validated in [71] against real test data for the specific battery cell A123 ANR26650M1A [72]. A battery aging is the damage to its life by variable (L), which can vary from 0 (new) to 1 (dead). Normally, L=0.2 is used as a measure for end of useful life, resulting in 80% of the nominal capacity [72, 73]. L is calculated using the aging model in three stages: the first stage (L1) accounts for swing in SoC, the second one (L2) 28 accounts for average SoC, and the third stage (L3) accounts for temperature effects. They are given by [69]: (SoCdev−1)Tref/Kex/TB L1 = Kco. N. e (2.9) KSoC(SoCavg−0.5) L2 = L1. e . (1 − Li) (2.10) Tref KT(TB−Tref) L3 = L2. e . (2.11) TB The accumulative change in life aging parameter after M time intervals is given in [69]: M L(M) = ∑ L3(m) (2.12) m=1 All the empirical constants Kco, Kex, KSoC, and KT are battery specific, and must be obtained from test data. A battery life test was conducted in [71]. The values of battery aging model constants are tabulated in Table 2.3. The State-of-Health (SoH) of a battery is related to aging by [74]: L SoH = 1 − (2.13) 0.2 which decreases over time. SoH is equal to 1 at the beginning-of-life (BoL) and equal to 0 at end of life (EoL), i.e. L=0.2. With the help of the above validated models, a detailed simulation has been carried out for battery life estimation of the fuel cell-battery hybrid DERs for the microgrid operation. A sample SoH plot is shown in Figure 2.7, illustrating how Batt1 reaches its EoL before Batt2. 29 Table 2.2: Battery aging model constants Constant Name Value -5 Kco Coefficient of throughput 3.66 x 10 Kex Exponent of DoD 0.717 KSoC Coefficient of SoCavg 0.916 KT Thermal coefficient 0.0693 o Tref Ref battery temperature 35 C 1 Batt1 0.8 Batt2 0.6 0.4 0.2 State-of-Health SoH State-of-Health 0 0 5000 10000 15000 20000 25000 Life (cycles) Figure 2.7: Sample state-of-health (SoH) curves 2.5.2 Supercapacitor Aging Model Although the cycle life of a supercapacitor is relatively longer when compared with the other energy storage devices, its aging estimation and cycle life prediction cannot be neglected [75]. Supercapacitors are generally capable of enduring several thousands of deep charge and discharge cycles due to the absence of chemical reactions at the electrodes. Hence, cycling per se does not solely affect the supercapacitor life. Instead, the temperature and cell voltage are the key factors when it comes to life estimation of supercapacitors. 30 At increased temperature and cell voltage, the aging processes are accelerated because of higher reactivity of the chemical components. These effects result in a decrease of the capacitance and an increase of internal resistance [54]. The supercapacitor aging model employed is given by [75]: 100 − 퐶퐸표퐿 N퐸표퐿 = [ 푇−푇푟푒푓 ] (2.14) ( ) 훼 10 푑푇푟푒푓 where NEoL is the predicted cycle life, CEoL is the capacitance retention at the end-of-life, T is the temperature of the supercapacitor, Tref is the reference temperature, dTref is the degradation ratio at the reference temperature, and α is the acceleration factor. Equation (2.14) can be rewritten to find the capacitance retention CT as a function of operating cycles N, as given in (2.15). T−Tref ( ) α 10 (2.15) CT = 100 − dTref √N The supercapacitor reaches its end-of-life (EoL) when its capacitance falls below 80% of the initial value. Therefore, CEoL is set to 80% in (2.14) [54]. All the supercapacitor aging model constants are tabulated in Table 2.3. Figure 2.8 shows sample capacitance retention curves for two supercapacitors having different temperatures. In this figure, SC1 operating at a higher temperature reaches its EoL before SC2, which operates at a lower temperature. 31 100 SC1 SC2 95 90 85 Capacitance Retention (%) 80 0 10000 20000 30000 40000 N (cycles) Figure 2.8: Sample capacitance retention curves Table 2.3: Supercapacitor aging model constants Constant Value CEoL 80% o Tref 40 C dTref 0.08 α 1.2 2.6 Summary This chapter discussed the two main roles that energy storage systems (ESS) play in modern microgrids, viz. to supply energy for short-term applications and to supply energy for long-term applications. The differences between these two roles and the energy storage technologies that suit each of them have been summarized. This chapter has further classified various applications for ESS in microgrids according to their benefits. The aging of energy storage devices has been discussed in detail, and aging models for various types of ESS have been presented. In the next chapter, a generalized method is proposed for ESS sizing in microgrids. It uses the load profile of the specific application 32 and calculates the ESS size tailored to it, taking into account operating temperature and aging factors. 33 Chapter 3 SIZING ENERGY STORAGE SYSTEMS The energy storage systems (ESS) play two main roles in modern microgrids (i) to supply energy for short-term requirements (i.e. power applications), and (ii) ~ for long- term requirements (i.e. energy applications). Section 2.2 discussed two main benefits for ESS in microgrids, viz., frequency-responsive spinning reserve and peak shaving. The ESS capacity is dissimilar for different applications. Therefore, an exact and accurate ESS sizing method is necessary to maximize the reliability and lower the cost of microgrid planning and design. In general, it is not possible to find one single energy storage unit, e.g., battery cell or supercapacitor cell, which has enough capacity and capability for power or energy applications. However, energy storage modules or packs are fabricated and utilized. An energy storage module comprises several identical cells connected in series and parallel combinations. Terminal voltage of such a module is the voltage of a single cell multiplied by the number of battery cells connected in series. The ampere-hour capacity of a series- connected string of cells is the same as that of an individual cell. If the required ampere- hour capacity is more than the capacity of the available module, then multiple strings with equal number of cells must be connected in parallel to expand the capacity. Thus, 34 the overall ampere-hour capacity of the module is the capacity of a single string multiplied by the number of strings connected in parallel. The IEEE Standard 485-2010 Recommended Practice for Sizing Lead-Acid Batteries for Stationary Applications offers a guideline to size a battery module [76]. However, its scope is not applicable to the entire spectrum of residential, commercial, and industrial loads. Hence, this dissertation suggests a generalized approach for energy storage sizing in a microgrid. In Section 3.1, a sizing practice for battery is introduced for long-term applications; Section 3.2 presents the sizing method for supercapacitor as short-term energy storage. 3.1 Energy Storage Sizing for Long-term Applications 3.1.1 Determine the Duty Cycle The first step in sizing ESS for long-term applications, e.g. peak shaving, is to find the duty cycle of that particular application. The IEEE Standard 485-2010 defines duty cycle as “the sequence of loads a battery is expected to supply for specific time periods” [76]. Duty cycle could be different for different applications [77]. It is determined based on the difference between demand and supply. When the resulting duty cycles are compared with the typical duty cycle described in IEEE Standard 485-2010, it has been found that the IEEE Standard does not sufficiently address such a practical scenario. In fact, the current IEEE Standard has a narrow view of the duty cycle definition. As there are periods of time in which the load is constant, and load increases or decreases in step 35 changes with long steady state periods in between. This is not always true, especially in industrial power systems. Moreover, the IEEE Standard 485-2010 determines the duty cycle in amperes, which can be inaccurate in a microgrid where electric power is converted between ac and dc in multiple stages. Further, the cell sizing worksheet of IEEE Standard 485-2010 requires the user to calculate the current levels at each section of the duty cycle. However, in practice, the current level is different when measured at different points in the microgrid. For instance, the dc current drawn by the battery is not the same as dc current at dc link, because the duty cycle of a dc/dc converter and the modulation index of a dc/ac inverter are changing based on the battery pack voltage, which is not constant. This dissertation proposes a solution to this issue by determining the duty cycle in watts (i.e. power). It makes the total energy in watt-hours applicable to both dc and ac sides of the microgrid. 3.1.2 Calculate the Required Overall Energy After determining the duty cycle of a specific application, total energy can be calculated by integrating the power duty cycles. The outcome will be the amount of energy required by the battery pack to supply the load for one full duty cycle, taking into account the energy transformation efficiency of the ESS (ηESS). However, it is sensible to include a capacity margin to allow for unforeseen additions to the load demand and losses in the system. A method of providing this design margin is to add 10% to the total energy determined by (3.1). 36 10% 24 hours Ereq(Wh) = ∫ (Pdemand − Psupply)dt ηESS 0 (3.1) for Pdemand ≥ Psupply 3.1.3 Determine Required and Available Capacity After calculating the required energy (in Wh) that takes into account the design margin, the required capacity in ampere-hours can be determined as shown in (3.2). The available capacity of one battery cell can be calculated from (3.3). Ereq(Wh) Capreq(Ah) = (3.2) Vcell(V) × N(cells) Batt Pack Cap(Ah) = Nparallel × Capper−cell(Ah) (3.3) where Ereq is the required energy calculated from the previous subsection, Vcell is the terminal cell voltage, N is the number of batteries in the module (in each pack, if it is assumed to have more than one string in parallel), Capper-cell is the nominal capacity in ampere-hour of the battery cell, and Nparallel is the number of strings connected in parallel within the battery pack. It should be noted that every string connected in parallel has the same number of cells. 3.1.4 Calculate the Battery Pack Size The required size of the battery pack can be calculated by equating (3.2) and (3.3). The resulting battery size is given in (3.4). Ereq(Wh) N(cell) = (3.4) Vcell(V) × Nparallel × Capnom(Ah) 37 3.1.5 Estimate the Battery Cycle Life It is well known that due to repeated cycling the capacity of battery decreases. The fading in the deliverable capacity is called battery aging [78, 79]. The United States Advanced Battery Consortium (USABC) recommends replacing the battery when its capacity fades below 80% of its nominal value [72, 73]. Therefore, to make sure that the battery meets its designed loads until its end-of-life, the battery’s size must be 125% of the determined number of cells. The IEEE Standard 485-2010 for battery sizing has simplified the representation of the battery life to only 125%, which limits the user’s ability to set their own expected cycle life for the battery in their systems. The cycle life of any battery depends generally on the storage technology and operating conditions. Discharge of a huge part of energy (deep discharge) is much more damaging to the battery as compared to discharging a small part (shallow discharge). The relation between the battery cycle life to the depth-of-discharge (DoD) is described in various publications (e.g., [80], [81]) and given by battery manufacturers [71]. Mathematically, the relationship between cycle life and DoD can be described by (3.5), where α and β are battery-specific parameters. β Ncycle = α × ∆DoD (3.5) The use of (3.5) for an acceptable range of value of α and β [82, 83] gives ~1500 cycles (i.e., ~3.5 years) for 100% DoD and ~2000 cycles (i.e., ~5 years) for 80% DoD, respectively. But what if the microgrid needs to run for a duration that exceeds these numbers? One can replace the energy storage at once after they have reached their end- of-life (EoL). The better solution, however, is to take the required operating duration into 38 consideration when sizing the energy storage. Hence, this dissertation introduces a new parameter based on (3.5), Klife, as tabulated in Table 3.1. Table 3.1: Target Cycle Life with the Corresponding Klife Klife Cycle Life Range 1 1300-1500 1.11 1500-1800 1.25 1800-2250 1.43 2250-3000 1.67 3000-4000 2 4000-5700 2.5 5700-9000 3.33 9000-17000 5 17000-42000 3.1.6 Additional Considerations Another consideration that should be taken into account in sizing batteries is the operating temperature, since it directly affects the deliverable capacity of a battery. The standard temperature for rating battery capacity in U.S. is 25°C [76]. If the operating temperature is not the standard temperature, then a temperature correcting factor KT needs to be considered. The values of KT are given in the IEEE Standard 485-2010. Equation (3.4) can be rewritten as shown in (3.6). It should be noted that the presented sizing concept in this dissertation is also applicable to other microgrid configurations. Ereq(Wh) × Klife × KT N(cell) = (3.6) Vcell(V) × Nparallel × Capnom(Ah) 39 3.2 Energy Storage Sizing for Short-term Applications The isolated microgrid, especially when providing energy for an industrial load, is subjected to fluctuating demand. Such a load fluctuation can cause the system’s frequency to drop. Furthermore, things can get worse, if a huge step change in demand takes place. In this case, the practical limitations of maximum mechanical torque causes the generator to stall5; thereby leading to a collapse of the microgrid’s frequency [84]. To avoid a frequency collapse scenario, frequency-responsive spinning reserve is needed to give power to the grid immediately, when a large increase in the load occurs. A good ESS candidate for this spinning reserve is supercapacitors as discussed in the Subsection 2.3.2. Since no published standard is available yet for the supercapacitors, the following method of sizing supercapacitors is proposed. The required capacitance of the supercapacitor (in Farads) can be related to the required power, duration, and voltage, as given in (3.7). As the designers know the required power and time duration, it is always safer to assume the worst-case scenario of a step change from zero to the rated power level. To calculate (3.7), the user has to pick a specific supercapacitor to find its voltage from the manufacturer’s datasheet. After the value of required capacitance from (3.7) has been calculated, the number of cells can be determined using (3.8). By plugging the numbers from the adopted microgrid and supercapacitor, the required size of supercapacitors is found. 5 For an engine driven generator, stalling is said to occur if a sudden change in load or overloading brings the generator to an abrupt halt. 40 Ereq(J) Preq(W) × treq(s) Creq(F) = 2 2 = 2 2 (3.7) VSC(V) VSC(V) Creq(F) n(cells) = (3.8) CSC(F) 3.3 Case Study: Sizing ESS for Residential/Commercial/Industrial Loads To demonstrate the proposed ESS sizing methods discussed in Sections 3.1 and 3.2, a microgrid test bed has been chosen as shown in Figure 3.1. The isolated microgrid comprises DERs and ESS for supplying power to a load. A 100-kW MTU Onsite Energy natural gas powered generator set controlled by an electronic isochronous speed governor is connected in parallel with a solar/photovoltaic (PV) system. The detailed modeling of internal combustion engine for the application of a natural gas genset was developed in [85]. The PV system consists of 50 x 410-W tenKsolar PV arrays [86], whose output power is shown in Figure 3.2. Power generation profile of the PV system has been obtained from [87]. As seen from Figure 3.2, it is clear that the solar energy is only available between 4:00AM and 7:30PM, and its peak occurs around the noon time. It is noteworthy that the ESS includes lead-acid battery and supercapacitor. Table 2 gives the specifications of commercial lead-acid battery to be used for the purposes of sizing in this section. 41 Natural Gas Genset 480V, 100kW Load Residential/ Solar/Photovoltaic (PV) Commercial/ Arrays 50 x 410W Industrial Battery Supercapacitor Figure 3.1: Schematic diagram of a microgrid test bed for peak shaving application showing the various distributed energy resources and energy storage options 20 15 10 Power Power (kW) Power (kW) Power 5 0 12AM 6AM 12PM 6PM 12AM Time Figure 3.2: Typical daily power output for Photovoltaic 42 Table 3.2: Specifications of Lead-acid Battery Specification Value Brand MARATHON Model M12V155FT Nominal Voltage 12 V Capacity 155 Ah Weight 119 lbs. Table 3.3. Specifications of Supercapacitor Specification Value Brand MAXWELL Model BCAP3400 P285 K05 Nominal Voltage 2.85 V Capacity 3400 F This case study features two applications; (i) for peak shaving using the lead-acid battery, and (ii) for sizing the supercapacitor toward frequency regulation/spinning reserve. The sizing of lead-acid battery for peak shaving is considered first. To find the duty cycles for residential, commercial or industrial loads, their corresponding typical daily load profiles are considered. For instance, the demand curve (red) in Figure 3.3 shows a typical daily residential load profile of a neighborhood consisting of 20 houses. The supply curve (green) is the available power in the microgrid, supplied from generator and PV system. Figure 3.3 also indicates that there is limited supply available during residential peak hours between 6PM and 11PM. The difference between demand and supply is the basis for determination of a residential duty cycle, see hatched area. Figure 3.4 shows the typical daily commercial profile where the peak-ON hours are usual business hours 43 between 8AM and 7PM. Likewise, the difference between demand and supply, which constructs the duty cycle of the commercial load, is hatched area in Figure 3.4. Figure 3.5 displays a large and fluctuating load profile of an industrial application. The load profile of Steel Mill Thüringen GmbH was used for this study [88]. Rolling mills are generally used to change the width of steel [89]. A typical steel rolling mill is expected to work continuously for various dimensions of steel. The variations in steel sizes and dimensions need the plant power source to have high energy and power capability [90]. This results in a very harsh and fast varying duty cycle. It must be noted that all the duty cycles mentioned are supplied when the microgrid is isolated from the main grid. The required energy Ereq for the residential, commercial, and industrial loads are 146 kWh, 483 kWh, and 562 kWh, respectively. By applying the described sizing procedure, different sizes of battery modules were obtained in Table 3.4 assuming default KT =1 and various values of Klife. Table 3.4: Battery Sizes (in number of cells) for Peak-shaving of Residential, Commercial, and Industrial Applications Lifetime (years) 3.8 4.5 5.5 7.2 9.6 Klife 1 1.11 1.25 1.43 1.67 Residential 79 88 99 113 132 Commercial 260 289 325 372 435 Industrial 303 337 379 434 507 44 150 Supply Demand 100 ) Power Power (kW) Power (kW Power 50 0 12AM 6AM 12PM 6PM 12AM TimeTime Figure 3.3: Typical daily residential loadsupply and demand curves with required energy for battery (hatched) Supply 150 Demand ) 100 Power Power (kW) Power (kW Power 50 0 12AM 6AM 12PM 6PM 12AM Time Figure 3.4: Typical daily commercial load—supply and demand curves with required energy for battery (hatched) 45 Supply 300 Demand 250 ) 200 150 Power Power (kW) Power (kW Power 100 50 0 12AM 6AM 12PM 6PM 12AM TimeTime Figure 3.5: Typical daily industrial load—supply and demand curves with required energy for battery (hatched) After accomplishing the battery sizing for the peak shaving, the supercapacitor is designed for frequency regulation/spinning reserve in the microgrid. The isolated microgrid, especially when providing energy for an industrial load, is subjected to fluctuating load demand (cf., Figure 3.5). Such a load fluctuation can cause the system’s frequency to drop. Assuming an 80-kW step increase in load of the genset shown in Figure 3.1, the speed of the generator and so the microgrid’s frequency drop [85]. Figure 3.6a shows the frequency response for microgrid when a large step change in load occurs. It is observed that the frequency drops from 60 Hz to 50 Hz, in the transient, until the genset is able to increase its power output. Furthermore, things can get worse, if a 90-kW step load increase is required. In this case, the genset stalls, thus leading to a collapse of the microgrid’s frequency as shown in Figure 3.6b; this leads to shut down for the entire microgrid [84]. To avoid a frequency 46 collapse scenario, ESS is needed to give power to the grid whenever a large increase in the load occurs. A good candidate for this frequency-responsive spinning reserve is supercapacitors. The required power and time duration selected are a step change from zero to full rated power of the genset, i.e., 100kW (c.f. Figure 3.1), and a required response time of 2 seconds. The supercapacitor considered is Maxwell BCAP3400 supercapacitor, whose parameters are tabulated in Table 3.3. By plugging the numbers from the adopted microgrid and supercapacitor in (3.7) and (3.8), the required size of supercapacitor was found to be 15 cells, see Table 3.5. After the supercapacitor is designed, the frequency regulation response provided the microgrid with a rapid back-up power as a bridge to support operations until the genset’s output rises, see Figure 3.7. It is noteworthy that the generator’s frequency gets restored back to 60 Hz relatively slowly (in about 100 seconds). 47 60 60 50 55 40 Frequency (Hz) Frequency (Hz) 50 30 0 1 2 3 4 0 1 2 3 4 Time (s) Time (s) (a) (b) Figure 3.6: Frequency of microgrid when (a) 80kW step load is applied and (b) 90kW step load is applied causing stalling of generator 60 60 55 59 Spinning reserve 0.2 0.6 1 response Frequency (Hz) 50 0 1 2 3 4 5 Time (s) Figure 3.7: Frequency of microgrid with supercapacitor is present as the spinning reserve when 90kW step load is applied Table 3.5: ESS Sizes (in number of cells) for frequency regulation/spinning reserve Applications ESS Technology Model No. Number of cells Supercapacitor MAXWELL BCAP3400 15 48 3.4 Summary In this chapter, a generalized method was proposed for ESS sizing for microgrids. The sizing methodology for long-term peak shaving applications used the specific load profile of the targeted industry, and calculated the battery size by taking into account operating temperature and aging factors. Furthermore, a new supercapacitor sizing methodology for short-term spinning reserve applications was presented. The proposed sizing methodology is used in the later chapters of this dissertation to design the ESS for industrial application. It is used to emphasize the problem of ESS cycle life discrepancies in microgrids. A solution to this problem is proposed in the next chapter using the novel framework called Flexible Distribution of EneRgy and Storage resources (FDERS). 49 Chapter 4 FLEXIBLE DISTRIBUTION OF ENERGY AND STORAGE RESOURCES (FDERS) As defined earlier in this dissertation, microgrids consist of several DERs together with loads and energy storage systems (ESS). The various kinds of primary energy sources employed include dispatchable as well as non-dispatchable resources, such as internal combustion engines, microturbines, photovoltaics, wind generators, and fuel cells [31]. Among them, even the dispatchables like fuel cells have a long start-up time and poor transient response due to their inherent characteristics [21]. As such, they are supplemented with ESS to meet the load demand under islanded-mode of operation, because the fuel cell falls short of supplying independently the transient load demand. Since ESS can offer rapid transient response with a higher specific power capacity compared to that of fuel cells, a combination of fuel cells with ESS is found to be helpful for many applications [91, 92]. The transient response of a typical fuel cell-energy storage hybrid DER is shown in Figure 4.1, where PL and PFC denote the step change in load power demand and the fuel cell response, respectively [93]. The difference between the two responses PL and PFC causes charging and discharging cycles in ESS. As discussed in Section 2.5, a repeated cycling of ESS causes the deliverable capacity of ESS to decrease. This capacity fade phenomenon is called aging 50 [78, 79]. Moreover, a power fade happens when the internal resistance of the cell increases [94]. Cycle life refers to the number of times a battery is cycled before the cell capacity fades below 80% of its nominal value according to the United States Advanced Battery Consortium (USABC) [72, 73]. Batteries contain heavy metals including lithium, mercury, lead, cadmium, and nickel. Such metals can be harmful to the environment when batteries are improperly disposed of [95]. The waste management of batteries must comply with regulations like the US Mercury-Containing and Rechargeable Battery Act of 1996 and the EU Directives 91/156 and 91/689 [96]. It is therefore expensive to employ batteries in many applications. For sustainable development, extending the battery life has become a critical concern. In this dissertation, a microgrid consisting of several small-rated fuel cell-battery hybrid DERs is considered for supplying a crusher-conveyor load when power from the main grid is not available. The energy storage life can be improved by creating a cooperative framework known as Flexible Distribution of EneRgy and Storage Resources (FDERS) [29]. This chapter introduces FDERS concept and the later chapters demonstrate how to use FDERS in improving cycle life of ESS. 51 Figure 4.1: Transient dynamic response of a fuel cell-battery hybrid system 4.1 Inspiration for FDERS FDERS was inspired by the V-shape formation of a flock of birds [97, 98] and peloton/echelon formation of a cycling racing team [99-101], see Figure 4.2. The front (leading) rider creates a slipstream and the others take advantage of it and aerodynamically reduce their wind resistance by drafting in the rear [100, 101]. The pecking order is determined based on their relative strength at any time. In a team of equals, a periodic rotation of positions helps in reinvigorating all members. However, in case their positions are unalterable over the entire journey, the leading bird/cyclist gets exhausted sooner than the drafting counterparts do. A network of smaller-rated DERs in a microgrid may be generally viewed as a fixed formation. In contrast, this dissertation proposes a flexible distribution of energy and storage resources for achieving increased resource lifetime, reduced energy storage requirement, enhanced controllability, and improved system robustness, particularly 52 when supplying the large and fluctuating loads commonly found at industrial sites when operating in the islanded mode. In this dissertation, similar cooperative strategies are evaluated for balancing the energy storage cycle life in a parallel configuration of small- rated fuel cell-energy storage hybrid DERs supplying a large and fluctuating crusher- conveyor load, especially when there is no power available from the main (utility) grid. Figure 4.2: Energy saving formations 53 4.2 State-of-the-Art in Integration of Small-Scale Distributed Energy Resources (DERs) The integration of DERs in a microgrid has been an active research topic in the recent past [21, 24-28]. In order to facilitate power sharing with decentralized controls and plug- and-play operation in the microgrid [21], every kth (k = 1, 2, …, n) DER is regulated by frequency/active power (ω/P) and voltage/reactive power (V/Q) [102, 103] droop controls, as shown in Figure 4.3. The behavior of steady-state active power sharing between the interconnected DERs in a microgrid is straightforward, and more or less identical to that between interconnected large rotating machine generators in a power system [102]. It can be calculated from the steady-state frequency droop curves of the various interconnected DERs. Normally, these droop characteristics are designed so that interconnected DERs under islanded conditions share the total power among themselves in a manner proportionate to their ratings. Moreover, each DER’s steady-state power share is independent from its locational placement within the microgrid. The dynamic behavior of a DER, on the contrary, is highly influenced by its ‘electrical’ locational placement within the microgrid as well as its controller design [103-105]. This is of particular concern to industrial power distribution systems at metal and mining industries containing large and fluctuating loads [89], especially if such loads have to be supplied from a network of multiple smaller-rated DERs in conjunction with local energy storage resources. 54 Figure 4.3: DER outer loop power controller block diagram (k = 1, 2, … , n) 4.3 Qualitative Evaluation In this section, the Quality Function Deployment (QFD) methodology is employed to evaluate the various features of FDERS. The QFD method employs a means of translating the demanded quality outcomes (“What’s” or customer requirements) into the provision for the demanded quality (“How’s” or design variables of the quality team) [106, 107]. QFD works through a set of matrices to quantify customer requirements (“What’s”) and engineering characteristics/technical descriptors (“How’s”), and for this reason it is also called the House of Quality. The various “What’s” and “How’s” along with the strengths of relationships between them for the FDERS system are tabulated in Table 4.1. A relative importance/weight of each technical descriptive toward production is found by simple algorithm that correlates “What’s” to “How’s”. The specific values for strengths of relationships used in the QFD are described in Table 4.2. 55 Table 4.1: Qualitative evaluation of FDERS design parameters Design Parameters (How’s) Reactance 5) - Customer Inertia Requirements (What’s) (1 Importance Physical Virtual Droop Frequency Virtual Resource Lifetime 5 3 3 9 9 Energy Storage Deployment 5 9 3 -3 3 Improve Controllability 4 -9 3 3 3 Robustness 3 -1 -1 3 -1 Cost 5 -9 -1 -1 -1 Reduce Power Loss 3 -1 3 -1 3 Complexity 5 3 -3 -3 -3 Maintain Steady State 4 -1 -1 -9 0 Absolute Weight -16 24 -8 58 Unlike the quantitative evaluation, the qualitative evaluation offers a relative assessment of some of the technical metrics towards meeting the customer’s requirements. The QFD of this chapter shows that by controlling the virtual reactance and virtual inertia, the customer needs can be met quite easily. Also, the house of quality shows that it is not feasible to implement FDERS by using physical reactances. This is mainly because of cost issues and other technicalities. Furthermore, a simple change in the droop gain gets a significant negative weight although it has an influence because it also affects the steady-state load sharing among various FDERS constituents (that is not desired). 56 Table 4.2: Relationships between “What’s” and “How’s” Relation Value Strong Positive +9 Medium Positive +3 Weak Positive +1 Neutral 0 Weak Negative -1 Medium Negative -3 Strong Negative -9 The results of this QFD made it clear which design parameters need to be focused on in the research [106]. To achieve this clarity, there was a need to assign an importance factor for each of the customer requirements and to compute the absolute weights for every design parameter. As seen in Table 4.1, the design parameters of virtual reactance and virtual inertia have resulted in a positive absolute weight. Of these two, the virtual inertia has the largest impact on meeting all the customer requirements. However, from several case studies, it has been observed that a combination of the virtual inertia and virtual reactance would give the best results. 4.4 Synthesis of Virtual Reactance and Virtual Inertia FDERS transforms a fixed electrical power network into a flexible one for achieving potential savings in microgrids. This flexibility is achieved by controlling the ‘electrical’ locational placements of DERs within the microgrid. It is realized, as shown in Figure 4.4, with synthesized ‘virtual’ reactances, Xk-add, which can be accomplished in ∗ ′∗ the DER controller by modifying its voltage reference from 푣⃗ (푡) to 푣⃗ (푡). As a result, 57 th the net reactance between the k DER and load will be Xk = Xk0 + Xk-add -- that can be smoothly adjusted. In this way, the pecking order of DER positions as seen from the load can be effortlessly changed. Though the concept of virtual reactance has been applied earlier by several authors for other purposes, it has been employed in the FDERS with a different intent: to facilitate a change of the pecking order or hierarchy among the interconnected DERS. At the instant of load change, the pecking order is determined by the ascending order of Xk’s. However, the only way to make the initially leading DERS (that has a low Xk) remain in the leading position for a sustained period is by also implanting a larger virtual inertia through the time-constant parameter Tpk > 0 in the active power controller (cf. Figure 4.3). Figure 4.4: Synthesis of variable reactance in the DER’s controller The control of PWM inverter-based DER units is carried out using a multi-loop regulator whose outer loop is as shown in Figure 4.4 and whose inner loop consists of voltage regulator for the inverter output. The bandwidth of inner loop voltage regulator is limited by the inverter switching frequency. It was observed, as experimental tests showed in 58 [103], that an inverter operating at 4-kHz switching frequency had limited the bandwidth of inner loop voltage regulator to around 200 rad.s-1. Accordingly, the transfer function ∗ between commanded voltage (푉푘 ) and actual voltage (푉푘) in Figure 4.4 (inset) can be -1 approximated as a first-order lag function with time-constant Tv = (200) s = 5ms. Also, a lead-lag transfer function was included to compensate for any delays in the current measurement. 4.5 Benefits of FDERS This dissertation describes multiple benefits of employing FDERS in microgrids. An important benefit is battery cycle life balancing in a microgrid consisting of multiple fuel cell-battery hybrid DERs supplying fast-varying load. These DERs are ‘electrically’ displaced by unequal reactances which leads to unequal utilization of batteries in these DERs. This part is covered in Chapter 5. Another benefit of applying FDERS is extending the operation of microgrids consisting of multiple DERs that have different prime-movers as presented in Chapter 6. FDERS is used to shift the entire burden of the load from the leading DER once its battery reached the end of life to the drafting DER. Furthermore, Chapter 6 also describes how FDERS can be employed for different technologies of ESS. Finally, Chapter 7 shows how the concept of FDERS can be advantageous in fleet vehicle-to-grid (V2G) system to balance and extend the cycle life of batteries in fleet electric vehicles. 59 Chapter 5 CYCLE LIFE IMPROVEMENT IN ENERGY STORAGE SYSTEMS This chapter demonstrates the FDERS to achieve the important benefit of balancing of cycle life for all ESS in the microgrid. Its content has been earlier published in the Journal of Power Sources [108] and was presented in IEEE Industrial & Commercial Power Systems Technical Conference (I&CPS) [109]. This chapter starts with a system description for the adopted microgrid. Then the problem of the discrepancy in the cycle life of ESS is described in detail. Several FDERS-based solutions are proposed to solve the problem. 5.1 System Description The microgrid system for a crusher-conveyor load comprises four small-rated fuel cell- battery hybrid DERs in an industrial plant as illustrated by Figure 5.1. Numerous publications have reported the extremely demanding load profiles observed at industrial sites [110-112]; A crusher-conveyor load profile, tabulated in Table 5.1, has been taken from [110]. The microgrid consists of four fuel cell-battery hybrid DERs, each including a Solid Oxide Fuel Cell (SOFC) along with an A123 Li-ion battery pack ANR26650M1A [113]. Fuel cells are considered to be one of the most promising alternative energy 60 resources due to their high energy density, high efficiency, zero or low emission, and flexible/modular structure [114]. Energy is produced in a fuel cell by oxidizing (O2) the - hydrogen gas (H2). The resulting ions (OH ) from the reaction at the anode fuse with the oxygen to become water (H2O), and the resulting electrons form an electrical field between the anode and cathode through the external circuit, see Figure 5.2 [115]. Despite the unique advantage that fuel cells offer, they have slower response time to step changes in load demand due to their inherent characteristics. This dissertation builds on a basic SOFC power dynamic model developed in [116] together with necessary enhancements for simulating the system’s control strategies [117]. Figure 5.1: Single-line diagram of a microgrid consisting of four fuel cell-battery hybrid DERs A block diagram illustrating the power flows and controls in each fuel cell-hybrid DER is displayed in Figure 5.3. As seen in this figure, a power management system regulates the fuel flow rate of fuel cell stack to meet the expected load demand [118]. The power 61 management system is aimed at splitting the power flow between the SOFC and Li-ion battery for sustainably meeting the requirements [92]. Moreover, it maximizes the fuel cell generation output to recharge the battery and cover parasitic losses of items such as pumps, fans, blowers and heat exchangers [119]. In addition, the power management system maintains a fast recharge of the Li-ion battery at a constant rate of 5 C-rate. As a result, the battery would eventually become fully charged in steady state when the fuel cell fulfills the load demand. It is must be noted that the SOFC is connected to dc-link through a unidirectional dc/dc converter, unlike the Li-ion battery that is connected to a bi-directional dc/dc converter to allow charging and discharging. Table 5.1. Motor loading profiles for a 3-phase, 480-V crusher-conveyor Recurring Cycles Time RMS Current Active Power (PL) 20 sec 424 A 300kW 120 sec 56 A 34 kW 20 sec 424 A 300 kW 120 sec 56 A 34 kW 20 sec 424 A 300 kW 120 sec 56 A 34 kW 20 sec 424 A 300 kW 120 min 56 A 34 kW 62 Figure 5.2: Basic hydrogen fuel cell Figure 5.3: Power flow and control structure of each fuel cell-battery hybrid DER To enable power sharing with decentralized controls of power electronics-based DERs in the microgrid [21], each DER unit is regulated with frequency/active power (/P) and voltage/reactive power (V/Q) droop controls, as shown before in Figure 4.3. The behavior of steady-state active power sharing between the interconnected DER units in a 63 microgrid and can be determined from the steady-state frequency droop curves of the interconnected DER units. Normally, these droop characteristics are designed in such way that interconnected DERs under islanded conditions share the total power among themselves in a manner proportionate with their ratings. Moreover, each DER’s steady- state power share is independent of its locational placement within the microgrid. It is to be noted that the reactive power control has no bearing on the battery lifetime/aging studies conducted in this dissertation, and hence the details are omitted. The DER units are ‘electrically’ displaced by unequal reactances with the DER1 being the ‘electrically’ closest as depicted in Figure 5.1 (i.e. X1 < X2 < X3 < X4). Such reactances are caused by the transformers, line filters, distribution lines, etc. For analyzing a more realistic system, DER units of unequal ratings were assumed. Each DER’s battery is sized according to its rated power. The DERs design parameters are given in Table 5.2, and their interface reactances Xk included practical considerations of X/R = 10%. Such microgrid models have been earlier experimentally validated in laboratory [103] and on CERTS Microgrid field test bed [120]. The circuit details of DER components like power electronic dc/ac inverters, filters, and transformers can be found in [121-123]. 64 Table 5.2: DER ratings and design parameters Rated Power X k k Battery Size DER k q p k (kW) (Ω) (V kVAR-1) (rad kJ-1) (cells) 1 30 0 0.01 0.0157 28 2 60 0.18 0.01 0.0079 56 3 90 0.27 0.01 0.0052 84 4 120 0.46 0.01 0.0039 112 5.2 Problem Identification A detailed and dynamic model for the four-DER parallel connected microgrid has been developed in MATLAB/ Simulink with the power network simulated in SimPowerSystems toolbox. The simulation is conducted in two stages, viz., (i) time- domain simulation and (ii) offline calculations to estimate battery life, as illustrated in Figure 5.4. First, the time-domain simulation of green colored blocks is carried out iteratively for the entire time. The generation output of each individual DER is computed when the system is subjected to the application load profile of a crusher-conveyor. This power is shared between the fuel cell and battery components. Later, the battery aging model, displayed in blue dashed lines in Figure 5.4, is run offline for estimating the battery life using the SoC and temperature data obtained from time-domain simulation over several cycles. A brief description of the utilized component models within each fuel cell-battery hybrid DER is described in this section. Upon microgrid system simulation, it was observed that the dynamic responses of the four fuel cell-battery hybrid DERs to the crusher-conveyor load are varied. Individual units’ responses are highly influenced by their relative ‘electrical’ locational placements 65 within the microgrid. This is due to the differing underlying interface reactances between the DERs and the crusher-conveyor load [29]. Ideally, it is desired for each DER’s power profile to follow the change in the load abruptly without any delays, as shown in Figure 5.5(a). This case is hypothetical since the current cannot change instantaneously across the reactances embedded in the system (cf. Figure 5.1). It will, therefore, be considered as an “ideal” case throughout the chapter, and the battery aging results for this case indicate that all the batteries age at an identical rate that they reach EoL after 21664 cycles as illustrated in Figure 5.5(b). The above ideal case is not a realistic scenario in an interconnected power system. Therefore, a time-domain simulation was carried out in MATLAB/ Simulink with the power network simulated in SimPowerSystems toolbox for a worst case scenario, taking into account the various network reactances in the microgrid (cf. Figure 5.1). For this case, labeled as baseline, the smallest DER was considered closer to the crusher- conveyor. Figure 5.6 shows the dynamic response of the microgrid supplying the crusher- conveyor load. The battery aging results for this realistic case indicates that each DER’s battery reaches its EoL at a distinct time. 66 Figure 5.4. Block diagram of simulation strategy for microgrid operation and battery life estimation 120 1 DER1 Batt1 100 DER2 0.8 Batt2 80 DER3 Batt3 0.6 DER4 Batt4 60 0.4 40 Power (kW) 0.2 20 State-of-Health SoH 0 0 138 140 142 144 0 5000 10000 15000 20000 25000 Life (cycles) Time (s) (a) (b) Figure 5.5: Ideal case (a) power profile, and (b) battery state of health (SoH) 67 250 DER1 DER2 200 DER3 DER4 150 100 Power (kW) 50 0 140 140.2 140.4 140.6 140.8 Time (s) Figure 5.6: Power response characteristic As shown in Figure 5.6, the difference in the transient responses resulted in significant variations in the utilization of the Li-ion batteries in different DERs. The more ‘electrically’ closer a DER unit is to the load, the more stress is placed on its battery. This causes difference in SoC curves of the various batteries as shown in Figure 5.7(a). It is due to a greater discharge current, which implies a higher rate of energy conversion leading to heat losses [124]. This results in a rise in the battery temperature as shown in Figure 5.7(b), where the temperature of Battery 1 shot up to a peak value of 60oC from an ambient temperature of 35oC. According to the battery aging model, SoC and temperature are the two main factors affecting the rate of aging of the battery. By applying the SoC and temperature data obtained earlier from time-domain simulation of microgrid to an offline battery aging model also developed in MATLAB, the resulting aging factor for the four DER batteries was obtained as shown in Figure 5.7(c). This figure illustrates a definite imbalance in the accumulation of the aging factor. By calculating the accumulated age 68 over the time, the aging results depicted in Figure 5.7(d) show that significant variations in battery age occur for the four DERs. Assuming that each battery replacement is carried out after its EoL, it has been observed that Battery 1 reached EoL after 11627 cycles, which is almost one half the hypothetical estimated lifetime in the “ideal” case. Further, Battery 2, Battery 3, and Battery 4 survived until 16928 cycles, 19791 cycles, and 25106 cycles, respectively. However, it must be noted that if Battery 1 is not replaced after its EoL of 11627 cycles, the supply network to the crusher-conveyor can actually collapse within right after Battery 1 reached its EoL. This is because the remaining three surviving batteries themselves may not be able to meet the challenging load requirement, thereby leading to a series of cascading failures and eventually resulting in the shutdown of the entire industrial power system. 5.3 Strategies for Balancing the Battery Cycle Life This chapter presents three approaches to carry out the balancing of battery aging to varying degrees in the microgrid. These are approaches labeled as A, B, C, and D. 5.3.1 Periodic Cycling – Approach A In approach A, all the DERs are cycled periodically to change their ‘electrical’ positions according to the four orders shown in Table 5.3, which has the various positions considered for the four DERs in the context of the microgrid in Figure 5.1. In the first order, i.e. “Order I,” no virtual reactances are added to the system. Accordingly, the response is unchanged from the baseline response (cf. Figure 5.6). In “Order II,” it is DER2’s turn to take the lead and DER1’s turn to draft in the last position. Likewise, the 69 sequence continues for “Order III” and “Order IV” where DER3 and DER4 take turns to lead, respectively. The values of virtual reactances (Xk-add) necessary to achieve these four cycles are shown in Table 5.3, and the sequence repeats itself never ending. The cycling frequency in this approach was set to one per cycle, meaning that all these four orders will take place in 4 consecutive cycles. The power profiles of the four orders are shown in Figure 5.8. 100 60 Batt1 Batt1 98 Batt2 55 Batt2 C) Batt3 o 96 Batt3 Batt4 50 Batt4 94 91 90 92 89 45 Temperature ( 90 40 State of Charge SoC (%) 88 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) x 10-5 4 1 Batt1 Batt 1 Batt 2 0.8 Batt2 3 Batt 3 Batt3 Batt 4 0.6 Batt4 2 0.4 1 0.2 State-of-Health SoH Life Aging Parameter L 0 0 1 2 3 4 5 6 7 8 0 5000 10000 15000 20000 25000 Life (cycles) Cycles (c) (d) Figure 5.7: Baseline case (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH 70 A sample of the batteries SoC under this approach with “Order II” is shown in Figure 5.9(a). A sample of temperature profiles of four consecutive cycles is also shown in Figure 5.9(b). After a detailed battery aging analysis as described in the previous section, the life aging parameters and the estimated batteries lifetimes under Approach A were obtained as shown in Figure 5.9(c) and Figure 5.9(d), respectively. The life of Battery 1 was extended to 17237 cycles from 11627 cycles of baseline. However, this was at the cost of the life of Battery 4, which suffered from a loss of about 3000 cycles as compared to its baseline lifetime. Table 5.3. Settings of parameters for cycling Order Reactance, X(Ω) DER1 DER2 DER3 DER4 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0 0 0 0 I Total (Xk = Xk0+Xk-add) 0 0.18 0.27 0.46 Position 1 2 3 4 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0.46 -0.18 -0.09 -0.19 II Total (Xk = Xk0+Xk-add) 0.46 0 0.18 0.27 Position 4 1 2 3 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0.27 0.28 -0.27 -0.28 III Total (Xk = Xk0+Xk-add) 0.27 0.46 0 0.18 Position 3 4 1 2 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0.18 0.09 0.19 -0.46 IV Total (Xk = Xk0+Xk-add) 0.18 0.27 0.46 0 Position 2 3 4 1 71 300 300 DER1 DER1 DER2 DER2 DER3 DER3 200 200 DER4 DER4 Power (kW) 100 Power (kW) 100 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Time (s) Time (s) (a) (b) 300 300 DER1 DER1 DER2 DER2 DER3 DER3 200 200 DER4 DER4 Power (kW) 100 Power (kW) 100 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Time (s) Time (s) (c) (d) Figure 5.8: Power profiles of (a) Order I, (b) Order II, (c) Order III, and (d) Order IV 5.3.2 Power Rating-Weighted Cycling – Approach B Another approach that provided an improved battery life performance uses the ratio of number of cycles for each Order to be proportional to the ratio of the DER unit’s rated power. Accordingly, the cycles are determined as (number of cycles for “Order I”): (~ “Order II”) : (~ “Order III”) : (~ “Order IV”) = 30kW : 60kW : 90kW : 120kW which can be further simplified as equal to 1:2:3:4. In other words, “Order I” occurs for one cycle, then “Order II” takes place for the next two cycles, and the sequence goes on in the same 72 manner for “Order III” and “Order IV,” respectively. The Orders have been described earlier in Table 5.3. 100 60 Batt1 Batt1 98 Batt2 Batt2 55 Order II C) Batt3 o Order I Order III 96 Order IV Batt3 Batt4 50 Batt4 94 92 91 92 90 45 Temperature ( 90 40 State of Charge SoC (%) 88 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 Batt1 3 Batt 1 Batt 2 0.8 Batt2 Batt 3 Batt3 Batt 4 0.6 2 48% Batt4 0.4 1 0.2 State-of-Health SoH Life Aging Parameter L 0 0 1 2 3 4 5 6 7 8 0 5000 10000 15000 20000 25000 Life (cycles) Cycles (c) (d) Figure 5.9: Approach A (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH The simulation results under Approach B show that the batteries demonstrated relatively more balanced SoC and therefore, balanced temperature as well. A sample of both SoC and temperature is shown in Figure 5.10(a) and Figure 5.10(b), respectively. As a result, the aging rates were also more evenly distributed, as shown in Figure 5.10(c). Since this pattern repeats itself for the system operation, the aging rates are expected to be 73 maintained constant over time. The battery state of health (SoH) curves for this approach are shown in Figure 5.10(d). It is evident from these results that Approach B provides improved balancing of battery lifetimes in comparison to Approach A. In particular, the estimated lifetime of Battery 1 is 18561 cycles for Approach B as compared to 17237 cycles in Approach A. Moreover, Battery 2 and Battery 3 have also gained more life under this approach. On the down side, Battery 4 sacrifices some of its lifetime to the advantage of the other three batteries: the estimated lifetime of Battery 4 dropped by 2000 cycles in this approach relative to Approach A. 5.3.3 Power Rating-Levelized Cycling – Approach C A third notable approach that gave promising results is based on isolating batteries into two levels, the higher rated DERs and the lower rated ones, to cycle within the leading two positions and lagging two positions, respectively. Accordingly, the leading/first two ‘electrical’ positions, i.e. positions 1 and 2, are operated by the higher rated DER4 and DER3, respectively. Similarly, the lagging/last two ‘electrical’ positions, i.e. positions 3 and 4 are operated by the lower rated DER2 and DER1, respectively. For the purpose of this approach, “Order III” was borrowed from Table 5.3 as it represents the condition in which DER3 leads DER4 in the upper level whereas DER2 leads DER1 in the lower level. The opposite of that is a newly introduced order given as “Order V”. Table 5.4 gives the details of the parameters needed for these two orders. The cycling in the approach alternates at the rate of 2:1 to stress DER4 more than DER3 because it has larger rated power. In other words, “Order III” occurs for one cycle, then “Order V” takes 74 place for the next two cycles; and this sequence repeats itself with a total isolation between the two levels. 100 60 Batt1 Batt1 98 Batt2 Order II Batt2 55 Order II C) Order III Batt3 o 96 Order I Batt3 Batt4 Batt4 50 94 92 91 45 92 90 Temperature ( Sequence 90 40 continues State of Charge SoC (%) 88 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 Batt1 3 Batt 1 Batt 2 0.8 Batt2 Batt 3 Batt3 Batt 4 0.6 2 60% Batt4 0.4 1 0.2 State-of-Health SoH Life Aging Parameter L 0 0 1 2 3 4 5 6 7 8 0 5000 10000 15000 20000 25000 Life (cycles) Cycles (c) (d) Figure 5.10: Approach B (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH This concept of levelizing based on DERs rated power paid off very well in further converging SoC resulting in more homogenous temperatures, as shown in Figure 5.11(a) and Figure 5.11(b), respectively. Therefore, the gap between the aging rates of all batteries in the microgrid became smaller, as indicated by the battery life-aging parameter, shown in Figure 5.11(c). By computing the accumulative age over the cycles 75 under this approach, the overall battery SoH curves were brought closer to ideal case results as illustrated in Figure 5.11(d). Particularly, the estimated lifetime of Battery 1 is 21408 cycles as compared to 17237 cycles and 18561 cycles in Approach A and Approach B, respectively. The estimated lifetimes of the four batteries are brought close to each other to achieve a 76% improvement in the system’s lifetime as compared to the baseline. Table 5.4. Settings of parameters for cycling – Approach C Order Reactance, X(Ω) DER1 DER2 DER3 DER4 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0.27 0.28 -0.27 -0.28 III Total (Xk = Xk0+Xk-add) 0.27 0.46 0 0.18 Position 3 4 1 2 Physical (Xk0) 0 0.18 0.27 0.46 Virtual (Xk-add) 0.46 0.09 -0.09 -0.46 V Total (Xk = Xk0+Xk-add) 0.46 0.27 0.18 0 Position 4 3 2 1 5.3.4 Adaptive Cycling – Approach D This approach is an enhanced approach where the decision making logic behind the pecking order of DERs for each cycle is shown in Figure 5.12. In this approach the estimated SoH of all batteries is read once after each load cycle, and then the DER formation is sorted according to their SoH. This implies that the most aged battery takes a drafting position, while the least aged battery takes a leading position. After assigning the ‘electrical’ positions, the necessary positive/negative values of virtual reactances are 76 added to achieve the assigned order. In this way, the pecking order of DER positions as seen from the load can be effortlessly changed. 100 60 Batt1 Batt1 98 Batt2 55 Batt2 C) Order III Batt3 o Batt3 96 Order V 92 Batt4 Order V 91 50 Batt4 94 90 Order V 92 45 Temperature ( Sequence 90 40 continues State of Charge SoC (%) 88 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 3 Batt 1 Batt 2 0.8 Batt 3 Batt 4 0.6 2 76% 0.4 Batt1 Batt2 1 0.2 Batt3 State-of-Health SoH Batt4 Life Aging Parameter L 0 0 1 2 3 4 5 6 7 8 0 5000 10000 15000 20000 25000 Life (cycles) Cycles (c) (d) Figure 5.11: Approach C (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH Figure 5.12: Block diagram for Approach D (of FDERS) 77 Table 5.5: Parameter settings for adaptive cycling – Li-ion batteries DER1 DER2 DER3 DER4 Accumulative Life 0 0 0 0 Position 1 2 3 4 Cycle 0 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) 0 0 0 0 Xk = Xk0+Xk-add(Ω) 0 0.18 0.27 0.46 Accumulative Life 3.01x10-6 2.31x10-6 2.05x10-6 1.67x10-6 Position 4 3 2 1 Cycle 1 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.46 +0.09 -0.09 -0.46 Xk = Xk0+Xk-add(Ω) 0.46 0.27 0.18 0 Accumulative Life 5.51x10-6 4.58x10-6 4.23x10-6 4.08x10-6 Position 4 3 2 1 Cycle 2 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.46 +0.09 -0.09 -0.46 Xk = Xk0+Xk-add(Ω) 0.46 0.27 0.18 0 Accumulative Life 7.94x10-6 6.88x10-6 6.50x10-6 6.75x10-6 Position 4 3 1 2 Cycle 3 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.46 +0.09 -0.27 -0.28 Xk = Xk0+Xk-add(Ω) 0.46 0.27 0 0.18 Accumulative Life 10.6x10-6 9.16x10-6 9.36x10-6 9.18x10-6 Position 4 1 3 2 Cycle 4 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.46 -0.18 0 -0.28 Xk = Xk0+Xk-add(Ω) 0.46 0 0.27 0.18 With the aid of the proposed Approach D, a more balanced and equalized utilization of the four batteries has been realized in Figure 5.13. Overall, a life extension of 80% was 78 achieved with FDERS (i.e. from 11627 cycles in baseline case to 20964 cycles in Approach D). These results of all the approaches are summarized in Table 5.6. 100 60 Batt1 Batt1 98 Batt2 55 Batt2 C) 96 Batt3 o Batt3 Batt4 50 Batt4 94 92 92 45 90 Temperature ( 90 40 State of Charge SoC (%) 88 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 1.5 Batt 1 Batt 2 0.8 Batt 3 Batt 4 1 0.6 80% Batt1 0.4 Batt2 0.5 0.2 Batt3 State-of-Health SoH Batt4 Life Aging Parameter L 0 0 1 2 3 4 0 5000 10000 15000 20000 25000 Life (cycles) Cycles (c) (d) Figure 5.13: Approach D (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH It has been observed that the FDERS-based approaches have a major impact on the SoC (and ΔDoD). This is because all the batteries are stressed in a more balanced manner. The ΔDoD of each battery is changed according to its ‘electrical’ position at each cycle without implementing any physical change to the system, lowering the average SoC and the temperature and thereby extending life of batteries. In all approaches, the 79 accumulated batteries’ lifetimes were balanced to varying degrees depending on the cycling sequence, pattern, and occurrence. Approach A presented a first possible implementation of FDERS toward achieving an equalization of the battery age rates. While Approach A may work for the case of four identical DERs, its ability to balance a broad ranging and realistic scenario of multiple DERs would not be sufficient. Changing the ‘electrical’ positions based on DER power ratings, in Approach B, resulted in further convergence of batteries lifetimes towards the theoretical results. It was originally believed that Approach B would work well in the case of multiple DERs of dissimilar ratings. However, Approach C offered the best solution to the discrepancy in batteries lifetime problem through levelizing the DERs according to their ratings. This approach resulted in the better equalizing of the batteries lifetimes in values close to the ideal case. Also, the concept introduced in Approach C can be easily scaled to a larger microgrid with multiple DERs by assuming more levels and assigning the DER units to these levels according to their power ratings. Unlike the previous three passive approaches, Approach D offers an active solution by estimating SoH of all batteries and then sort them ‘electrically’ according to their age. In all the analysis carried out so far, a key assumption made is that each battery is replaced immediately after its EoL is reached. It must be noted that if the replacement of Battery 1, for example, is not made at the designated time, the supply network to the industrial plant crusher-conveyor can actually collapse within a shorter period than was estimated. This is because the remaining three surviving batteries may not be able to meet the challenging load requirement, thereby leading to a series of cascading failures and 80 eventually resulting in the shutdown of the microgrid. Moreover, the nearly equalized battery life of all four DER units makes it easy for field service engineers to make all battery replacements at the same time. Table 5.6: Comparison of estimated life of batteries (in cycles) – Unequal ratings Cycle Life Life Extension Baseline 11627 0% Approach A 17237 48% Approach B 18561 60% Approach C 20453 76% Approach D 20964 80% Ideal Case 21664 86% 5.4 Industrial Microgrid with Equal-Rated Distributed Energy Resources This section builds on the success of the FDERS-based solutions shown in Section 5.3. This section applies Approach D to the same microgrid considered before, c.f. Figure 5.1. The only difference is that the DERs ratings are equal with equal sizes of batteries, and the design parameters that reflect this condition are given in Table 5.7. As seen in Figure 5.14, the baseline case of operating as in a conventional microgrid results in a varied battery aging even when four equal-rated DERs are employed to supply the crusher-conveyor load. This is similar to the baseline performance observed for unequal- rated DERs in Section 5.3. 81 Table 5.7: DER ratings and design parameters – Equal ratings Battery Rated Power X k k DER k q p Size k (kW) (Ω) (V kVAR-1) (rad kJ-1) (cells) 1 75 0 0.01 0.0039 70 2 75 0.18 0.01 0.0039 70 3 75 0.27 0.01 0.0039 70 4 75 0.46 0.01 0.0039 70 100 Batt1 Batt1 Batt2 Batt2 98 50 C) Batt3 o Batt3 Batt4 Batt4 96 94 93 45 94 92 Temperature ( 40 State of Charge SoC (%) 92 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 Batt1 1 Batt 1 Batt 2 0.8 Batt2 0.8 Batt 3 Batt3 Batt 4 0.6 Batt4 0.6 0.4 0.4 0.2 0.2 State-of-Health SoH Life Aging Parameter L 0 0 1 2 3 4 0 10000 20000 30000 Life (cycles) Cycles (c) (d) Figure 5.14: Equal Ratings Baseline (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH 82 100 Batt1 Batt1 Batt2 98 50 Batt2 C) Batt3 o Batt3 Batt4 Batt4 96 94 45 93 94 92 Temperature ( 40 State of Charge SoC (%) 92 0 100 200 300 400 500 0 200 400 600 800 Time (s) Time (s) (a) (b) -5 x 10 1 Batt1 1 Batt 1 Batt 2 0.8 Batt2 0.8 Batt 3 Batt3 Batt 4 0.6 42% Batt4 0.6 0.4 0.4 0.2 0.2 State-of-Health SoH Life Aging Parameter L 0 0 1 2 3 4 0 10000 20000 30000 Life (cycles) Cycles (c) (d) Figure 5.15: Equal Ratings Approach D (a) battery SoC, (b) battery temperature, (c) life aging parameter for batteries, and (d) battery SoH With the aid of the proposed Approach D a more balanced and equalized utilization of the four batteries has been realized in Figure 5.15. Figure 5.15(a) displays a more balanced and equalized utilization of the four batteries with respect to their SoC. Figure 5.15(b) shows that the temperatures become cooler and more homogenous in comparison to baseline. From the detailed battery aging analysis, the life aging parameters and the estimated batteries lifetimes under FDERS were obtained as shown in Figure 5.15(c) and Figure 5.15(d), respectively. A life extension of 42% was achieved 83 with the use of Approach D (i.e. from 18637 cycles in baseline case to 26456 cycles in FDERS). These results are summarized in Table 5.8. Table 5.8: Comparison of estimated life of batteries (in cycles) – Equal ratings Cycle Life Life Extension Baseline 18637 Approach D 26456 42% 84 Chapter 6 DISCUSSION ON PRACTICAL CONSIDERATIONS The battery cycle life analysis presented in Chapter 5 assumed that the industrial microgrid consisted of a crusher-conveyor with load profile given in Table 5.1. Similar analysis can also be performed for any other practical load levels that are typical of an industrial power system. In addition, the effect of DERs’ interface reactances that are realistic in the microgrid can also be investigated. This section presents selected results after detailed and exhaustive analysis was conducted to test the effectiveness of the proposed FDERS strategy (Approach D) against a spectrum of variables. For simplicity, similar to the analysis presented in Section 5.4, it has been assumed that the DERs are equal-rated or similar to each other. 6.1 Different Levels of Fluctuations in Load In this analysis, the fluctuating load levels in Table 5.1 have been changed keeping all remaining system variables exactly as tabulated in Table 5.7. Specifically, the fluctuating load was reduced below 300kW. At each load level, the batteries’ cycle life simulation has been carried out for the baseline case and under FDERS balancing strategy (Approach D). The results of this analysis are shown in Figure 6.1, where the dashed 85 lines are the baseline results (for Batt1), and the solid line is the balanced lifetime after the proposed FDERS approach has been applied. From this figure, it is clear that the value of FDERS balancing is greater for an industrial power system containing extremely harsh and larger fluctuating loads (as compared to the smaller fluctuating loads), when the batteries get stressed enormously and their cycle life is low. Baseline 300 FDERS 250 Load Level (kW) Level Load 200 18637 Cycles (w/o FDERS) 26456 Fluctuating Cycles 150 (w/ FDERS) 0 50000 100000 150000 Life (Cycles) Figure 6.1: Analysis for different levels of load 6.2 Different Values of DER Interface Reactances (Xko) In this section, the load levels have been fixed back to the values in Table 5.1. However, the values of system reactances, viz., X2o, X3o, and X4o have being varied by an incremental amount (X) in each simulation study to understand better their impact on the system behavior and to test the effectiveness of the proposed FDERS strategy in balancing the battery life in microgrid. The results of this analysis are shown in Figure 6.2. These results show that the discrepancy between batteries’ lifetimes increases 86 when the reactances increase. This happens because a more drafting effect of the ‘electrically’ farther DERs increases burden on the leading DER unit. Again, the main outcome has demonstrated that the proposed strategy works well here too. 0.35 Baseline 0.3 FDERS 0.25 ) 0.2 X ( X 0.15 18637 26456 Cycles Cycles 0.1 (w/o FDERS) (w/ FDERS) 0.05 0 15000 20000 25000 30000 Life (Cycles) Figure 6.2: Analysis for different values of DER interface reactances 6.3 System of Different Types of DERs This section presents another application of employing FDERS in industrial microgrids to achieve the target of extending the microgrid operation. It has been accepted for publication in the IEEE Industry Applications Magazine [29]. The section starts with a system description for the microgrid considered. Then, the problem is investigated in detail and a solution is proposed. The adopted microgrid is 2-DERS system, as shown in Figure 6.3. All the components of the microgrid have been modeled and simulated in MATLAB/Simulink/ SimPowerSystems toolbox. The microgrid design parameters are given in Table 6.1. 87 Afterwards, the 2-DERS system was subjected to the crusher-conveyor load with one half the profile given in Table 5.1 in the previous chapter. Figure 6.3: Single line diagram of a 2-DERS system supplying a crusher-conveyor load Table 6.1. Design parameters for 2x75-kW DERS system shown in Figure 6.3 Reactance DERSk T b (Xk) pk pk (s/rad.) (rad./J) DERk DESk Ohms p.u. DER1 = MT Batt1 0 0 0 π/(75K) DER2 = FC Batt2 0.46 0.15 0 π /(75K) The dynamic response of the DER1 after the load change event is illustrated in Figure 6.4(a), whereas DER2 dynamic response is shown in Figure 6.4(b). As seen in these plots, the leading DER1 connected right across the load terminals takes the strike at the instant of the load change, and DER2 is able to draft behind. Although the final 88 steady-state power sharing between the two DERS is identical when they take up 75-kW each, their individual dynamic responses are distinct. This happens because the drafting DER2 is electrically displaced from the location of load change event by inductive reactance X2, and the entire burden initially fell on the leading DER1. This discrepancy in transient load sharing between the two DERS, combined with the different response rates of their prime-movers/energy sources (viz. MT and FC, respectively), is expected to result in varied stresses for the battery of leading DER1 as against that of drafting DER2. In each DER, the flow of power from the battery is regulated by a battery management system as illustrated in Figure 6.5 [119]. The battery sizing and the design specifications are given in Table 6.2. By using the simulation models described in Chapter 3, the battery SoC and temperature for the recurring crusher load cycles have been obtained as shown in Figure 6.4(c) and Figure 6.4(d). It clearly indicates a disparity in SoC as well as temperature variations of the two batteries, viz., Batt1 and Batt2. Evidently, the battery of leading DER1 (i.e. Batt1) undergoes a larger variation in the SoC and temperature due to its proximity to the fluctuating load in addition to the slow response characteristics of MT. Such a cycling of battery temperature and SoC is known to affect adversely its lifetime [125, 126], and so this particular aspect has been investigated further in battery aging analysis. After simulating the battery aging model presented in Chapter 3 in MATLAB/Simulink, the accumulated aging life (L) and state of health (SoH) have been plotted in Figure 6.4(e) and Figure 6.4(f) over several thousand crusher load cycles for Batt1 and Batt2. It is clear that the 2-DER microgrid operation causes significant variations in aging 89 140 140 120 120 DER2 100 DER1 100 80 80 Batt2 60 Batt1 60 Power (kW) Power (kW) 40 40 FC 20 MT 20 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Time (s) Time (s) (a) (b) 100 35 Batt1 33 Batt2 C) 95 o 31 90 Batt1 29 Batt2 Temperature ( 27 State of Charge SoC (%) 85 25 0 100 200 300 400 500 0 100 200 300 400 500 Time (s) Time (s) (c) (d) x 10-4 1 1 Batt1 Batt 1 Batt 2 0.8 Batt2 0.8 0.6 0.6 0.4 0.4 Batt1 needs to be replaced or 0.2 0.2 State-of-Health SoH else the system Life Aging Parameter L will shutdown 0 0 1 2 3 4 0 5000 10000 15000 Cycles Life (cycles) (e) (f) Figure 6.4: Battery life analysis for 2-DERS microgrid in Figure 6.3 90 of Batt1 and Batt2. Batt1 in the leading position has been over-utilized whereas Batt2 in the drafting position has been relatively under-utilized. In particular, Batt1 reached its end of life (EoL) after 10100 cycles. This can potentially cause tripping of the leading DER1 at the next load change event since the microturbine (its primary energy source) cannot independently meet the transient load demands without Batt1’s energy storage support. At that time, the industrial power system can shut down, if Batt1 is not immediately replaced by a new battery pack. Figure 6.5: Block diagram illustrating division of power flow in each DERS Table 6.2: Battery sizing particulars for the 2x75-kW DERS system DERSk No. of Strings in No. of Cells/ Total No. of DERk DESk Parallel String Cells Batt DER = MT 1 3 55 165 1 (6.67 Ah) Batt DER = FC 2 2 55 100 2 (3.89 Ah) The microgrid operation of the previous section can be viewed as a form of teaming up among the 2 DERS to supply the large and fluctuating industrial load. One can see many 91 parallels between such a DER formation and the birds/V-formation or riders/peloton formation that was referred to in Chapter 4. For example, in case the positions of birds/riders in their formation are fixed over the entire journey, the leading bird/rider gets exhausted sooner than its or his drafting counterparts do. This is similar to Batt1 (of leading DERS1) reaching its end of life (EoL) before the Batt2 (of drafting DERS2) in the 2-DERS microgrid. To solve this problem and to extend endurance limits of the collective network, all the team members cooperate and rotate their positions within the formation. In this dissertation, FDERS incorporates flexibility in the integration of DERs through synthesized virtual reactances, virtual inertias, and adaptive DERS controls. Consider the two-DERS system as shown in Figure 6.3 with physical interface reactances X1 = 0, and X2 = 15%. At the instant of load change, the pecking order is determined by the ascending order of Xk’s. However, the only way to make an initially leading DER (that has a low Xk) stay in that leading position for a sustained period is by implanting a larger virtual inertia through the time-constant parameter Tpk 0 in the active power controller (cf. Figure 4.4). This technique has been utilized to change the pecking order of the 2-DER system, specifically to bring the DER2 into the leading position after Batt1 reached its EoL (i.e. after 10100 cycles). By so doing, it can be ensured that from the next load change event Batt2 will get utilized more than Batt1. For the particular set of design parameters given in Table 6.3, the dynamic response to the next load change was obtained as shown in Figure 6.6, where Batt1 (that had reached its EoL) is not burdened anymore and all the energy storage needs of the 2-DER system are directed toward Batt2 alone. Then, the 92 battery parameter estimation and aging analysis was again performed for the 2-DERS system. The outcome of making such a FDERS intervention, as clearly showed by the variation of battery aging shown in Figure 6.7, was an extension of the 2-DERS system operation without shutdown until Batt2 also reached its end of life (EoL). In particular, FDERS enabled continued operation of the 2-DER system beyond 10100 cycles up to 13260 cycles, which is around 31% increased utilization of available resources before the Batt2 also reaches its EoL, and the system shuts down. Because the FDERS offered an extended operation resulting in complete utilization of the two batteries before causing a system stoppage, it has a significant impact on extending the mean replacement time of the industrial power system. This fact is graphically illustrated in Figure 6.8, where the battery replacement times with FDERS are compared against those without FDERS. As seen in Figure 6.8, the time at which the battery replacements need to be conducted by service and maintenance engineers at the industrial sites employing FDERS is about twice longer than the period taken in the conventional approach (without FDERS). Table 6.3: Design parameters for obtaining results shown in Figure 6.6 DERSk Reactance (Xk =Xko+ Xk-add) Ohms p.u. Tpk bpk (s/rad.) (rad./J) DERk DESk Physical Virtual Total Total (Xko) (Xk-add) (Xk) (Xk) DER1 = MT Batt1 (reached EoL) 0 0.46 0.46 0.15 0 π/(75K) DER2 = FC Batt2 0.46 -0.46 0 0 2.67π π /(75K) 93 150 Increasing T2 and X1 Leading FC DER2 100 Batt2 MT Drafting DER1 Power (kW) 50 Increasing T2 (Batt1 reached and X1 its EoL) 0 10 20 30 40 50 60 Time (s) Figure 6.6: Response of the 2-DERS system with the implementation of FDERS 1 Batt1 0.8 Batt2 0.6 Batt2 reached 31% 0.4 its EoL Batt1 0.2 reached State-of-Health SoH its EoL 0 0 5000 10000 15000 Life (cycles) Figure 6.7: Battery aging in a 2-DERS system illustrating the benefits of FDERS 94 1 0.8 Replace Replace Batt1 Batt2 0.6 Replace Replace Batt1 Batt1 0.4 Replace Batt2 0.2 State-of-Health SoH 0 0 0.5 1 1.5 2 2.5 3 3.5 4 Life (cycles) x 10 (a) 1 0.8 0.6 Replace Replace Batt1 & Batt1 & 0.4 Batt2 Batt2 0.2 State-of-Health SoH 0 0 0.5 1 1.5 2 2.5 3 3.5 4 Life (cycles) x 10 (b) Figure 6.8: Comparison between the battery replacement times for the 2-DERS system – (a) conventional (without FDERS) and (b) with FDERS 6.4 Balancing the Supercapacitor Cycle Life This section provides an analysis to the ability of FDERS-based solution (Approach D) in cycle life equalization of supercapacitors. Supercapacitors are a practical and viable option for short-term high-power applications [35]. For this reason, supercapacitors have 95 a great potential to be utilized in industrial applications [54]. In this section, the considered microgrid system comprises four identical fuel cell-supercapacitor DERs, as illustrated in Figure 6.9. This section shows that the balancing strategies and the concept of changing the ‘electrical’ placement of the DERs can be applied to supercapacitors as successfully as it is applied to batteries. The content of this section have been accepted for publication in the IEEE Transactions on Industry Applications [127]. The same microgrid described in Section 5.1 is used in this section. The only difference is that in this section, four identical fuel cell-supercapacitor DERs are considered rather than fuel- cell battery DERs, as shown in Figure 6.9. A block diagram summarizing the system modeling strategy is shown in Figure 6.10. A dynamic MATLAB/ Simulink model was built for the microgrid and a time-domain simulation was carried out. The output power of each DER was determined for the load profile given in Table 5.1. Subsequently, the supercapacitor aging model was run offline to estimate the capacitance retention CT by using the temperature data obtained from time-domain simulation. This dissertation adopts the supercapacitor model explained earlier in Subsection 2.3.2 and the supercapacitor aging model explained in Subsection 2.5.2. Maxwell BCAP1200 has been chosen after the evaluation of various models of supercapacitors from different manufacturing companies because of its lightweight, optimal size, and minimal losses [128, 129]. The specifications of Maxwell BCAP1200 are given in Table 6.4. 96 Figure 6.9: Single-line diagram of a microgrid consisting of four fuel cell-supercapacitor hybrid DERs supplying a crusher-conveyor Table 6.4: Supercapacitor Specifications [128] Manufacturer Maxwell Capacitance 1200 F Mass 260 g Rated Cell Voltage 2.70 V Internal Resistance 0.58 mΩ 97 Figure 6.10: Block diagram of modeling strategy for supercapacitor life estimation Table 6.5: DER ratings and design parameters for microgrid in Figure 6.3 Rated Power k k SC Module DER X (Ω) q p k (kW) ko (V kVAR-1) (rad kJ-1) (cells in series) 1 75 0 0.01 0.0039 213 2 75 0.18 0.01 0.0039 213 3 75 0.27 0.01 0.0039 213 4 75 0.46 0.01 0.0039 213 Similar to the findings of Chapter 5, the simulation results for the four DERs showed a varied dynamic response, which is strongly influenced by their ‘electrical’ locational placement within the microgrid. This is caused by inherent reactances of the transformers, line filters, distribution lines, etc. that are generally present in the network. The dynamic response of microgrid supplying the load is depicted in Figure 6.11; this is the baseline response. A difference in the transient responses resulted in significant variations in the utilization of various supercapacitors within the microgrid. The more 98 ‘electrically’ closer a DER was to the crusher-conveyor load, the more stressed is its supercapacitor. This caused differences in cell voltage (cf., Figure 6.12 (a)) and current drawn (cf., Figure 6.12 (b)) for all supercapacitors, thus implying a lower state of charge (SoC), as depicted in Figure 6.14 (a). The higher drawn current led to a higher rate of energy conversion resulting in heat losses [124]. This eventually increased the temperature of supercapacitor as shown in Figure 6.14 (b). The temperature of Supercapacitor 1 shot up to 59.2oC from an ambient temperature of 40oC. During the same time, the temperature of Supercapacitor 4 increased up to only 48.9oC. 300 DER1 DER2 200 DER3 DER4 100 Power (kW) 0 0 0.5 1 Time (s) Figure 6.11: DER power profiles for baseline case By applying the SoC and temperature data obtained from earlier conducted time-domain simulation to an offline supercapacitor aging model, which was also developed in MATLAB, the resulting capacitance retention CT for the four DER supercapacitors was obtained, as shown in Figure 6.14 (c). This figure shows a very clear imbalance in the accumulation of the aging factor. It is observed that the system will shut down as soon as Supercapacitor 1 reaches its end-of-life (EoL) after 4.1239x105 cycles. 99 580 400 SC1 SC1 SC2 SC2 300 560 SC3 SC3 SC4 SC4 200 540 Current (A) 100 String Voltage (V) 520 0 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Time (s) Time (s) (a) (b) Figure 6.12: (a) Voltage and (b) current of the supercapacitor module for one cycle Approach D, cf. Subsection 5.3.4, was utilized to solve the supercapacitor life discrepancy problem. The pecking order of DERs for each cycle was determined by the decision-making logic shown in Figure 6.13. A sample of three consecutive cycles is shown in Table 6.6. All the SCs of DER units started with capacitance retention of 100%, i.e., without any virtual reactances added to the system. Then, after the first cycle, DER1 was found to age the most, since it is in the first position, while DER4 aged the least. These changes in capacitance retention are reflected in Cycle 2, when DER1 was made to draft to the last position by adding a virtual reactance (i.e., X1-add = X4o - X1o), while DER4 was moved to the leading position by adding a negative virtual reactance (i.e., X4- add = X1o - X4o, since X1o < X4o). In the same manner, DER2 and DER3 also switched their respective ‘electrical’ locations. Subsequently, the same sequence is followed for Cycle 3 and thereafter. The results showed a significant balance in the different variables that affect supercapacitor aging. The SoCs and temperatures of the supercapacitors after applying the periodic cycling strategy are shown in Figure 6.15 (a) and Figure 6.15(b). It 100 is evident from these figures that the periodic cycling produced a cooler and more homogenous temperature in comparison to the baseline case (cf., Figure 6.14 (b)). After conducting the detailed supercapacitor aging analysis, similar to that performed for the baseline case, the life aging calculations have been carried out. The resulting capacitance retention under FDERS was obtained as shown in Figure 6.15(c). The life of the system was extended to 4.8131x104 cycles from 4.1239x104 cycles in baseline (i.e. an addition of 17% life). Figure 6.13: Block diagram for the adaptive cycling approach to equalize the lifetimes of supercapacitors in the microgrid From the results of Figure 6.15, it can be observed that the proposed FDERS-based solution had a major impact on the utilization of the supercapacitors in the systems. This happens because all the supercapacitors were stressed in a more balanced manner. This lowered the average temperature and consequently balanced the cycle life of all supercapacitors in the microgrid. 101 100 60 95 C) 55 o 90 50 SC1 SC1 SC2 SC2 85 45 SC3 SC3 Temperature ( SC4 State of Charge SoC (%) SC4 80 40 0 100 200 300 400 500 0 200 400 600 Time (s) Time (s) (a) (b) 100 SC1 SC2 95 SC3 SC4 90 SC1 reached its EoL 85 Capacitance Retention (%) 80 0 10000 20000 30000 40000 N (cycles) (c) Figure 6.14: Equal ratings - baseline case (a) supercapacitor SoC, (b) temperature and (c) capacitance retention 102 100 60 95 C) 55 o 90 50 SC1 SC1 SC2 SC2 85 45 SC3 SC3 Temperature ( SC4 State of Charge SoC (%) SC4 80 40 0 100 200 300 400 500 0 200 400 600 Time (s) Time (s) (a) (b) 100 SC1 95 SC2 SC3 SC4 90 17% 85 Capacitance Retention (%) 80 0 10000 20000 30000 40000 N (cycles) Life (cycles) (c) Figure 6.15: Equal ratings – FDERS solution (a) supercapacitor SoC, (b) temperature and (c) capacitance retention 103 Table 6.6: Parameter settings for adaptive cycling – Supercapacitors DER1 DER2 DER3 DER4 Cap. Retention (%) 100 100 100 100 Position 1 2 3 4 Cycle 1 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) 0 0 0 0 Xk = Xk0+Xk-add(Ω) 0 0.18 0.27 0.46 Cap. Retention (%) 99.87 99.90 99.92 99.95 Position 4 3 2 1 Cycle 2 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.46 +0.09 -0.09 -0.46 Xk = Xk0+Xk-add(Ω) 0.46 0.27 0.18 0 Cap. Retention (%) 99.78 99.84 99.85 99.74 Position 3 2 1 4 Cycle 3 Physical Xk0 (Ω) 0 0.18 0.27 0.46 Virtual Xk-add (Ω) +0.27 0 -0.27 0 Xk = Xk0+Xk-add(Ω) 0.27 0.18 0 0.46 104 Chapter 7 SHARING STRATEGY FOR FLEET VEHICLE-TO-GRID (V2G) SYSTEMS As mentioned earlier in Chapter 1 and Chapter 2, the electric power grid is currently witnessing a massive deployment of renewable energy resources. Therefore, energy storage is recommended for grid stabilization [130], smoothening the generation intermittency from renewables [131] and making efficient usage of electricity [132]. The energy storage devices such as batteries have become an essential component of electric vehicles that are gaining acceptance because of environmental concerns and regulations [133]. Electric vehicle batteries can be utilized for stabilizing the power grid [134, 135]. This is advantageous considering the fast response, charge/discharge efficiency, and mobility advantage of batteries [136]. 7.1 Motivation Research showed that in the U.S. most passenger vehicles are not used 96% of the time [134]; this figure is even lower in Germany: 89% [137]. Consequently, the utilization of a passenger vehicle’s power capacity is significantly low, thereby becoming the main motivation for vehicle-to-grid (V2G) systems [138]. V2G is advantageous because electric vehicles can be plugged-in to the distribution grid during long parking hours, 105 when their battery utilization is otherwise low. Such an electric vehicle may be a battery electric vehicle (BEV), plug-in hybrid electric vehicle (PHEV), or fuel cell electric vehicle (FCEV) [139]. The major goals of V2G systems are to increase reliability, enhance stability, cut the energy storage costs, and provide a backup for renewables in an islanded microgrid [140, 141]. Earlier publications, viz., [142, 143], reported various V2G case studies. Adding PHEVs to the conventional generation scheduling problem can reduce the operation costs [142]. A charging strategy was presented in [143] for the energy storage in PHEVs to decrease the peak load. 7.2 State-of-the-art strategies in V2G systems State-of-the-art control strategies treats all the electric vehicles connected by V2G alike, disregarding their diverse driving histories and unequal battery aging. Very few publications investigated the impact of battery aging. V2G charging processes to slow down the battery aging were presented in [74, 144]. Guenther et al. [74] proposed a battery model that considers lifetime effects to design battery system for various ambient conditions and user behavior, and also for calculating aging-dependent economic metrics. More recently, the effect of V2G strategy on the lifetime of Li-ion batteries was studied in [144]. It has been concluded that the cycling aging (and not the calendar aging) has a greater impact on the overall battery aging. Hence, dissimilar driving cycles in electric vehicles should cause a heterogeneous battery aging. This dissertation proposes to coordinate the battery aging in fleet electric vehicles based load sharing strategy for V2G. It has been simulated and tested using 106 MATLAB/Simulink for the physical layer and Java Agent DEvelopment Framework (JADE) [145] for programming the agents in cyber layer. The communication between power system and cyber system was facilitated by TCP/IP protocol. Multi-agent systems help realize distributed intelligence in the controls of power system, in contrast to the conventional centralized controls [146-148]. In the proposed V2G strategy, the load shared by each fleet electric vehicle’s battery is controlled based on its battery’s state-of-health (SoH). This means that a battery with higher SoH will contribute more than a lower SoH battery. This is shown to result in an even distribution of load between the different vehicles. This strategy is derived from the flexible distribution of energy and storage resources (FDERS) concept [29], which was inspired by the energy saving formations in bird flocks and cycling teams, as discussed in Chapter 4. 7.3 Fleet Vehicle-to-Grid (V2G) System and Adopted Models The V2G system is part of an isolated microgrid comprising solar/photovoltaic (PV) and fuel cell (FC) types of distributed energy resources (DERs). There are four electric vehicle charging stations (CS) feeding the fleet vehicles to a commercial office building load demand. The structure and topology of CS has been adopted from [149], and the load profile of commercial office building was taken from [150]. A small-scale 4-kW solid oxide FC was modeled according to [116]. For the PV system, the power generation profile has been obtained from [87]. The PV system operates at its maximum power point, and the FC meets the balance when it is not enough to supply the load. 107 Figure 7.1 illustrates the aggregated supply and demand in the commercial office building, and Figure 7.2 shows the single line diagram for the adopted V2G system. In Figure 7.2, the green shaded area shows an excess energy that can charge the vehicles’ batteries, and the red shaded area denotes a lack of energy that needs support from the same batteries. It is noteworthy that all the CSs in the microgrid have power electronic inverters that are controlled by using active power-frequency and reactive power-voltage droop controls. Figure 7.1: Supply and demand of the microgrid In this dissertation, the considered battery electric vehicle (BEV) is Nissan Leaf. The electric vehicle modeling is based on [151] with all the specifications and parameters of the Nissan Leaf found on the manufacturer website. Besides, the battery electro-thermal model was built using the method given in [51], and battery aging model used that of [69]. It must be noted that the battery aging model of [69] has been earlier validated in [71] against real test data. 108 Battery aging is indicative measure of the damage to battery life. It is denoted by the life aging parameter L, which can vary between 0 (for new battery) to 0.2 (for dead battery). A value of L=0.2 is deemed the end of useful life [73]. The L is calculated by solving the equations explained earlier in Subsection 2.5.1. Figure 7.2: A microgrid for testing the fleet V2G systems consisting of PV, FC, charging stations for the fleet vehicles at a commercial office building In the fleet V2G system (see Figure 7.2), the four different BEVs were assumed to have different driving cycles. The driving cycle of the first vehicle, viz., BEV1, is the Urban Dynamometer Driving Schedule (UDDS), which represents an urban driving scenario with frequent stops. The driving cycle for the second vehicle, viz., BEV2, is Highway Fuel Economy Test (HWFET), which characterizes a highway driving scenario. For the 109 purpose of driving cycles of the third and fourth vehicles, half of UDDS and half of HWFET are considered to BEV3 and BEV4, respectively. These driving profiles are U.S. certification cycles that represent different standard driving patterns and can be found on U.S. Environmental Protection Agency (EPA) website (see Figure 7.3). 100 50 Speed (km/h) 0 0 500 1000 1500 Time (s) (a) 100 50 Speed (km/h) 0 0 200 400 600 800 Time (s) (b) Figure 7.3: Driving cycles – (a) UDDS (b) HWFET It is fair to assume that different drivers have their own unique driving behaviors. However, the driving actions can be characterized with the main variables of speed, acceleration, and duration. Simulation on the models described in previous section was carried out using MATLAB/Simulink for testing the four driving cycles. While driving the BEVs, the SoCs (i.e., state-of-charge) of vehicle batteries change in accordance with a driving cycle. For instance, the urban driving cycle (UDDS) requires lesser energy, since speed limits are lower; this manifests itself in lower depth-of- 110 discharge or DoD (i.e., higher SoC). This is so when compared to more aggressive driving cycles of higher speed limits (i.e., HWFET). Figure 7.4(a) shows the SoCs corresponding to the four driving cycles. As seen in the figure, SoC was 93.18% for battery of BEV with UDDS, 89.22% for battery of BEV with HWFET, 96.18% for battery of BEV with half UDDS, and 95.29% for battery of BEV with half HWFET. Further, a larger discharge current implies higher rate of energy conversion leading to heat losses that raise the temperature of the battery packs. When driving cycle was HWFET, the temperature reached 46oC, as seen in Figure 7.4(b). In contrast, driving with multiple stops in UDDS cycle allowed the battery to cool, and so the battery temperature did not rise beyond 35oC (see Figure 7.4(b)). Most publications on V2G systems assume that the battery cycle life is very long (for instance, [152]). As such, they generalize the battery charging/discharging for the entire electric vehicle fleet disregarding the individual vehicle driving histories that can cause varied battery aging [153, 154]. When the BEV batteries are plugged-in to supply the energy deficit during evening hours (cf., red shaded area in Figure 7.1), the load is often divided equally among the four BEVs [155]. This does not take into consideration the differences observed in SoC and temperature among the BEV batteries (cf., Fig. 5). Such an approach was found to result in the same DoD (i.e., ΔSoC) for all batteries during the discharging to support the microgrid, as shown in Figure 7.5(a). However, according to the battery aging model, the SoC and temperature are the two main factors affecting the rate of aging of the battery. After conducting the battery aging analysis for the four BEVs described in the earlier section, the results showed huge 111 discrepancies in their life aging parameters. Figure 7.5(b) shows the life aging parameter over 7 consecutive days of operation. Over a longer time, this trend led to increased deviation in the battery aging of four BEVs. This was confirmed from the SoH values calculated by using battery aging model presented in Subsection 2.5.1; the results showed a marked deviation as displayed in Figure 7.5(c). Another point to note is that after BEV2 reached its EoL (i.e., SoH = 0), the slopes of the SoH curves of remaining 3 BEVs became steeper. This happens because the load burden is now divided into these 3 BEVs’ batteries rather than the earlier 4. Similar observation was found when BEV1 reached its EoL. Therefore, the results shown in Figure 7.5(c) prove that the conventional V2G controls approach to ignore the vehicle driving histories does not lead to an optimal solution. Such a discrepancy in battery aging has been first reported by the authors in [156]. In the following section, an agent-based load sharing strategy is proposed for coordinating the battery aging in the fleet V2G system. 100 50 UDDS Half UDDS 45 Full C) o HWFET 95 Half 40 HWFET Full UDDS Half 35 UDDS 90 Full HWFET Temperature ( Half 30 State of Charge SoC (%) HWFET Full 85 25 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 Time (s) Time (s) (a) (b) Figure 7.4: Batteries parameters for different driving cycles 112 7.4 Load Sharing Strategy and its Benefits This section proposes a V2G load sharing strategy that considers battery aging history of all the fleet vehicles connected to the commercial office building. For this purpose, it is assumed that each BEV is equipped with an on-board device that keeps a record of its battery state-of-charge (SoC) and battery temperature. Furthermore, the device estimates the battery SoH after computing the life aging parameter (L). The proposed scheme has been built using software agents in Java Agent DEvelopment Framework (JADE). The main advantage of using a multi-agent system (MAS) for V2G is in realizing the system goals with coordinated control of local variables. This helps achieve a decentralized operation by using all available resources efficiently. Such distributed architectures help in achieving optimal controls even as the number of BEVs connected to the V2G system is increased by a large amount. In the proposed scheme, a software agent is assigned to each BEV and each charging station (CS). In particular, the block diagram depicting action of BEV agent is shown in Figure 7.6. At first, every BEV agent senses its battery temperature and current; then, it calculates SoC using I (t) SoC(t) = SoC(t − 1) + batt ∆t (7.1) Qn where Qn is the nominal capacitance of the battery. At the end of every charging/discharging cycle, the same agent calculates the aging parameter using equations of battery aging models presented earlier in Subsection 2.5.1. Finally, the agent determines the accumulated life aging parameter (L) using (7.2). Later, when the BEV is 113 connected to the CS, i.e., plugged-in to the microgrid, the same agent transmits the value of L to the CS agent through lines of communication. Figure 7.7 shows the block diagram illustrating a CS agent action. The CS agent, in its turn, registers the value of L from the corresponding BEV agent. Besides, it also gathers L-values from all other BEV agents in the V2G system through LAN communication lines. With all the L-values available in hand, the CS agent determines the load share for its battery after calculating Leq and D values according to the following equations n 1 1 = ∑ (7.2) Leq Lk k=1 Leq Dk = × 100%; for k = 1,2, … , n (7.3) Lk It is noteworthy that once a BEV agent determines that its battery has reached the end-of- life (EoL), it transmits information to the corresponding CS agent. This will lead to excluding itself from (7.2) and setting its share factor (D) to be zero, the implication being that it cannot share the load burden with others any more. The general equation for sharing any change in the load is given by ∆Pk = Dk × ∆PL (7.4) 114 -4 x 10 100 4 UDDS Half UDDS 80 3 Full HWFET Half HWFET 2 Full 60 1 Life Aging Parameter L State of Charge SoC (%) 40 0 18 20 22 24 1 2 3 4 5 6 7 Time (s) Days (a) (b) 1 UDDS Half 0.8 UDDS Full HWFET 0.6 Half HWFET Full 0.4 0.2 State-of-Health SoH 0 0 1000 2000 3000 4000 Life (days) (c) Figure 7.5: Batteries (a) SoC when connected to the microgrid, (b) life aging parameter, and (c) overall SoH for baseline scenario. 115 Current sensor Temperature Calculate sensor SoC Record T and SoC Yes I≠0? End of a single No (dis)charge cycle Calculate Aging parameter for Lcycle this cycle Compute L Add to the accumulative aging parameter Figure 7.6: Block diagram of the BEV agent. From other agents L1 from BEV1agent L2 L3 … LN … BEV1 battery reached EoL Yes L≥0.2? No Set Calculate LAN D1=0 Leq Calculate Share factor for D1 BEV1 Figure 7.7: Block diagram of charging station (CS) agent 116 7.5 Results The overall performance of fleet V2G controls described in the earlier sections are illustrated in Figure 7.8. Variation of load share factors (Dk), facilitated by the proposed scheme, has led to diverging SoCs during discharge in Figure 7.8(a). It was observed that the SoCs of the two least aged batteries, i.e., those undergoing half driving cycles, are lower than of the remaining two more aged batteries. This means that BEV3 and BEV4 – that have reduced stress due to half driving cycles – are further exploited (than BEV1 and BEV2) for supplying the power shortage of commercial office building load in the evening hours (cf., red shaded area of Figure 7.1). The dashed red line in Figure 7.8(a) indicates the SoC for the baseline, i.e., conventional equal load sharing approach, was earlier shown in Figure 7.5(a). Figure 7.8(b) shows the accumulated aging parameters of the four BEVs’ batteries. It is clear that the L-values get closer to each other by comparison to the baseline case displayed in Figure 7.5(b). The resulting aging of the batteries due to the proposed strategy is shown in Figure 7.8(c). The benefits of proposed load sharing strategy are evident from the contrast between the battery aging results of Figure 7.8(c) and those from the baseline case shown in Fig. 6(c). Table 7.1 gives a side-by-side comparison of these two cases. Applying the proposed control strategy has boosted the cycle life of the batteries of all BEVs in the system. It has also became clear that the cycle lives of batteries became closer to each other (i.e., became harmonized or coordinated). In the baseline scenario, the battery of BEV2 reached its EoL ahead of the battery of BEV4 by 1.85 years. However, the cycle lives of the four batteries after the proposed strategy had been carried out were found to be within 117 1.44 years of each other. This is a 22.2% enhancement in lifetime balancing. Besides the balancing feature, the cycle lives of all BEV batteries of fleet vehicles have increased by 2% to 8%. -4 x 10 100 UDDS 4 Half UDDS Half UDDS Full UDDS HWFET Full 80 Half 3 HWFET HWFET Full Half HWFET Baseline 2 Full 60 1 Life Aging Parameter L State of Charge SoC (%) 40 0 18 20 22 24 1 2 3 4 5 6 7 Time (s) Days (a) (b) 1 UDDS Half 0.8 UDDS Full HWFET 0.6 Half HWFET Full 0.4 0.2 State-of-Health SoH 0 0 1000 2000 3000 4000 Life (days) (c) Figure 7.8: Batteries (a) SoC when connected to the microgrid, (b) life aging parameter, and (c) overall SoH after applying the proposed strategy 118 Table 7.1: Comparison of Battery Life (in years) for the Proposed Solution against Baseline Case Proposed Battery Driving Baseline Vehicle Solution Lifetime Cycle (Years) (Years) Enhancement BEV 1 Full UDDS 9.27 9.56 3.13 % BEV 2 Full HWFET 8.01 8.64 7.87 % BEV 3 Half UDDS 9.78 9.98 2.05 % BEV 4 Half HWFET 9.86 10.08 2.23 % Δ Lifetime (Years) 1.85 1.44 22.2 % 7.6 Discussion An important point to note is that the proposed load sharing strategy is effective only when the BEVs are plugged-in during the 15:00 24:00 hours for discharge. Since this scheme has no influence over the driving behavior – that impacts the battery aging – the SoH curves in Figure 7.8(c) can never converge to the same point. Nevertheless, the share factors (Dk) get adjusted on-the-fly based on changing driving pattern to bring the battery cycle lives closer to each other. This is further proved in another more realistic case study below. An assumption was made in the earlier analysis that all the BEVs had new batteries in the beginning, i.e., SoH = 1. Since this may not always be true, it is assumed in the following study that the BEVs start with different SoH values. In particular, battery of BEV1 starts with SoH = 0.90, ~ BEV2 starts with SoH = 0.8, and ~ both BEV3 and BEV4 start with SoH = 1. The corresponding results are shown in Figure 7.9. Conventional V2G scheme (i.e., baseline) results displayed in Figure 7.9(a) indicate that BEV2 (i.e., full HWFET) reached EoL much faster, because it started with the lowest SoH in addition to its 119 aggressive driving cycle. The proposed strategy yielded significant benefits, as shown in Figure 7.9(b). Because BEV2 had the lowest SoH (i.e., highest L), it was assigned the minimum sharing factor D, refer to (7.3). In so doing, the life of BEV2 battery got extended by about 40% as compared to baseline. The final results, tabulated in Table 7.2, show that the aging of all batteries became 72% more harmonized in this case. Thus, it was proved that the proposed approach works satisfactorily under both changing driving cycles and dissimilar battery aging histories. 1 1 UDDS UDDS Half Half 0.8 UDDS 0.8 UDDS Full Full HWFET HWFET 0.6 Half 0.6 Half HWFET HWFET Full Full 0.4 0.4 0.2 0.2 State-of-Health SoH State-of-Health SoH 0 0 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Life (days) Life (days) (a) (b) Figure 7.9: Batteries SoH when starting at dissimilar SoH values Table 7.2: Comparison of Battery Life (in years) for the Proposed Solution against Baseline Case (Dissimilar Initial SoH) Proposed Battery Driving Baseline Vehicle Solution Lifetime Cycle (Years) (Years) Enhancement BEV 1 Full UDDS 9.27 9.56 3.13 % BEV 2 Full HWFET 8.01 8.64 7.87 % BEV 3 Half UDDS 9.78 9.98 2.05 % BEV 4 Half HWFET 9.86 10.08 2.23 % Δ Lifetime (Years) 1.85 1.44 22.2 % 120 Chapter 8 CONCLUSIONS AND FUTURE WORK Microgrids are being increasingly favored to address the growing needs of high energy consumers in the modern power grid and to facilitate higher levels of renewable energy penetration. The output power of renewables is intermittent, because of the uncertain nature of solar and wind energy. In addition to them, microgrids have to deal with small- rated DERs such as fuel cells and microturbines that have long start-up time and poor transient response due to their intrinsic characteristics. Therefore, renewables and slow- responding DERs are generally supplemented with energy storage systems (ESS). 8.1 Conclusions This dissertation has evaluated energy storage systems (ESS) in modern microgrid applications. It has covered the two main roles that energy storage systems (ESS) play (i) for short-term requirements (i.e., power applications), and (ii) for long-term requirements (i.e., energy applications). The differences between these two roles and the corresponding energy storage technologies that match with the requirements have been summarized. Supercapacitors have been found to be a great candidate for power applications, while batteries have been found to meet energy applications better. 121 A generalized method has been proposed in this dissertation for ESS sizing for microgrids that uses energy balance concept rather than current levels. This method has used the load profile of the targeted industry and calculated the ESS size by taking into account operating temperature and aging factors. A practical solution has been presented for sizing batteries for peak shaving and supercapacitors for frequency regulation. This dissertation has also analyzed battery cycle life in a microgrid featuring large and fluctuating loads like crushers, excavators and steel rolling mills at any mining or metal industrial sites. Such loads are critical loads that demand highly reliable power. Their industrial power system design requires a precise and painstaking planning at the early stages. This dissertation has introduced the new framework of Flexible Distribution of EneRgy and Storage resources (FDERS) for reliably supplying power to the extremely harsh and demanding loads from a network of multiple smaller-rated distributed energy and storage resources especially when power from the main grid is not available. This framework was inspired by the V-shape formation of flocks of birds and peloton/echelon formation of cycling racing teams for extending their endurance limits. Similar ideas have been applied in FDERS to establish a new paradigm in integration of distributed energy and storage resources for offering greater sustainability in terms of benefits such as increased resource lifetime, optimal energy storage deployment, enhanced controllability, and improved system robustness. FDERS is a cooperative framework among multiple interconnected DERs that achieves these goals with reconfiguration, virtual reactances, virtual inertias, and adaptive DERs controls. The flexibility is incorporated in the controllers by synthesizing virtual reactances, virtual inertias, and 122 adaptive frequency/active power (/P) controls. The pecking order or hierarchy of distributed energy and storage resources can be based on various factors such as energy resource availability, response characteristics, and lifecycle costs of participating DERs. Specific examples of FDERS application for extending the operation of an industrial power system and thereby reducing its mean repair frequency have been demonstrated through the analysis of aging of various batteries in the system. A novel method for balancing cycle life of batteries in industrial microgrids based on FDERS framework has also been presented. This method solves the battery life discrepancy problem for an industrial crusher plant. This method has featured multiple solution approaches to this problem. In Approach A, the DERs were periodically cycled to change their ‘electrical’ positions and this resulted in 48% extension in microgrid operation. In Approach B, the ratio of number of cycles for each Order was made proportional to the ratio of the DER unit’s rated power. This approach resulted in 60% extension. The cycling in Approach C was based on isolating into two levels the higher rated DERs and the lower rated ones, to cycle within the leading two positions and lagging two positions, respectively. The microgrid operation has been extended by 76% using Approach C. Furthermore, Approach D has presented an active cycling where the decision making logic behind the pecking order of DERs was made cycle by cycle depending on their accumulative State-of-Health SoH. These solution approaches have been tested on equal-rated DERs and on a more generalized case of non-equal-rated DERs. Approach D has shown the best results of all and has been able to extend the operation of the microgrid by 80%. 123 Another concept of employing FDERS in an industrial microgrid has been proposed to extend the operation of microgrids that contains DERs with different types of prime- movers. This technique has been applied to change the pecking order by bringing the drafting DER into the leading position after the battery of the leading DER reaches its EoL. Extension of operation results in a complete utilization of the batteries before causing a system stoppage, it has a significant impact on extending the mean replacement time of the industrial power system. In addition, practical considerations of employing FDERS have been discussed. An analysis was performed for different levels of fluctuations in load. It was found that the value of FDERS balancing is greater for an industrial power system containing extremely harsh and larger fluctuating loads, when the batteries get stressed enormously and their cycle life is low. Another practical consideration that has been presented here is to investigate the effect of DERs’ interface reactance in the microgrid. This has revealed that the discrepancy between batteries’ lifetimes increases when the reactances increase. This happens because the drafting effect of ‘electrically’ farther DERs increases transient burden on the leading DER unit. Furthermore, the practical scenarios have been extended to include different technologies like supercapacitors. The operation of this microgrid has successfully been extended by 17%. Another aspect that has been covered in this dissertation is the fleet vehicle-to-grid (V2G) systems. The fleet V2G systems are advantageous because of significantly low utilization of fleet vehicle’s power capacity. In a V2G system the electric-powered vehicles can also deliver power to the grid when plugged-in since their battery utilization ratio is low when 124 they are parked for long periods. The state-of-the-art control strategies implemented for coordinating the charging/discharging of the batteries in V2G systems treated all the electric vehicles alike, when connected to the microgrid without any consideration to their varied driving history and therefore dissimilar battery aging. In contrast, a novel control strategy was proposed for the fleet V2G systems that contain multiple electric vehicles with different driving cycles. It has employed a programmable on-board smart device that employs software agents to observe battery state-of-charge (SoC) and temperature, and has detected its state-of-health (SoH). When the fleet electric vehicles are connected to the charging stations at the commercial office building microgrid, their share of the load demand is based on their latest SoH value. This ensures that a vehicle battery with higher SoH contributes more for the building load as compared to lower SoH battery. It resulted in fair sharing of the load between the different vehicles, and therefore extended the lifetime of the vehicle’s batteries by up to 8%. 8.2 Contributions Evaluated the energy storage systems (ESS) according to their applications o Short-term applications (power applications) – spinning reserve. o Long-term applications (energy applications) – peak shaving. Proposed generalized methodology for ESS sizing in microgrids o Design guidelines for battery sizing in microgrid applications that were not covered well in the prevalent IEEE Standard 485-2010. 125 o Used the energy balance against the current levels in developing improved guidelines. o Considered aging factor and target lifetime into account. o Covered both batteries and supercapacitors. o Design objectives included peak shaving and spinning reserve applications. Conducted quantitative analysis of FDERS o Adopted the necessary mathematical models. o Investigated quantitatively the effect of “virtual” reactance and “virtual” inertia on a microgrid featuring large and fluctuating loads. Conducted qualitative analysis of FDERS o Employed the Quality Function Deployment (QFD) methodology to evaluate the performance of various design variables in FDERS. o Observed that the customer needs can be satisfactorily met by using the design variables of virtual inertia and virtual reactance. Proposed novel approaches to balance cycle life of batteries in industrial microgrids o Balancing strategies to the battery life discrepancy problem were proposed by employing FDERS. o Approach A – Periodic Cycling resulted in 48% life extension. o Approach B – Power Rating-Weighted Cycling resulted in 60% life extension. 126 o Approach C – Power Rating-Levelized Cycling resulted in 76% life extension. o Approach D – Active Cycling resulted in 80% life extension. Discussed practical considerations in the various applications o Performed analysis for different levels of fluctuations in load. o FDERS balancing strategies are of great value for industrial power systems containing extremely harsh and larger fluctuating loads. o Investigated the effect of different DER interface reactances on the FDERS performance. o The discrepancy between batteries’ lifetimes increases when the reactances increase. Analyzed FDERS on Different Types of Prime-Movers o Analyzed the change of the pecking order for a 2-DER system to bring the drafting DER into the leading position after the battery of the leading DER reached its EoL. o Extended the mean replacement time for batteries in the industrial power system. Demonstrated the FDERS on different ESS technologies o Both batteries and supercapacitors were tested with FDERS techniques. o The framework of FDERS was successfully utilized in an industrial microgrid with four fuel cell-supercapacitor DERs. 127 o The operation of an industrial microgrid has successfully been extended by 17%. Load sharing strategy for fleet Vehicle-to-Grid (V2G) systems o Proposed a novel control strategy for fleet V2G system that uses on-board smart devices in fleet electric vehicles to estimate the battery state-of- health (SoH). o The load sharing strategy insures that a fleet vehicle with higher SoH battery contributes more than others to supply the commercial office building load. 8.3 Future Work The work presented in this dissertation opens new research opportunities. A potential project is to extend the sizing methodologies and develop new methodologies for different applications in addition to peak shaving and spinning reserve/frequency regulation. Such applications can include the electric energy time-shift, which involve purchasing inexpensive electric energy during periods when price is low to use or sell the stored energy when the price is high. Further ancillary services like backup supply and transmission support could be considered in sizing. In addition, the sizing practice can include ESS prices to obtain the optimal ESS size at the best price. Additional work on the proposed ESS sizing methodology may be to convert it into a simple worksheet for use by practicing engineers. 128 Another research potential is to build and analyze more configurations of ESS that could enhance the overall performance of microgrids. Similar to solar and wind forecasting methods in renewable sources of energy, a driving cycle forecast would benefit the proposed load sharing strategy for a V2G system and would give better results. 129 REFERENCES [1] Global Wind Energy Council (GWEC), "Global wind report, annual market update 2013," http://www.gwec.net/wp-content/uploads/2014/04/GWEC-Global- Wind-Report_9-April-2014.pdf, 2013]. [2] G. Barbose, N. Darghouth, S. Weaver, and R. Wiser. Tracking the sun VI: An historical summary of the installed price of photovoltaics in the United States from 1998 to 2012 [Online]. Available: http://emp.lbl.gov/sites/all/files/lbnl- 6350e.pdf [3] Greentech Media Inc. and Solar Energy Industries Association, U.S. solar market insight, 2014. [4] R. D. Masiello, K. Vu, L. Deng, A. Abrams, K. Corfee, and J. Harrison, Research evaluation of wind generation, solar generation, and storage impact on the California grid: PIER final project report, CEC-500-2010-010, California Energy Commission, 2010. [5] E. Muljadi, C. P. Butterfield, B. Parsons, and A. Ellis, “Effect of variable speed wind turbine generator on stability of a weak grid,” IEEE Trans. Energy Convers., vol. 22, no. 1, pp. 29-36, 2007. [6] P. Louka, G. Galanis, N. Siebert, G. Kariniotakis, P. Katsafados, I. Pytharoulis, and G. Kallos, “Improvements in wind speed forecasts for wind power prediction purposes using Kalman filtering,” J. Wind Eng. Ind. Aero., vol. 96, no. 12, pp. 2348-2362, 2008. [7] J. D. Maddaloni, A. M. Rowe, and G. C. van Kooten, “Network constrained wind integration on Vancouver Island,” Energy Policy, vol. 36, no. 2, pp. 591-602, 2008. [8] T. M. Weis, and A. Ilinca, “The utility of energy storage to improve the economics of wind–diesel power plants in Canada,” Renewable Energy, vol. 33, no. 7, pp. 1544-1557, 2008. [9] P. Denholm, “Improving the technical, environmental and social performance of wind energy systems using biomass-based energy storage,” Renewable Energy, vol. 31, no. 9, pp. 1355-1370, 2006. [10] E. D. Castronuovo, and J. A. Peas Lopes, “On the optimization of the daily operation of a wind-hydro power plant,” IEEE Trans. Power Syst., vol. 19, no. 3, pp. 1599-1606, 2004. [11] N. J. Schenk, H. C. Moll, J. Potting, and R. M. J. Benders, “Wind energy, electricity, and hydrogen in the Netherlands,” Energy, vol. 32, no. 10, pp. 1960- 1971, 2007. 130 [12] S. Shaw, and E. Peteves, “Exploiting synergies in European wind and hydrogen sectors: A cost-benefit assessment,” Int. J. Hydrogen Energy, vol. 33, no. 13, pp. 3249-3263, 2008. [13] J. P. Barton, and D. G. Infield, “Energy storage and its use with intermittent renewable energy,” IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 441-448, 2004. [14] C. Abbey, and G. Joos, “Supercapacitor energy storage for wind energy applications,” IEEE Trans. Ind. Appl., vol. 43, no. 3, pp. 769-776, 2007. [15] R. Cardenas, R. Pena, G. Asher, and J. Clare, “Power smoothing in wind generation systems using a sensorless vector controlled induction Machine driving a flywheel,” IEEE Trans. Energy Convers., vol. 19, no. 1, pp. 206-216, 2004. [16] Greentech Media Inc. and Solar Energy Industries Association, U.S. energy storage monitor, 2014. [17] N. Jenkins, J. B. Ekanayake, and G. Strbac, Distributed generation: Institution of Engineering and Technology, 2010. [18] D. Pudjianto, C. Ramsay, and G. Strbac, “Virtual power plant and system integration of distributed energy resources,” IET Renewable Power Generation, vol. 1, no. 1, pp. 10-16, 2007. [19] H. L. Willis, and W. G. Scott, Distributed power generation planning and evaluation: Marcel Dekker, Inc, New York, 2000. [20] R. H. Lasseter, "Control of distributed resources," in Proc. Bulk Power Syst. Dyn. Control IV, Santorini, Greece, 1998, pp. 323-329. [21] R. H. Lasseter, "MicroGrids," in Proc. IEEE Power Eng. Soc., 2002, pp. 305– 308. [22] J. Cardell, and R. Tabors, “Operation and control in a competitive market: distributed generation in a restructured industry,” Energy J., pp. 111-136, 1997. [23] M. Donnelly, J. Dagle, D. Trudnowski, and G. Rogers, “Impacts of the distributed utility on transmission system stability,” IEEE Trans. Power Syst., vol. 11, no. 2, pp. 741-746, 1996. [24] R. H. Lasseter, “Smart distribution: Coupled microgrids,” Proc. IEEE, vol. 99, no. 6, pp. 1074-1082, 2011. [25] N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, “Microgrids,” IEEE Power Energy Mag., vol. 5, no. 4, pp. 78-94, 2007. [26] G. Venkataramanan, and M. Illindala, "Microgrids and sensitive loads," in Proc. IEEE Power Eng. Soc., 2002, pp. 315-322. [27] F. Katiraei, and M. R. Iravani, “Power management strategies for a microgrid with multiple distributed generation units,” IEEE Trans. Power Syst., vol. 21, no. 4, pp. 1821-1831, 2006. [28] N. D. Hatziargyriou, and A. S. Meliopoulos, "Distributed energy sources: technical challenges," in Proc. IEEE Power Eng. Soc., 2002, pp. 1017-1022. [29] M. S. Illindala, H. J. Khasawneh, and A. Renjit, “Flexible distribution of energy and storage resources,” IEEE Ind. Appl. Mag., vol. 21, no. 5, 2015. 131 [30] IEEE, "IEEE standard for interconnecting distributed resources with electric power systems," IEEE Std 1547-2003, 2003, pp. 1-28. [31] R. H. Lasseter, "Microgrid: a conceptual solution," in Proc. IEEE PESC, 2004, pp. 4285-4290. [32] J. A. Peas Lopes, C. L. Moreira, and A. G. Madureira, “Defining control strategies for MicroGrids islanded operation,” IEEE Trans. Power Syst., vol. 21, no. 2, pp. 916-924, 2006. [33] J. Eyer, and G. Corey, Energy storage for the electricity grid: benefits and market potential assessment guide, SAND2010-0815, Sandia National Laboratory, 2010. [34] S. B. Peterson, J. F. Whitacre, and J. Apt, “The economics of using plug-in hybrid electric vehicle battery packs for grid storage,” J. Power Sources, vol. 195, no. 8, pp. 2377-2384, 2010. [35] B. Kroposki, C. Pink, R. DeBlasio, H. Thomas, M. Simões, and P. K. Sen, “Benefits of power electronic interfaces for distributed energy systems,” IEEE Trans. Energy Convers., vol. 25, no. 3, pp. 901-908, 2010. [36] B. Bolund, H. Bernhoff, and M. Leijon, “Flywheel energy and power storage systems,” Renew. Sustainable Energy Rev., vol. 11, no. 2, pp. 235-258, 2007. [37] S. Vazquez, S. Lukic, E. Galvan, L. Franquelo, and J. Carrasco, “Energy storage systems for transport and grid applications,” IEEE Trans. Ind. Electron., vol. 57, no. 12, pp. 3881-3895, 2010. [38] J. Stevens, and B. Schenkman, DC energy storage in the CERTS microgrid, 2008. [39] H. Nikkhajoei, and R. Lasseter, “Distributed generation interface to the CERTS microgrid,” IEEE Trans. Power Del., vol. 24, no. 3, pp. 1598-1608, 2009. [40] B. J. Kirby, Spinning reserve from responsive loads, ORNL/TM-2003/19, Oak Ridge National Laboratory, 2003. [41] J. H. Eto, Use of frequency response metrics to assess the planning and operating requirements for reliable integration of variable renewable generation, LBNL- 4142E, Ernest Orlando Lawrence Berkeley National Laboratory, 2010. [42] J. Boyes, Energy storage systems program report for FY99, SAND2000-1317, Sandia National Laboratories, 2000. [43] C. Platt, P. Taylor, L. Charles, and P. Butler, Report on the energy storage systems program executive meetings project, SAND97-2700, Sandia National Laboratory, 1997. [44] D. Conover, S. Ferreira, A. Crawford, D. Schoenwald, and V. Viswanathan, Protocol for uniformly measuring and expressing the performance of energy storage systems, PNNL-22010 Rev. 1., Pacific Northwest National Laboratory (PNNL), Richland, WA (US), 2014. [45] A. Oudalov, R. Cherkaoui, and A. Beguin, "Sizing and optimal operation of battery energy storage system for peak shaving application," in Proc. IEEE Power Tech Conf., 2007, pp. 621-625. [46] C. Parker, “Lead–acid battery energy-storage systems for electricity supply networks,” J. Power Sources, vol. 100, no. 1-2, pp. 18-28, 2001. 132 [47] X. Hu, S. Li, and H. Peng, “A comparative study of equivalent circuit models for Li-ion batteries,” J. Power Sources, vol. 198, pp. 359-367, 2012. [48] H. Chen, T. N. Cong, W. Yang, C. Tan, Y. Li, and Y. Ding, “Progress in electrical energy storage system: A critical review,” Progress in Natural Science, vol. 19, no. 3, pp. 291-312, 2009. [49] M. Armand, and J. M. Tarascon, “Building better batteries,” Nature, vol. 451, no. 7179, pp. 652-657, 2008. [50] H. J. Khasawneh, “Analysis of heat-spreading thermal management solutions for lithium-ion batteries,” Thesis, Mechanical Eng. Dept., The Ohio State University, Columbus, OH, 2011. [51] Y. Hu, S. Yurkovich, Y. Guezennec, and B. J. Yurkovich, “Electro-thermal battery model identification for automotive applications,” J. Power Sources, vol. 196, no. 1, pp. 449-457, 2011. [52] G. L. Plett, “High-performance battery-pack power estimation using a dynamic cell model,” IEEE Trans. Veh. Technol., vol. 53, no. 5, pp. 1586-1593, 2004. [53] F. P. Incropera, D. P. D. Witt, T. L. Bergman, and A. S. Lavine, Fundamentals of heat and mass transfer, 6 ed.: John Wiley & Sons, 2006. [54] D. Linzen, S. Buller, E. Karden, and R. W. D. Doncker, “Analysis and evaluation of charge-balancing circuits on performance, reliability, and lifetime of supercapacitor systems,” IEEE Trans. Ind. Appl., vol. 41, no. 5, pp. 1135-1141, 2005. [55] S. Buller, E. Karden, D. Kok, and R. W. D. Doncker, “Modeling the dynamic behavior of supercapacitors using impedance spectroscopy,” IEEE Trans. Ind. Appl., vol. 38, no. 6, pp. 1622-1626, 2002. [56] S. van der Linden, “Bulk energy storage potential in the USA, current developments and future prospects,” Energy, vol. 31, no. 15, pp. 3446-3457, 2006. [57] R. B. Schainker, "Executive overview: Energy storage options for a sustainable energy future," in Proc. IEEE Power Eng. Soc., 2004, pp. 2309-2314. [58] R. H. Schulte, J. Nicholas Critelli, K. Holst, and G. Huff, Lessons from Iowa: Development of a 270 Megawatt compressed air energy storage project in Midwest independent system operator, SAND2012-0388, Sandia National Laboratories, 2012. [59] S. Nomura, Y. Ohata, T. Hagita, H. Tsutsui, S. Tsuji-Iio, and R. Shimada, “Wind farms linked by SMES systems,” IEEE Trans. Appl. Supercond., vol. 15, no. 2, pp. 1951-1954, 2005. [60] W. V. Hassenzahl, D. W. Hazelton, B. K. Johnson, P. Komarek, M. Noe, and C. T. Reis, “Electric power applications of superconductivity,” Proc. IEEE, vol. 92, no. 10, pp. 1655-1674, 2004. [61] B. Frenzel, P. Kurzweil, and H. Rönnebeck, “Electromobility concept for racing cars based on lithium-ion batteries and supercapacitors,” J. Power Sources, vol. 196, no. 12, pp. 5364-5376, 2011. 133 [62] S. M. Lukic, S. G. Wirasingha, F. Rodriguez, J. Cao, and A. Emadi, "Power management of an ultracapacitor/battery hybrid energy storage system in an HEV," in Proc. IEEE Veh. Power Propulsion Conf., 2006, pp. 1-6. [63] EastPenn. UltraBattery [Online]. Available: www.eastpennmanufacturing.com/ [64] D. Linden, and T. B. Reddy, Handbook of batteries, 3rd ed.: McGraw-Hill Professional, 2001. [65] M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, and R. J. Staniewicz, “Aging mechanism in Li ion cells and calendar life predictions,” J. Power Sources, vol. 97, pp. 13-21, 2001. [66] M. Broussely, P. Biensan, F. Bonhomme, P. Blanchard, S. Herreyre, and K. Nechev, “Main aging mechanisms in Li ion batteries,” J. Power Sources, vol. 146, no. 1, pp. 90-99, 2005. [67] J. Vetter, P. Novák, M. R. Wagner, C. Veit, K. C. Mölle, J. O. Besenhard, M. Winter, M. Wohlfahrt-Mehrens, and A. H. C. Vogler, “Ageing mechanisms in lithium-ion batteries,” J. Power Sources, vol. 147, no. 1, pp. 269-281, 2005. [68] G. Ning, R. E. White, and B. N. Popov, “A generalized cycle life model of rechargeable Li-ion batteries,” Electrochim. Acta., vol. 51, no. 10, pp. 2012-2022, 2006. [69] A. Millner, "Modeling Lithium Ion battery degradation in electric vehicles," in Proc. IEEE CITRES, 2010, pp. 349-356. [70] S. Santhanagopalan, Q. Zhang, K. Kumaresan, and R. E. White, “Parameter estimation and life modeling of lithium-ion cells,” J. Electrochem. Soc., vol. 155, no. 4, pp. A345-A353, 2008. [71] S. Peterson, J. Apt, and J. Whitacre, “Lithium-ion battery cell degradation resulting from realistic vehicle and vehicle-to-grid utilization,” J. Power Sources, vol. 195, no. 8, pp. 2385-2392, 2010. [72] H. J. Bergveld, W. S. Krujit, and P. H. L. Notten, Battery management systems: design by modelling: Springer, 2002. [73] M. Dubarry, V. Svoboda, R. Hwu, and B. Y. Liaw, “Capacity and power fading mechanism identification from a commercial cell evaluation,” J. Power Sources, vol. 165, no. 2, pp. 566-572, 2007. [74] C. Guenther, B. Schott, W. Hennings, P. Waldowski, and M. A. Danzer, “Model- based investigation of electric vehicle battery aging by means of vehicle-to-grid scenario simulations,” J. Power Sources, vol. 239, no. 1, pp. 604-610, 2013. [75] M. Uno, and K. Tanaka, “Accelerated charge–discharge cycling test and cycle life prediction model for supercapacitors in alternative battery applications,” IEEE Trans. Ind. Electron., vol. 59, no. 12, pp. 4704-4712, 2012. [76] IEEE, "IEEE recommended practice for sizing lead-acid batteries for stationary applications," IEEE Std 485-2010, 2011, pp. 1-90. [77] D. Conover, V. Viswanathan, and A. Crawford, Determination of duty cycle for energy storage systems integrated with microgrids, PNNL-23390, Pacific Northwest National Laboratory (PNNL), Richland, WA (US), 2014. 134 [78] B. Saha, and K. Goebel, "Modeling Li-ion battery capacity depletion in a particle filtering framework," in Proc. Annu. Conf. Prognost. Health Manage. Soc., 2009. [79] R. Spotnitz, “Simulation of capacity fade in lithium-ion batteries,” J. Power Sources, vol. 113, no. 1, pp. 72-80, 2003. [80] G. Ning, and B. N. Popov, “Cycle life modeling of lithium-ion batteries,” J. Electrochem. Soc., vol. 151, no. 10, pp. A1584-A1591, 2004. [81] G. Sarre, P. Blanchard, and M. Broussely, “Aging of lithium-ion batteries,” J. Power Sources, vol. 127, no. 1, pp. 65-71, 2004. [82] F. R. Kalhammer, B. M. Kopf, D. H. Swan, V. P. Roan, and M. P. Walsh, Status and prospects for zero emissions vehicle technology, Report of the ARB Independent Expert Panel 2007, Sacramento, California 2007. [83] A. A. Pesaran, T. Markel, H. S. Tataria, and D. Howell, "Battery requirements for plug-in hybrid electric vehicles-analysis and rationale," presented at the Elect. Veh. Symp., Anaheim, CA, USA, 2007. [84] A. Mondal, D. A. Klapp, M. S. Illindala, and J. H. Eto, "Modeling, analysis and evaluation of smart load functionality in the CERTS Microgrid," in Proc. IEEE ECCE, 2014, pp. 4628-4634. [85] A. Renjit, M. Illindala, and D. Klapp, “Graphical and analytical methods for stalling analysis of engine generator sets,” IEEE Trans. Ind. Appl., vol. 50, no. 5, pp. 2967-2975, 2014. [86] tenKsolar. XT-A Module 410W [Online]. Available: http://www.tenksolar.com/ [87] M. Uzunoglu, O. C. Onar, and M. S. Alam, “Modeling, control and simulation of a PV/FC/UC based hybrid power generation system for stand-alone applications,” Renewable Energy, vol. 34, no. 3, pp. 509-520, 2009. [88] Steel Mill Thüringen. [Online]. Available: https://www.gossenmetrawatt.com [89] V. B. Bohorquez, "Fast varying loads," in Proc. IEEE EPQU, 2007. [90] J.-K. Ji, and S.-K. Sul, “DSP-based self-tuning IP speed controller with load torque compensation for rolling mill DC drive,” IEEE Trans. Ind. Electron., vol. 42, no. 4, pp. 382-386, 1995. [91] E. Achenbach, “Response of a solid oxide fuel cell to load change,” J. Power Sources, vol. 57, no. 1-2, pp. 105-109, 1995. [92] L. Gao, Z. Jiang, and R. A. Dougal, “An actively controlled fuel cell/battery hybrid to meet pulsed power demands,” J. Power Sources, vol. 130, no. 1, pp. 202-207, 2004. [93] A. Jossen, J. Garche, H. Doering, M. Goetz, W. Knaupp, and L. Joerissen, “Hybrid systems with lead–acid battery and proton-exchange membrane fuel cell,” J. Power Sources, vol. 144, no. 2, pp. 395–401, 2005. [94] G. L. Plett, and L. Gregory, “Extended Kalman filtering for battery management systems of LiPB-based HEV battery packs: Part 3. State and parameter estimation,” J. Power Sources, vol. 134, no. 2, pp. 277-292, 2004. [95] A. M. Bernardes, D. C. R. Espinosa, and J. A. S. Tenório, “Recycling of batteries: a review of current processes and technologies,” J. Power Sources, vol. 130, no. 1, pp. 291-298, 2004. 135 [96] M. Contestabile, S. Panero, and B. Scrosati, “A laboratory-scale lithium-ion battery recycling process,” J. Power Sources, vol. 92, no. 1-2, pp. 65-69, 2001. [97] P. B. S. Lissaman, and C. A. Schollenberger, “Formation flight of birds,” Science, vol. 168, pp. 1003-1005, 1970. [98] H. Weimerskirch, J. Martin, Y. Clerquin, P. Alexandre, and S. Jiraskova, “Energy saving in flight formation,” Nature, vol. 413, pp. 697-698, 2001. [99] C. R. Kyle, “Reduction of wind resistance and power output of racing cyclists and runners traveling in groups,” Ergonomics, vol. 22, pp. 387-397, 1979. [100] A. G. Edwards, and W. C. Byrnes, “Aerodynamic characteristics as determinants of the drafting effect in cycling,” Med. Sci. Sports Exercise, vol. 39, no. 1, pp. 170-176, 2007. [101] J. Brisswalter, and C. Hausswirth, “Consequences of drafting on human locomotion: benefits on sports performance,” Int. J. Sports Physiol Perform, vol. 3, no. 1, pp. 3-15, 2008. [102] A. J. Wood, and B. F. Wollenberg, Power generation operation and control, 2 ed.: John Wiley & Sons, 1996. [103] M. S. Illindala, “Vector control of PWM-VSI based distributed resources in a microgrid,” Dissertation, University of Wisconsin-Madison, Madison, WI, 2005. [104] G. Venkataramanan, and M. S. Illindala, "Small signal dynamics of inverter interfaced distributed generation in a chain-microgrid," in Proc. IEEE Power Eng. Soc., 2007, pp. 1-6. [105] M. S. Illindala, and G. Venkataramanan, “Small signal stability of a microgrid with parallel connected distributed generation,” J. Intelligent Automation Soft Computing, vol. 16, no. 2, pp. 235-254, 2010. [106] L. P. Sullivan, “Policy management through quality function deployment,” Quality Progress, vol. 21, no. 6, pp. 18-20, 1988. [107] D. H. Besterfield, C. Besterfield-Michna, G. Besterfield, and M. Besterfield- Sacre, Total quality management, 3rd Edition ed.: Prentice Hall, 2002. [108] H. J. Khasawneh, and M. S. Illindala, “Battery cycle life balancing in a microgrid through flexible distribution of energy and storage resources,” J. Power Sources, vol. 261, no. 1, pp. 378-388, 2014. [109] H. J. Khasawneh, and M. S. Illindala, "Equalization of battery cycle life through flexible distribution of energy and storage resources," in Proc. IEEE I&CPS Conf., 2014, pp. 1-9. [110] R. E. Henry, "Cable sizing for fast transient loads," in Proc. IEEE I&CPS Conf., 2003, pp. 18-32. [111] C. Hsu, “Transient stability study of the large synchronous motors starting and operating for the isolated integrated steel-making facility,” IEEE Trans. Ind. Appl., vol. 39, no. 5, pp. 1436-1441, 2003. [112] C. Hsu, C. Chen, and C. Lin, “Electric power system analysis and design of an expanding steel cogeneration plant,” IEEE Trans. Ind. Appl., vol. 47, no. 4, pp. 1527-1535, 2011. 136 [113] A123 Systems Data sheet MD1000001-02 for high power lithium ion battery cell ANR26650M1. [Online]. Available: www.a123systems.com [114] K. Subramanyan, and U. M. Diwekar, “Characterization and quantification of uncertainty in solid oxide fuel cell hybrid power plants,” J. Power Sources, vol. 142, no. 1, pp. 103-116, 2005. [115] H. Misawa, and Y. Hirotake, Solid oxide fuel cell, U.S., 1992. [116] J. Padulles, G. W. Ault, and J. R. McDonald, “An integrated SOFC plant dynamic model for power systems simulation,” J. Power Sources, vol. 86, no. 1, pp. 495- 500, 2000. [117] Y. Zhu, and K. Tomsovic, “Development of models for analyzing the load- following performance of microturbines and fuel cells,” Elect. Power Syst. Res., vol. 62, no. 1, pp. 1-11, 2002. [118] Z. Jiang, L. Gao, M. J. Blackwelder, and R. A. Dougal, “Design and experimental tests of control strategies for active hybrid fuel cell/battery power sources,” J. Power Sources, vol. 130, no. 1-2, pp. 163-171, 2004. [119] Z. Jiang, L. Gao, and R. A. Dougal, “Adaptive control strategy for active power sharing in hybrid fuel cell/battery power sources,” IEEE Trans. Energy Convers., vol. 22, no. 2, pp. 507-515, 2007. [120] A. Renjit, M. Illindala, R. Lasseter, M. Erickson, and D. Klapp, "Modeling and control of a natural gas generator set in the CERTS microgrid," in Proc. IEEE ECCE, 2013, pp. 1640-1646. [121] S. J. Ahn, J. W. Park, I. Y. Chung, S. I. Moon, S. H. Kang, and S. R. Nam, “Power-sharing method of multiple distributed generators considering control modes and configurations of a microgrid,” IEEE Trans. Power Del., vol. 25, no. 3, pp. 2007-2016, 2010. [122] J. K. Steinke, “Use of an LC filter to achieve a motor-friendly performance of the PWM voltage source inverter,” IEEE Trans. Energy Convers., vol. 14, no. 3, pp. 649-654, 1999. [123] Y. P. Chang, and C. J. Wu, "Effect of utility system X/R ratio on harmonic amplification characteristics of a plastic plant," in Proc. IEEE Power Eng. Soc., 2003, pp. 68-72. [124] D. Bernardi, E. Pawlikowski, and J. Newman, “A general energy balance for battery systems,” J. Electrochem. Soc., vol. 132, no. 1, pp. 53-63, 1985. [125] G. Ning, B. Haran, and B. N. Popov, “Capacity fade study of lithium-ion batteries cycled at high discharge rates,” J. Power Sources, vol. 117, no. 1, pp. 160-169, 2003. [126] D. P. Abraham, J. L. Knuth, D. W. Dees, I. Bloom, and J. P. Chrisophersen, “Performance degradation of high-power lithium-ion cells—Electrochemistry of harvested electrodes,” J. Power Sources, vol. 170, no. 2, pp. 465-475, 2007. [127] H. Khasawneh, and M. Illindala, “Supercapacitor cycle life equalization in a microgrid through flexible distribution of energy and storage resources,” IEEE Trans. Ind. Appl., vol. 51, no. 3, 2015. 137 [128] Maxwell Supercapacitors. [Online]. Available: http://www.maxwell.com/products/ultracapacitors/ [129] M. Corley, J. Locker, S. Dutton, and R. Spee, "Ultracapacitor-based ride-through system for adjustable speed drives," in Proc. IEEE PESC, 1999, pp. 26-31. [130] R. H. Lasseter, "Extended CERTS microgrid," in Proc. IEEE Power Eng. Soc., 2008, pp. 1-5. [131] E. Sortomme, M. M. Hindi, S. D. J. MacPherson, and S. S. Venkata, “Coordinated charging of plug-in hybrid electric vehicles to minimize distribution system losses,” IEEE Trans. Smart Grid, vol. 2, no. 1, pp. 198-205, 2011. [132] O. Sundstrom, and C. Binding, “Flexible charging optimization for electric vehicles considering distribution grid constraints,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 26-37, 2012. [133] W. Kempton, and J. Tomić, “Vehicle-to-grid power fundamentals: Calculating capacity and net revenue,” J. Power Sources, vol. 144, no. 1, pp. 268-279, 2005. [134] W. Kempton, and S. E. Letendre, “Electric vehicles as a new power source for electric utilities,” Transp. Res. D, vol. 2, no. 3, pp. 157-175, 1997. [135] W. Kempton, and J. Tomić, “Vehicle-to-grid power implementation: From stabilizing the grid to supporting large-scale renewable energy,” J. Power Sources, vol. 144, no. 1, pp. 280-294, 2005. [136] Z. Wang, X. Sun, and L. Liu, "Study on control strategy of V2G in power peaking," in Proc. Int. Conf. Power Electron. Syst. Appl., 2011, pp. 1-3. [137] N. Hartmann, and E. D. Özdemir, “Impact of different utilization scenarios of electric vehicles on the German grid in 2030,” J. Power Sources, vol. 196, no. 4, pp. 2311-2318, 2011. [138] S. Wencong, H. Eichi, Z. Wente, and C. Mo-Yuen, “A survey on the electrification of transportation in a smart grid environment,” IEEE Trans. Ind. Informat., vol. 8, no. 1, pp. 1-10, 2012. [139] D. P. Tuttle, and R. Baldick, “The evolution of plug-in electric vehicle-grid interactions,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 500-505, 2012. [140] W. Yao, J. Zhao, F. Wen, Y. Xue, and G. Ledwich, “A hierarchical decomposition approach for coordinated dispatch of plug-in electric vehicles,” IEEE Trans. Power Syst., vol. 28, no. 3, pp. 2768-2778, 2013. [141] D. Wu, D. C. Aliprantis, and L. Ying, “Load scheduling and dispatch for aggregators of plug-in electric vehicles,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 368-376, 2012. [142] E. Talebizadeh, M. Rashidinejad, and A. Abdollahi, “Evaluation of plug-in electric vehicles impact on cost-based unit commitment,” J. Power Sources, vol. 248, no. 1, pp. 545-552, 2014. [143] A. Sheikhi, S. Bahrami, A. M. Ranjbar, and H. Oraee, “Strategic charging method for plugged in hybrid electric vehicles in smart grids; a game theoretic approach,” Int. J. Elec. Power, vol. 53, no. 1, pp. 499-506, 2013. 138 [144] A. Marongiu, M. Roscher, and D. U. Sauer, “Influence of the vehicle-to-grid strategy on the aging behavior of lithium battery electric vehicles,” Appl. Energy, 2014. [145] JADE (Java Agent DEvelopment Framework), [Online]. Available: http://jade.tilab.com [146] H. N. Aung, A. M. Khambadkone, D. Srinivasan, and T. Logenthiran, "Agent- based intelligent control for real-time operation of a microgrid," in Proc. Joint Int. Conf. Power Electron. Drives Energy Syst. Power, 2010, pp. 1-6. [147] R. Giovanini, K. M. Hopkinson, D. V. Coury, and J. S. Thorp, “A primary and backup cooperative protection system based on wide area agents,” IEEE Trans. Power Del., vol. 21, no. 3, pp. 1222-1230, 2006. [148] Y. Tomita, C. Fukui, H. Kudo, J. Koda, and K. Yabe, “A cooperative protection system with an agent model,” IEEE Trans. Power Del., vol. 13, no. 4, pp. 1060- 1066, 1998. [149] M. Yilmaz, and P. T. Krein, “Review of battery charger topologies, charging power levels, and infrastructure for plug-in electric and hybrid vehicles,” IEEE Trans. Power Electron., vol. 28, no. 5, pp. 2151-2169, 2013. [150] J. A. Jardini, C. M. V. Tahan, M. R. Gouvea, S. U. Ahn, and F. M. Figueiredo, “Daily load profiles for residential, commercial and industrial low voltage consumers,” IEEE Trans. Power Del., vol. 15, no. 1, pp. 375-380, 2000. [151] K. L. Butler, M. Ehsani, and P. Kamath, “A Matlab-based modeling and simulation package for electric and hybrid electric vehicle design,” IEEE Trans. Veh. Technol., vol. 48, no. 6, pp. 1770-1778, 1999. [152] C. Guille, and G. Gross, “A conceptual framework for the vehicle-to-grid (V2G) implementation,” Energy Policy, vol. 37, no. 11, pp. 4379-4390, 2009. [153] A. Subramanian, M. J. Garcia, D. S. Callaway, K. Poolla, and P. Varaiya, “Real- time scheduling of distributed resources,” IEEE Trans. Smart Grid, vol. 4, no. 4, pp. 2122-2130, 2013. [154] S. Wencong, and C. Mo-Yuen, “Performance evaluation of an EDA-based large- scale plug-in hybrid electric vehicle charging algorithm,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 308-315, 2012. [155] C. Ahn, C.-T. Li, and H. Peng, “Optimal decentralized charging control algorithm for electrified vehicles connected to smart grid,” J. Power Sources, vol. 196, no. 23, pp. 10369-10379, 12/1/, 2011. [156] H. J. Khasawneh, and M. S. Illindala, "State-of-health based load sharing strategy in vehicle-to-grid systems," in Proc. IEEE ITEC, 2014, pp. 1-6. 139