1158 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO.
4, DECEMBER 2016
Coordinated Energy Management in
Resilient Microgrids for Wireless
Communication Networks
Youngsung Kwon, Student Member, IEEE, Alexis Kwasinski, Senior Member, IEEE,
and Andres Kwasinski, Senior Member, IEEE
Abstract This paper proposes a novel approach to power
wireless communication network base stations (BSs) in order to
achieve higher resilience. In the proposed dc microgrid, resilience
is improved with the coordinated operation of renewable energy
sources, energy storage devices, and BS traffic using an integrated
controller that adjusts BS traffic and distributes stored energy
among cell sites based on weather forecast and other operational
conditions. The essential role played by power electronics active
power distribution nodes in this resilient microgrid is also
explained. Resilience improvements are evaluated quantitatively
with metrics analogous to that of availability, which considers
energy storage levels, power generation, and load. The analysis
shows that the proposed system allows reducing service restora-
tion times by a factor of 3 or improve resilience by double-
digit percentage points, while the battery life is extended by
about 10%. Additionally, battery bank size needs and PV array
footprints can be reduced without compromising resilience. The
proposed microgrid is implemented for wireless communica-
tion networks because both energy and communications are
identified in the U.S. Presidential Policy Directive 21 as the
two infrastructures that are especially critical for community
resilience. Nevertheless, this same technology can also be used
in other applications, such as residential neighborhoods or
industrial campuses.
Index Terms Active power distribution nodes (APDNs),
dc microgrids, energy sharing, power electronics, resiliency,
traffic shaping, wireless communications.
N OMENCLATURE
PPD-21 Presidential Policy Directive 21.
BS Base station.
RWCM Resilient wireless communication microgrid.
QoE Quality of experience.
DSM Demand side management.
DR Demand response.
SOC State of charge.
APDNs Active power distribution nodes.
NB Nave Bayes.
Manuscript received December 30, 2015; revised March 13, 2016 and
June 10, 2016; accepted September 11, 2016. Date of publication
September 28, 2016; date of current version October 28, 2016. This work
was supported by NSF under Award 1331788. Recommended for publica-
tion by Associate Editor Sayed Ali Khajehoddin. (Corresponding author:
Alexis Kwasinski.)
Y. Kwon is with The University of Texas at Austin, Austin, TX 78701 USA.
A. Kwasinski is with the University of Pittsburgh, Pittsburgh,
PA 15261 USA (e-mail: akwasins@pitt.edu).
A. Kwasinski is with the Rochester Institute of Technology, Rochester,
NY 14623 USA.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JESTPE.2016.2614425
2168-6777 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
GHI Global horizontal irradiance.
CTF Cycles to failure.
DOD Depth of discharge.
QoS Quality of service.
PSNR Peak signal-to-noise ratio.
LTE Long-term evolution.
C1 Configuration 1.
C2 Configuration 2.
CTF Cumulative distribution function.
MDT Mean downtime.
I. I NTRODUCTION
T HIS paper explores an integrated energy management
controller for a dc microgrid that improves power sup-
ply resilience in wireless communication networks. In this
paper, the concept of resilience is considered based on the
definition provided by the U.S. PPD-21 [1], which indicates
that resilience is the ability to prepare for and adapt to
changing conditions and withstand and recover rapidly from
disruptions. That is, resilience is characterized not only by
how quickly an infrastructure system can recover from a
disruptive event but also by how well it can withstand such
event. Effective planning contributes to improved resilience by
affecting preparedness and adaptation. Recognizing the impor-
tance of communications and energy infrastructures during
natural disasters, we will focus this paper on the power supply
of wireless communication networks. This importance is also
recognized in [1] by identifying energy and communications
systems as uniquely critical due to the enabling functions they
provide across all critical infrastructure sectors. Recently,
there has been increased interest in studying wireless com-
munication network resilience during natural disasters, in
part, motivated by several intense extreme events that in the
past decade affected all corners of the world. Some notable
examples include hurricanes Katrina, Ike, and Sandy in the
USA, and earthquakes and tsunamis that affected Chile, Japan,
and New Zealand. In all these events, important service dis-
ruptions were documented in both power and communication
infrastructures [2][7]. In particular, the two main causes of
communication network disruptions were power supply issues
from extensive electric grid outages and network congestion
due to increased traffic.
In recent past natural disasters, wireless communication
network BSs were especially vulnerable to electric grid power
outages [2], [5], [6]. That is, as Fig. 1 exemplifies, although
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
Fig. 1. Undamaged cell site next to destroyed electric power infrastructure
in the Bolivar Peninsula after Hurricane Ike.
in recent natural disasters a very large percentage of the cell
sites survived the disruptive event undamaged, they lost service
due to issues related to their power supply. Although BSs
are equipped with batteries, usually these batteries provide an
autonomy of 28 h, which is less than the duration of the
long power outages that typically follow a natural disaster.
The conventional solution for powering BSs during these
long power outages is to use diesel gensets. Yet, in many
countries, the BSs are not typically equipped with a permanent
genset, which forces network operators in these locations to
endure the significant logistical burden of deploying portable
gensets [2], [5]. As it is increasingly seen in the USA, even in
the cases of the BS equipped with permanent gensets or other
type of permanent backup power generator, the use of gensets
present some issues during extreme disruptive events [6], [8].
These issues include a relatively high probability of fail-
ure to start [9], low availability when operated for more
than 24 h [10], and the creation of significant logistical effort
in order to keep all gensets fueled and operating.
References [3], [11][13] suggest that one alternative to the
issues presented by conventional backup power plants is to use
microgrids. However, as studied in [12], microgrids need to be
correctly designed in order to be resilient. As explained in [12],
one of the challenges of designing resilient microgrids is to
address the issue that many local power sources depend on a
lifeline that can be affected by the disruptive event or natural
disaster in the same way that conventional power grids are
affected during these events. For example, as it is explained
in [11], microgrids relying on natural gas to fuel microturbines,
engine generators, or fuel cells with local reformers may likely
have their natural gas supply interrupted for an extensive
period of time after an earthquake.
As suggested in [11], [12], and [14][18], because renew-
able energy sources do not require a lifeline to operate, they are
a good alternative to power generation units that do depend on
another infrastructure. However, the use of renewable energy
in wireless communication networks present some challenges.
One of these challenges is that the most popular lifeline-free
1159
renewable energy sources, such as photovoltaic (PV) arrays
and wind turbines, exhibit a variable output. To address this
variable output and maintain a more continuous operation,
PV arrays or wind turbines need to be paired with energy
storage, e.g., batteries, or some other nonrenewable source,
such as a main grid or a generator with an internal combustion
engine. Moreover, renewable energy sources have a relatively
large footprint compared with wireless networks loads. For
example, a typical BS may consume up to about 5 kW/m2
consumed power over a single rack or cabinet, whereas a
typical PV module may generate about 250 W/m2 a footprint
20 times larger than that of the load [2], [14], [15], [18].
Although the relatively large footprints of PV modules may not
be an issue when trying to collocate them at a cell site in a rural
environment, such large footprints are a significant limitation
in city centres where space is limited. Also importantly, system
operators would be concerned with the generated power output
variability, which may cause low power supply availability to
the BSs. Although energy storage can assist in improving foot-
print and output power variability issues [14], [15], in order to
improve availability, it is still necessary to consider the effect
of these operating profiles on battery life [19]. As a result of
these issues, the use of renewable energy to power wireless
communication networks is limited. However, these issues can
be improved if loads, energy storage, and power output from
renewable sources are managed in an integrated way [14], [15],
[18], [20]. BSs power consumption optimization has been
explored in past works. For example, [21] and [22] present
an approach to reduce BSs power consumption by switching
OFF (set to sleep mode) BSs in a cellular network selected
according to traffic conditions. In addition, [23] proposes an
efficient cell zooming technique that adjusts a BS coverage
area so as to balance the traffic load and reduce power
consumption. Various other techniques, such as interference
management and resource management by power control, are
discussed in [24]. These techniques might be helpful when the
BSs are powered through conventional means. However, these
methods are limited in their use to times when the traffic going
through the BS is small enough to avoid network congestion.
This paper discusses a microgrid that is formed by
interconnecting a group of nearby BSs in a common micro-
grid [14], [15], [20], named herein as an RWCM. In order
to address the issues associated with the use of renewable
energy sources in wireless cell sites, BS load, energy stored in
batteries, and power generation output from renewable sources
are controlled in an integrated and coordinated way [15]. This
coordinated control considers that the generation of distributed
renewable sources and stored energy in battery banks is shared
while the traffic loads are controlled by considering quality
of experience (QoE) constraints [25] and battery bank SOC.
Due to the importance of energy storage for the operation of
the proposed RWCM, the focus of this paper is on battery
management. In particular, this paper focuses on the high-
level controller of a hierarchical integrated traffic power energy
management platform based on the predicted renewable energy
and battery bank SOC. The proposed control can be considered
a particular case of a DSM or DR. These techniques have
been discussed in the past for power grids [26], [27] even
1160 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016

considering cases involving battery management in electric

vehicles [28], [29]. However, there are some important differ-
ences between these past works and the one presented in this
paper. One important difference is that these DSM strategies
have operational targets, such as cost and price reductions [30],
that are not necessarily related to the goal of this paper of
improving resilience. Moreover, grid applications, such as in
DSM for air conditioning systems [31], often consider loads
that can be momentarily completely interrupted [32], whereas
in the presented wireless communication system application,
although loads can be reduced, they cannot be completed inter-
rupted. Furthermore, DSM approaches discussed for micro-
grids include dispatchable sources [33], which is not the case
of the renewable energy sources considered in this paper. Fig. 2. Example of an RWCM in which PV arrays are placed where there
Use of microgrids in order to improve resilience has been is sufficient space, which in this case is primarily in a central base station.
proposed in the past [12], [34][36], including [37] focusing Other cell sites with insufficient space for PV arrays are powered through the
dc microgrid.
in particular on dc microgrids and [38] on the use of renew-
able energy sources. However, they do not provide resilience
metrics and, thus, improvements in resilience provided by
microgrids could not be measured. Energy management con-
trollers for microgrids have also been discussed in [39][41],
but these do not focus on operation for improved resilience
and, in addition, [40] and [41] focus only on the power
generation sources without considering loads as an additional
controllable asset. Decentralized control of microgrids has also
been studied in [35][37], [40], and [41], but they do not
integrate load management as part of the proposed hierarchical
control as it is being proposed in this paper. Technologies
applicable to dc microgrids have also been discussed in a
variety of applications [42], [43], some of them even in
nonstationary uses, such as for automobiles [44][46] and
ships [47], but none of these works focuses on resilience.
This paper is structured in the following way. Section II
describes the relevant power distribution and control architec-
tures of the RWCM and summarizes resilience metrics used in
the analysis and a weather forecasting tool considered in the
proposed control strategy. Section III discusses the microgrid Fig. 3. Example of the electric power schematic for a portion of an RWCM
planning approach and explains the proposed control strate- with APDNs.
gies for effective energy management in the RWCM, and
Section IV uses actual weather data to evaluate the concept power outages may last many days, for the resilience analysis
validity. Section V concludes this paper by presenting the main discussed in this paper, it is considered that a connected
findings. power grid is not present. As it is well known, cell sites
are typically equipped with batteries. While in conventional
II. F UNDAMENTAL P LANNING , D ESIGN , communication system power plants these batteries are used
AND O PERATIONS N OTIONS to power a given cell site during a main power grid outage,
Fig. 2 shows an overview of the proposed resilient microgrid in the proposed microgrid, the batteries are used to enhance
that powers a cluster of cellular BSs using a dc power resilience through a coordinated management of power gen-
distribution architecture. This configuration assumes that BSs eration from renewable energy sources and BSs load control
are located relatively nearby, which is consistent with the irrespective of whether a potential grid connection is in service
assumption that the RWCM operates in an urban or a suburban or not. This load control is implemented by considering that
area where the large footprint of renewable energy sources a BSs power consumption depends on the traffic and the
makes it difficult to collocate them at each site and where emitted radio signal power, which can be modified within a
effective energy storage management is a more critical need. reasonable margin without significantly affecting the QoE, i.e.,
Each of these RWCMs is formed by a few BSs (e.g., 6 or 7). the user perception of service quality. Such a QoE depends
Power for the BSs is obtained primarily from PV modules on the type of traffic and on the network QoS, which are
and wind turbines. The RWCM can still have a grid tie as measures of network performance (e.g., data delay, bit rate,
represented in the electrical schematic in Fig. 3. However, and application-specific benchmarks). Still, the QoS can be
since during natural disasters it is expected that electric grid reduced from the best excellent level to values that yield
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
Fig. 4. Example of a basic topology of an APDN.
QoE changes that are near imperceptible for the end user or
modify QoE so that it still is within the range that corresponds
to acceptable performance.
A. Power Distribution Architecture
As it was indicated, Fig. 3 shows an example of a possible
power distribution architecture for the RWCM that intercon-
nects a few nodes, most of them BSs, through a dc microgrid.
In it, power electronic dcdc converters are used to interface
each node with the rest of the microgrid. Although these
converters could be realized with conventional single-input
single-output topologies, a more cost effective and opera-
tionally flexible approach is to use multiple-input multiple-
output (MIMO) bidirectional dcdc converters called APDNs,
exemplified in Fig. 4 and studied in [48][51]. As it was
discussed in these papers, APDNs allow power flow control
in all of their ports, which is a very desirable feature for the
RWCM because controlled power flow among nodes enables a
more effective energy management and operation. Moreover,
APDNs can incorporate embedded energy storage, which,
as described in [48], creates the possibility of improving
availability. Hence, the use of APDNs enables operating each
node with a differentiated availability target, so operation
of more critical nodes, e.g., a BS serving a hospital, can
be prioritized over other nodes in the RWCM. This is an
important feature within the context of realizing a resilient
microgrid, because as will be discussed later, the metrics used
to measure resilience are analogous to those used to evaluate
availability [17]. Furthermore, APDNs are essential system
components in order to implement the energy sharing algo-
rithm explained later in this paper. In order to provide an equal
basis for comparison of all control approaches discussed in
this paper, APDNs are assumed to have ideal power efficiency.
Nevertheless, at the proposed power levels, APDNs, such as
those in Fig. 4, can be designed with high efficiency. Moreover,
the system cabling layout is expected to be designed following
the communication network practices that avoid fully cascaded
power distribution architectures. This design in addition to the
use of a dc power distribution architecture leads to a reduced
1161
Fig. 5. Control structure of the RWCM.
number of conversion stages, which may be even lower than
in conventional wireless cell site backup power plants.
B. Control Architecture
Fig. 5 shows the general architecture of the RWCM energy
management and system control platform. As Fig. 5 shows,
the proposed controller has a hierarchical structure in which
the top level is used to coordinate the operation of all power
sources, energy storage devices, and loads in order to optimize
RWCM operation. At a lower level, local decentralized con-
trollers located at each node (usually at a cell site, although it
is possible to have nodes without a load) are in charge of acting
on its corresponding BS traffic and/or local power generation
or energy storage devices based on the regulation set points
received from the high-level controller. As indicated in [15],
it is assumed that these lower level local controllers operate
based on droop laws so that in the case of a failure in the
communication link with the high-level controller or in the
high-level controller itself, the local controllers are still able
to manage local resources and load at their corresponding
BSs but at a suboptimal level. Since decentralized control
approaches for microgrids in general and for APDNs in
particular have been presented in the past [15], [40], [49],
the focus of this paper is on the higher level controller that
optimizes the operation of the RWCM for enhanced resilience.
C. Planning and Forecasting
Development of a resilient microgrid involving the use of
renewable energy sources requires considering weather fore-
casting both for long-term planning and for shorter term oper-
ation. Anticipating weather profiles when planning an RWCM
configuration serves to ensure that the microgrid will have
sufficient resources, particularly energy storage, in order to
meet availability and resiliency requirements without oversiz-
ing the system. Short-term weather forecasting serves to ensure
that existing resources are used effectively. For example,
two-day-ahead weather forecasting influencing controller
actions serve in an RWCM to keep batteries with adequate
charge levels while anticipating potential future weather con-
ditions that would require more intense use of batteries, while
at the same time reducing chances of early battery failure due
to inadequate usage profiles.
In the proposed RWCM, the controller achieves higher
resiliency by optimizing the use of batteries based on weather
forecasting in combination with the integrated management
1162 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
of BSs power consumption and of power output from renew-
able sources. Power output from each renewable source (solar
and wind) is predicted based on an NB classifier model
presented in [18] and implemented based on the National
Oceanic and Atmospheric Administration (NOAA) weather
information [52]. This weather information is used to antici-
pate GHI and wind speed conditions and the main variables
influencing PV and wind power generation, respectively.
The used NB model [18] is an effective probabilistic classi-
fication algorithm. Based on Bayes theorem, the NB method
performs the classification with the assumption that the fea-
tures (the weather variables) are independent of each other in a
given class (e.g., the value of the GHI). The fundamental idea
of this application of Bayes theorem is that the probability of
a class C {1, . . . , l} can be estimated based on observations
features. The value of l is an integer indicating the class levels,
i.e., the set of values that GHI can take; in this paper, l is
assumed being equal to 100. In order to apply the proposed
algorithm, four weather variables are considered: temperature,
relative humidity, dew point, and sky coverage. In the case
of GHI prediction, the proposed NB model has three steps.
In steps 1 and 2, the training data sets (weather variables)
are partitioned into several hourly subsets and the subsets are
filtered based on the sky coverage variable, respectively, in
order to avoid an overlapping classification and improve the
classification accuracy. In step 3, the NB model applies Bayes
theorem based on the assumption of independent variables as
follows:
P(W1 |C) P(W2 |C) P(W3 |C) P(C)
P(C|W1 , W2 , W3 ) =
P(W1 , W2 , W3 )
(1)
where the weather variables, W1 , W2 , and W3 , represent the
temperature, relative humidity, and dew point, respectively.
Additionally, the prior probability of C is indicated by P(C)
and P(C|W1 , W2 , W3 ). Each conditional probability term of
P(Wi |C) is then estimated with a Gaussian kernel [18], and
finally, the posterior probability of P (C|W1 , W2 , W3 ) in (1)
is updated based on [18]
 is analogous only to the mathematical definition of availability.
GHINB = arg max P(C) P(Wi |C) (2)
C{1,...,l} i availability is computed based on mean uptime and mean down
where GHINB stands for the target class value, chosen to be
the one maximizing the forecast probability as indicated in (2).
On a clear day, the values of GHI over time are deterministic
(a bell-shaped curve), increasing until noon and decreasing
thereafter until sunset. However, on a cloudy day, GHI shows
only a partly deterministic characteristic that is difficult to
be explained with a simple parametric family distribution,
such as a normal distribution N(, 2 ). Thus, in order to
predict GHI more accurately, nonparametric kernel function
is used, which can deal with large variations and, therefore,
achieve a more accurate GHI prediction. In addition, since the
GHI classification is a nonlinear problem, the kernel density
function may be preferable to the estimation of GHI variations.
Wind speed classification could be performed with a similar
methodology. However, contrary to the case of the GHI
prediction that requires using four weather variables, because
meteorological agencies do not produce a forecast for the GHI,
wind speed forecast can be typically directly obtained from the
anticipated weather information. Then the proposed NB model
can be used to establish the relationship between forecasted
wind speed and actual wind power.
D. Resiliency Metrics
In the past, it has been proposed that microgrids can
improve resilience [34][38]. However, such statements were
not supported with quantitatively metrics that allowed assess-
ing by how much resilience was improved or that provided
a way for evaluating different technological alternatives in an
objective way. In this paper, based on the aforementioned def-
inition of resiliency from PPD-21 [1], power supply baseline
resilience B of a BSi.e., the resilience of the proposed
microgrid as seen by each of the BSs in the RWCM without
accounting for the contribution of energy storage devicesis
measured based on a metric analogous to that of availability
as [17]
TU
B = (3)
TU + TD
where TU is the time when the BS is receiving the necessary
power from the microgrid and TD is the downtime when the
BS is not receiving power from the microgrid either due to
failure, damage, maintenance, or repairs. As detailed in [15],
this metric of resiliency analogous to that of availability has
been independently proposed in both [53] and [54]. In these
works, it is explained how TU and TD relate to the attributes
of resiliency indicated in PPD-21, such as withstanding capa-
bility and recovery speed, respectively. Both TU and TD are
evaluated over the entire period of consideration, which is
represented by the sum of TU and TD in the denominator
of (3). In order to consider the effect of renewable energy
sources on availability, TU is associated with the time when the
power generated by the local sources is equal or higher than
the load and TD corresponds to the time when the opposite
condition is observed. Although (3) seems to be indicating an
availability measurement, the baseline resilience metric in (3)
The main difference between availability and resilience is that
time calculated over a very large number of failure and repair
cycles, whereas resilience in (3) may be calculated over one
cycle of potential failure and operation during a disruptive
event or may also be calculated over a very large number of
cycles. Hence, in this paper, the term availability is still used
when considering periods of time sufficiently long, such as
in planning assessments, because in such conditions, metrics
for availability and resilience are indistinct. In the same way,
resilience is defined in an analogous way to availability;
in this paper, fragility F is considered to be analogous to
unavailability. That is, F = 1 B . This definition for
fragility originates in the lexical definitions of both resilience
and fragility. However, it is relevant to indicate that in some
past works [55], the term fragility has been associated with a
failure probability depending on a hazard intensity. Likewise,
it is important to note that the term resilience is distinct
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
from that of resistance. As it is explained in [17] and [56],
resistance is the capacity of withstanding the disruptive
effects of a given event, which is one of the components
of the definition of resilience indicated above. However, this
is not the only component of resilience. Quick recovery when
there is a disruption is another component of resilience, when
indicating that resilience is the ability to ..... recover rapidly
from disruptions. That is, the definition of resilience accepts
the possibility that the system under evaluation experiences
disruptions. Additionally, the concept of resilience indicates
that a more resilient system is the one capable of recovering
faster to those disruptions when they occur.
As it was suggested in the first paragraph of this section,
(3) measures the microgrid power supply resilience as seen by
a BS without accounting for the effect of energy storage in
batteries. When the effect of energy storage on power supply
resilience is also considered, it leads to the definition of a local
resilience L , which is calculated as [17]
Tbank
L = 1 (1 B )e (4)
where is the inverse of the downtime TD for the BS
power supply in (3) and Tbank indicates the battery bank
capacity based on the autonomy it provides to its load. That is,
resilience can be enhanced with the use of local energy storage,
usually in the form of batteries. In other words, local resilience
increases with battery autonomy. Such autonomy extension
can be achieved in the planning stage by allocating larger
battery banks or can be achieved through a more effective
operation by the integrated management of power generation
and load. The latter is the approach explored in the proposed
RWCM discussed in this paper.
E. Battery Bank Life Time Model
Battery life is one important concern for wireless network
operators, particularly when batteries are being charged and
discharged more often than in conventional backup power
plants. When assessing battery life, it is possible to consider
cycle life and calendar life. While cycle life is typically depen-
dent on how charges and discharges are controlled and usage
profiles, calendar life is also dependent on other factors, such
as environmental conditions and, in particular, temperature.
Control of these additional factors is out of the scope of this
paper, so the focus in this analysis is on cycle life and on the
control strategies that can be implemented to limit loss of cycle
life due to unsuitable usage profiles. The benchmarking project
described in [57] defined battery bank stress factors, which
causes an aging process by chemical mechanisms. Based
on [57] and [58], lifetime throughput (kWh) is considered here
in order to evaluate the effect of traffic shaping and power
sharing control on battery bank lifetime. Consider Fig. 6 as
a motivating example for this discussion. Fig. 6 shows the
expected number of CTF and lifetime throughput versus DOD
for a US 250E XC2 battery. The CTF and throughput curves
are estimated from the manufacturer data in [59], which is
measured at a 20-h rate based on nominal 225 Ah and 6 V
references. First, the CTF curve is found by fitting a double
1163
Fig. 6. Expected life cycle and lifetime throughput versus DOD for the
deep-cycle battery model US 250E XC2.
exponential curve [57]
CTF = 1 ea2 x + 3 ea4 x (5)
where i is the fitting coefficients (1 = 3.388 104 ,
2 = 0.223, 3 = 4415, and 4 = 0.025) and x is the
percentage DOD values. Based on the CTF curve, the lifetime
throughput Q batt is calculated as [57]
z
((DODi )(CTFi )(Cn ))
Q batt = i=1 (6)
z
where the subscript i is a trial number, z is the total
number of trial, C T F i is the cycles to failure at DO D i ,
which is determined by rainflow counting method [57], and
Cn is the nominal battery capacity (in this example is
1.35 kWh = 6 V 225 Ahit is worth noting that this
approach for calculating energy capacity in kilowatt hours
implies the assumption of using a simplified battery model
suitable with the scope of this paper). Note in Fig. 6 that
when battery life is measured based on its lifetime throughput,
as done in [57], then the relationship between battery life
and DOD or number of cycles is not monotonic. Even when
for deep discharges it is true that deeper DOD or more
cycles lead to shorter lifetime throughput, for shallow or
moderate depth discharges, it is possible to find higher values
of Q batt for higher DOD. That is, while the CTF curve
shows a clear dependency with the DOD, the curve for Q batt
indicates a weaker dependence on the DOD. For this reason,
HOMER [58] calculates the lifetime of a battery bank by just
accounting for how much energy is cycled through the battery.
Thus, a battery bank life measured in years is calculated as
Q batt N
Lifebank = (7)
Q bank
where N is the number of batteries in a battery bank, Q bank
is the total throughput of the batter bank in a year, and Q batt
is the lifetime throughput of a single battery.
III. I NTEGRATED T RAFFIC P OWER M ANAGEMENT
This section describes how the net power, i.e., PV and
wind power generationless BS power consumption, is managed
depending on predicted available renewable energy and battery
1164 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
bank SOC. Specifically, a real-time cellular traffic shaping
algorithm is investigated in order not only to increase the use
of renewable sources but also to decrease potential excessive
battery bank discharges. For simplicity of calculations, this
paper assumes an RWCM with two BS and the focus is on
the high-level controller.
A. Net Power Generation Model
The net power generated in the RWCM at any given time
is defined as
Pnet [n] = PPV [n] + PWind [n] PBS [n] (8)
where PWind [n] is the power generated by wind turbines,
PPV [n] is the power generated in PV arrays, and PBS [n] is
the power consumed by the BS, all of them considered at time
step n. If the effects of temperature on PV cell performance
are neglected in order to simplify the calculations, the power
generated at a PV module of the RWCM is given by
PPV = (GHI)(A)(PV ) (9)
where A is a total PV module area (m2 ) and P V is the
conversion efficiency. For simplicity in the calculations, it is
assumed that all arrays in the RWCM receive the same GHI
and uses the same PV module, a USA MX60-240 PV module,
which has a size of 1.7 m2 , with an efficiency of 14% and a
maximum power output of 240 W. In order to characterize
PWind , a theoretical turbine wind speedpower output curve is
used [60]. Like the PV modules, it is assumed that all the wind
turbines used in the RWCM are the same: a Bergey BWE10
wind turbine is considered, which composes of a three-blade
horizontal axis (7-m rotor diameter) with a 9-km/h cut-in wind
speed and a maximum power of 10 kW at 41.8 km/h.
For the BS, it is assumed that the RWCM is operating in an
urban area in which each cell cite has a coverage area given by
a circle with a radius of 250 m. As a macro-BS, transmitting
three sectors with 2 2 MIMO (two transmitters and two
receivers) antenna configuration and the transmitting power
of 40 dBm are assumed [61]. In addition, the frequency reuse
factor of 3 and a 10-MHz system bandwidth are assumed as
in [61]. In this paper, a BS power consumption is calculated
using the model from [14], which can shape the traffic at
the load based on the predicted renewable energy and battery
bank SOC. The power consumption is expressed by
PBS [n] =NRT (PT [n] [n]+PB ) (10)
where NT R is the number of active radio transmitters (in this
case is 6; the result of multiplying the number of antennas per
sector, 2, with the number of sectors per BS, 3), [n] is the
normalized traffic profile at the nth time period, PT represents
a factor (in this paper considered equal to 97) that accounts
for the power consumed at the BS depending on the traffic
profile, and PB = 65 W is a constant power term accounting
for pilot signals, cooling, processing, base band interface, and
so on.
The traffic shaping factor [n] controls the volume of
cellular traffic handled by each BS. Settings with [n] < 1
indicate that the traffic through a BS is limited to smaller
levels than when not doing traffic shaping. Limiting the traffic
resulted in increased data delay and real-time video with lower
quality (because it needs to be compressed more), which
results in end users experiencing a controlled reduction in
their QoE. As mentioned earlier, the metric used to measure
QoE depends on the type of traffic. For a real-time video, the
QoE shall measure the playback quality of the video, and for
data traffic, QoE is measured in terms of delay. Reducing the
traffic through the BS with a setting of [n] < 1 results in a
reduction of the QoE to active calls. For a real-time video, a
[n] < 1 reduces the playback quality (measured here as the
PSNR) because video is compressed more. For data traffic,
[n] < 1 increases the delay because it yields longer transmit
times and queuing waiting times. Nevertheless, as is the case
in this paper, by judiciously setting [n] , the QoE can be
reduced in a way that is nearly imperceptible or acceptable
for end users.
Following (10), the power consumed by the BS can be
controlled by setting [n]. The proposed traffic management
improves resilience in two ways. On the one hand, reduced
load makes Tbank in (4) to increase. On the other hand, reduc-
ing BS load by limiting traffic increases Pnet as given in (8).
As a result, the uptime tends to increase when traffic is limited
and the downtime tends to decrease. Limiting traffic may
be beneficial even when applying fixed preset values during
power outages events. Indeed, as discussed in [62], traffic
shaping can achieve important reductions in battery bank size
without reducing autonomy or in PV array footprint without
compromising performance. Although these improvements are
beneficial, it is desirable to realize more flexible controllers
that provide similar advantages without having to continuously
limit traffic. This is the motivation for the controller discussed
next.
B. Traffic Shaping Strategy
The previous paragraphs describe the value of integrated
traffic and energy management in order to improve resilience
while still providing acceptable QoE to network users. How-
ever, network operators are also interested in optimizing oper-
ation to prevent inadequate energy storage sizing or reduced
battery life. Such an optimized operation is achieved through
the RWCM higher level controller. The general controller
strategy based on (8) and (10) is summarized in Table I
and represented in Fig. 7. In this control strategy, a Markov
chain is used in order to estimate the likelihood of batteries
being at each of the possible SOC levels. That is, in the
Markov chain, each state represents a discrete SOC level,
from state #1, representing the discharged state, to state #N,
representing the fully charged state. States #2 to # N 1
represent intermediate SOC levels, usually spaced based on
an equal energy difference of W that can be considered to
equal a power difference  P times a time step period T , which
equals the duration of one control action step or period. Hence,
the battery capacity equals [19]
Cbank = (N 1) P T. (11)
Then, the information about the likelihood of batteries being
at each of the possible SOC levels is used to adjust the load
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
TABLE I
T RAFFIC S HAPING S CENARIO
Fig. 7. Real-time iteration algorithm for a maximizing utilization of energy
storage.
using the traffic shaping factor [n + 1] in order to optimize
battery use while keeping users QoE within acceptable levels.
In this paper, for simplicity, T is considered to be equal to 1 h
although other time step durations can easily be considered
instead.
Fig. 7 represents the real-time iterative process using a
Markov chain model used to choose the control traffic shaping
factor [n + 1]. In order to consider the potential effect
of future weather conditions, this process begins by calcu-
lating the net power profiles up to M hours in the future
(Pnet [n+1], . . . , Pnet [n+M]) based on the proposed NB model
summarized in Section II-C, which also includes estimation
for BS power consumption based on (8) and (10). The
integer Mis determined corresponding to the battery bank
capacity [14], [19] and these profiles are used to calculate
the net power transition probabilities pi . Then, the M-step
net power transition matrix P M , which includes M-hour-
ahead anticipated battery bank SOC transition information is
constructed in order to determine the state probability vector
for a Markov chain in which each state represents a discretized
SOC level
v[n + M] =v[n] P M (12)
1165
where v[n] is the battery bank SOC at time step n. Thus, the
vector v[n + M] represents
v[n + M] = [v 1 [n + M], v 2 [n + M], . . . , v N1
[n + M], v N [n + M]]1N (13)
where each v i [n + M] represents the probability of each of
the SOC states. Once v [n + M] in (13) is calculated, the
maximum probability value in v[n + M] is selected as the
maximum likelihood battery bank SOC at time step n + M.
This value is then used to determine the most suitable value
for the traffic shaping factor in the next control decision step,
[n + 1], in order to prevent excessive battery discharges.
In order to evaluate the value of [n + 1], an algorithm
summarized in Table I is used. For simplicity in the explana-
tion, consider that a BS station is powered only by a PV array
coupled with batteries. Then, the battery energy charge level
at time step n, S[n], is given by

n
S[n] = S[0]+T (PP V [k] NTR (97 [k] [k] + 65)). (14)
k=1
Then, at each decision time n, the high-level controller con-
siders a target change in the energy stored in the batteries,
s  S[n] S[n 1], which is given by
s( [n])  T (PPV [n] NTR (97 [n] [n] + 65)). (15)
That is, s( [n]) = Pnet [n]T . Values of s( [n]) > 0
represent cases when the batteries are charged and
s( [n]) < 0 are cases when the batteries are discharged.
Hence, a large positive setting for s( [n]) corresponds to a
situation when the battery is rapidly charged, likely implying
that traffic is shaped and is significantly reduced or that traffic
is already low due to low network usage at the time. Since
the former of these two cases is the one of interests in terms
of controller design, this is the case considered at this point.
When the former case happens, the power consumption at the
BS is reduced at the expense of reducing video quality and
increasing data delay. Hence, it may be desirable to charge
the batteries at a slower rate so that end user experience is not
affected as much as when the batteries are rapidly charging.
On the contrary, a small value for s( [n]) likely imply
minimal traffic shaping, or no traffic shaping at all. A value
much smaller than zero implies that the batteries may be
discharging excessively fast when no or little traffic shaping
is likely being applied. Thus, a maximum value for s( [n]),
smax , is defined as in [14], representing the maximum value
of Pnet [n + 1] that is set with the maximum traffic shaping
(corresponding to the smallest setting of [n + 1] = 0.3).
This is because the battery SOC is low and the batteries need
to be charged fast or because the batteries are being discharged
excessively fast and so the load would have to be reduced by
applying maximum traffic shaping (in which case smax 0).
Hence, as indicated in Table I, the higher level controller
algorithm first evaluates whether smax is greater than 0 as
an Event 1 trigger condition. Once Event 1 is evaluated,
each control decision of determining [n + 1] can be chosen
considering three cases: when the SOC[n] < SOClim , when
SOC[n] SOCc , and when SOClim SOC[n]<SOCc , in
1166 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
which for simplicity the SOC at time step n of a node is
defined based on a linear assumption as
S[n]
SOC[n] = (16)
Cbank
In these cases, SOClim represents a lower limit for battery bank
SOC that is selected based on various aspects, such as battery
technology, load criticality, and battery replacement cost. That
is, although SOClim does not represent a fully discharged
battery bank, it indicates a condition in which its charge level
is considered at a critically low level. On the other end, SOCc
and all higher SOCs represent SOC levels sufficiently high that
allows operating the microgrid without implementing traffic
shaping.
As shown in Table I, the control strategy when the current
SOC, SOC[n], is below SOClim (case 1) or above SOCc
(case 3) is straightforward. In case 1, the command sent by the
higher level controller is to limit traffic as much as possible,
so [n + 1] takes the minimum set value. The minimum value
for [n + 1] is chosen by considering the subjective nature
of QoE and that communication networks may operate in
conditions where the QoE for end users is reduced to levels
that, while not as good as in normal conditions, they are still
acceptable to end users if the reduction in QoE is transient.
In this paper, the minimum value for [n + 1] was chosen to
be 0.3. This is following the results in [14] that show that for
this minimum setting, video traffic distortion and data traffic
delay would be at the limit of QoE acceptability. In case 3,
when the battery SOC is sufficiently highin this paper, it is
considered that the battery has at least 90% of its maximum
chargethen traffic is not limited and [n + 1] is set equal
to 1. Note that a traffic shaping scheme that avoids setting
[n + 1] to its minimum all the time is needed because end
users would not accept a service with the minimum acceptable
QoE all the time when an operation at better-than-acceptable
QoE is possible. In this way, the control strategy serves the
purpose of reducing the impact of traffic shaping to end
users.
For case 2, the proposed algorithm includes both short-
term (1-time-step) and longer term (M-time-step) strategies.
Consider first when smax > 0 so that batteries are being
charged. In this situation, if the longer term evaluation of the
SOC at time step n + M indicates that the SOC will be above
SOClim , then a 1-step look-ahead traffic shaping strategy based
on a logistic function and introduced in [14] is implemented.
However, if the evaluation of the SOC at time n+M yields that
the SOC[n + M] < SOClim , then the algorithm recognizes that
although within the next time step, batteries are being charged,
in M time steps not only batteries would have discharged
but they would have discharged below SOClim . As a result, a
better strategy than using the logistic function is to limit traffic
more aggressively using the M-steps look-ahead Markov chain
strategy, which makes the battery to be charged faster in an
attempt to prevent the SOC to fall below SOClim in M time
steps. The use of this more aggressive traffic shaping approach
than just using the logistic function introduced in [14] is
consistent with the focus of this paper on resilient operation at
the expense of a reduced user QoE because [14] sought a less
aggressive approach when reducing QoE in order to reduce
environmental impact of powering wireless networks.
However, when smax 0 and batteries are discharged, the
algorithm follows a 1-time-step look-ahead strategy based
on equally spaced intervals in which traffic is limited more
as SOC decreases. This strategy is chosen over the logistic
curve strategy because although both strategies have the same
1-time-step look-ahead horizon, at higher SOCs, the intervals-
based strategy limits traffic more aggressively than the logistic
curve-based strategy, whereas at a lower SOC, both strategies
perform similarly. The Markov-chain-based strategy could also
have been used during discharges, but the intervals-based
strategy was chosen for two reasons. Since the Markov-
chain-based approach is an M-time-step look-ahead strategy,
it tends to limit traffic more aggressively than the 1-step look-
ahead intervals-based strategy. That is, during discharges, the
intervals-based strategy tends to affect QoE in an intermediate
way compared with the effects of the logistic curve strategy
or the Markov chain strategy.
Still, there is one more important reason for the proposed
approach. In reality, the previous discussion assumes perfect
weather forecasting, whereas in reality, weather forecast is
subject to uncertainties that may lead to wrong forecasting.
Such uncertainties and forecasting errors increase as the time
horizon for the forecasting increases. That is, in the Markov
chain model, as M increases, the chances for the controller of
making unnecessarily aggressive or not sufficiently aggressive
traffic shaping decisions also increases. For example, if the
M-time-step look-ahead forecast suggests that the SOC may
fall below SOClim and the observed weather matches the
forecast, then the Markov chain strategy would be desirable.
However, if the weather ends up being better than the forecast,
then QoE would have been unnecessarily reduced. Likewise,
if the forecast anticipate favorable weather conditions so that
SOC[n + M] > SOClim , but the actual weather is worse than
the forecast, then the batteries may end up being discharged
more than if a 1-step look-ahead strategy was used. Hence,
the choice for a method to estimate the weather forecast,
such as the NB approach discussed above, is an important
aspect affecting system operation. Although other weather
prediction methods can be chosen instead of the NB approach
considered here, the NB method has shown to be considerably
more accurate than other methods presented in the past [18].
Moreover, weather data accuracy may present some issues.
In our case, use of weather data from the USA is considered
to present a more challenging design environment than the
one found in other regions, such as Europe, which is broadly
considered to be more accurate. Still, discussion of the forecast
estimation errors is outside of the scope of this paper, and for
the NB method, it has been presented in [18].
Evidently, the key issue involving strategy selection is that
the controller makes a decision about traffic shaping before
the forecast is confirmed or that different weather conditions
are observed. Although applying a more aggressive M-time-
step look-ahead strategy like the one based on a Markov chain
may be more rewarding than a 1-step look-ahead strategy like
the one based on intervals, if the observed weather follows
the forecast. Although these rewards may be greater with
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS 1167

higher values of M, the risk of having the actual weather

missing the forecast increases as the number of look-ahead
steps considered in the strategy increases. Thus, the proposed
controller aims not only at providing some balance between
QoE and resilience, but also at assuming a moderate risk
taking approach by considering the moderately aggressive
1-time-step look-ahead algorithm when smax 0 and a
lower risk 1-time-step look-ahead algorithm (logistic function)
when smax > 0, except in a case when it is predicted that
SOC[n + M] < Slim in which case a more aggressive M-time-
step Markov chain strategy is used. This choice of assuming
a more risky strategy when smax > 0 is explained by the
fact that when the decision for which value [n + 1] should
take is made, Pnet is positive, and thus, the potential impact
of a wrong forecast is more limited than the effects when the
wrong forecast occurs with smax 0 and the batteries are
discharging.
In this paper, the focus is primarily on the Markov chain
strategy under the condition of smax > 0 in Case 2,
because the logistic function was presented in [14] and the Fig. 8. One-day results for traffic shaping strategies in September 4, 2013.
(a) Battery bank SOC. (b) Traffic shaping factor (control factor). (c) Actual
intervals-based strategy is straightforward. The operation of and predicted renewable energy.
the Markov-chain-based algorithm can be exemplified in the
following way. Assume that the SOC of the battery is rep-
between the predicted and observed generated power outputs.
resented through 24 discrete levels from the fully discharged
This prediction uncertainty exemplifies the aforementioned
state #1 to the fully charged state #24. If the energy differ-
detailed reasoning for the choice of strategies in Table I. The
ence between two adjacent SOC levels is 350 Wh and the
achieved improved utilization of energy storage resources has
considered time step is 1 h, then, from (11), the battery bank
important benefits in terms of resilience. Before the details
capacity (Cbank ) is about 8.1 kWh (equal to 23350 W 1 h),
are discussed, it is important to note that battery SOC is
which is consistent with 8 h of battery backupi.e.,
also dependent on generated power from renewable sources.
Tbank = 8 hfor a BS with an average power consumption
As Fig. 8 indicates, SOC tends to increase and traffic tends
of 1 kW. If the battery bank SOC at time step nis represented
to be less limited when the power generated in the renewable
by the vector v[n], then the M-h ahead state vector v[n + M]
energy sources increases. By inspecting Fig. 8(a), it is possible
is calculated by multiplying v[n] by the 24 24 transition
to observe that in this example, TU = 15 and TD = 9. Con-
matrix P M . If the state in v[n + M] with the maximum
sider also the battery bank and BS indicated above. Since the
probability corresponds to an SOC less than SOClim , [n + 1]
SOC with traffic shaping is on average 30 percentage points
will be adjusted (decreased) and the iterative Markov chain
higher than the SOC without traffic shaping, (4) yields that the
process in Fig. 7 is repeated until the state with the maximum
local resilience without traffic shaping is 0.8458; with traffic
probability corresponds to an SOC at least equal to one with
shaping, the local resilience improves to 0.8819. Moreover, the
charge SOClim . Hence, using the real-time Markov chain
extended autonomy represented in Tbank (about 30 percentage
model, [n + 1] is controlled in order to attempt satisfying
points) that is achieved with traffic shaping provides additional
the condition SOC[n + M] > SOClim . If the Markov chain
cushion time to react and solve potential power supply issues
model fails to satisfy this condition, then [n + 1] is set equal
that could be observed in the BS considered in this example.
to 0.3, which represents the maximum traffic shaping.
Thus, the use of the proposed algorithm increases resilience
Consider now Fig. 8, which shows a related example fol-
at a local level.
lowing the above explanation based on the same configuration
of a BS powered by six of the aforementioned PV modules
and one of the above-mentioned wind turbines. Fig. 8(a) C. Energy Sharing Strategy
is obtained based on hourly forecasted renewable energy The previous discussion assumes an independent operation
shown in Fig. 8(c) and considering battery SOC levels for the of all BSs in the RWCM. However, in order to achieve a
cases when traffic is not modified and when traffic is shaped more resilient microgrid, it is desirable that the battery banks
based on the integrated traffic power control algorithm and in the RWCM are controlled so they share their stored energy.
represented in Fig. 8(b). Fig. 8 shows that after several days of An example of a situation when it is desirable to operate each
operation without traffic shaping, the battery bank experiences battery bank so that they share stored energy is observed when
discharges under SOClim , including times when the batteries one battery bank SOC is expected to be lower than the lower
are fully discharged. However, as Fig. 8(a) shows, when the limit represented by SOClim while another battery banks is
traffic shaping algorithm is implemented so [n] follows the expected to have sufficient energy. Another complementary
profile in Fig. 8(b), the battery charge improves on average example is when one battery bank SOC is expected to be
about 30 percentage points. Fig. 8(c) shows the difference higher than the fully charged state while another battery bank
1168 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
TABLE II
P OWER S HARING S CENARIO
is not expected to be fully charged. In these cases, it is
desirable to establish power flows between BSs so that energy
storage is used more effectively in the entire RWCM and not
in one node in particular. APDNs become, then, important
components of this resilient microgrid because they allow for
a flexible and regulated power flow control between specific
RWCM nodes.
Energy sharing is performed only after the traffic shaping
scenarios in Table I are applied. For multiple cell sites, the
average high-level battery bank SOC at time step n + 1 is
defined based on
h
j =1 SOC H, j [n + 1]
SOC H [n + 1] = , 1 h B (17)
h limited.
where B is the number of RWCM nodes with batteries,
SOC H, j [n + 1] represents the expected SOC of a battery
bank that is equal or above a high-charge reference level, and
h indicates the number of RWCM nodes with batteries that
have a charge level equal or above the high-charge reference
level. In this paper, it is assumed that an SOC level of 0.6
or more with respect to the maximum possible charge is the
limit considered for batteries with high SOC. Likewise, it is
possible to define an average low-level battery bank SOC at
time step n + 1 as
l
j =1 SOC L , j [n + 1]
SOC L [n+1] = , 1l B (18)
l the generality of the conclusions. In the following discussion,
where SOC L , j [n+1] represents the expected SOC of a battery
bank that is equal or above a low-charge reference level and l
indicates the number of RWCM nodes with batteries that have
a charge level below the low-charge reference level, which in
this case is considered to be the level associated with SOClim .
Table II shows the energy sharing scenarios that are imple-
mented based on (17) and (18). In Table II, batt is the
battery chargedischarge efficiency and Cbank,L [n + 1] is the
equivalent average capacity of the battery banks that are
anticipated to have an SOC below SOClim at time step n + 1.
Likewise, Cbank,H [n+1] represents the equivalent average
battery bank capacity of battery banks with an expected SOC
above SOC H,lim at time step n + 1. The algorithm first
determines whether SOC H [n + 1] is greater than 1 or not.
That is, whether the majority of the RWCM nodes with SOC
over SOC H,lim would end up overcharged in time step n + 1.
If that is the case (event #2), then the excess energy over the
battery capacity is transferred to nodes with an SOC level
below SOClim . By distributing the potential excess charge in
this way, RWCM with a low battery back uptime would have
its autonomy extended, and thus, according to (4), resiliency
is improved. In order to realize this strategy, APDNs shown
in Figs. 3 and 4 are essential system components because
they allow the excess charge to be distributed to the specific
RWCM nodes with batteries with an SOC below SOClim and
prevent having those charges also distributed to other RWCM
nodes with batteries that could have charge levels between
SOClim and SOC H,lim as it would occur if power flow among
nodes cannot be independently controlled. Hence, APDNs are
key enablers of the proposed approach for enhance resilience.
In case the algorithm determines that SOC H [n + 1] is not
greater than 1 in step n + 1 (event #1), i.e., there may be
some nodes with batteries with an SOC above S H,lim , but
they would not be overcharged in the time step n + 1, then
only the energy necessary to bring batteries with an SOC
below SOClim to SOClim is transferred from the nodes with
batteries with an SOC above SOC H,lim to nodes with batteries
with an SOC below SOClim . Once again, APNDs are essential
system components in order to realize this operational strategy
to control power flows so that energy is shared among the
intended nodes. APDNs are not only essential in order to
enable the energy sharing algorithm. Additionally, APDNs
allow curtailing power generation when all batteries are fully
charged and none of the BS in the RWCM has their traffic
IV. R ESULTS
Two configurations are evaluated and compared
depending on the operational conditions in the RWCM.
In configuration #1 (C1), the power sources are 6 MX60-240
PV modules and an Excel 10-kW wind turbine, whereas in
configuration #2 (C2), the power sources are 20 MX60-240
PV modules. These configurations yield availability values
that are relatively low for communication network standards.
However, they were selected in order to show more clearly
the effect of the use of the energy management algorithm
presented in this paper and because these values do not affect
the size of battery bank is assessed first based on [19] and
considering resiliency targets and its metric analogous to that
of availability [17] introduced in (3) and [53] and [54]. Then,
the real-time battery bank operations are evaluated from the
perspective of resiliency and battery life. To this end, several
assumptions are considered.
1) For the energy storage, a deep-cycle lead acid battery
(US 250E XC 2) with a chargedischarge efficiency
of 85%.
2) For the traffic shaping algorithm, the variables are set
such as SOCc = 0.9, SOClim = 0.3, and min = 0.3,
and for the logistic function from [14], t = 0.4.
The weather data used in the analysis are based on publicly
available information from the NOAA [52] from January 2013
to March 2014 for the city of Austin, Texas, and U.S. data from
January 2013 to July 2013 were used to plan energy storage
needs. Data from August 2013 to March 2014 were used to
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
Fig. 9. Estimated energy harvested for each configuration during the
operational evaluation period.
Fig. 10. Battery capacity (SOC) depending on target power supply
availability.
evaluate the real-time energy management in an RWCM. The
estimated energy harvested on each of these months for each
configuration is shown in Fig. 9. BS power consumption is
calculated based on the average daily traffic profile in Europe
from [61].
The outcome of the planning portion of the analysis is
summarized in Fig. 10. As Fig. 10 shows, availabilityor
resilience as they are two analogous concepts within the
context of the presented discussionimprovement because
increased battery capacity has limits. This important obser-
vation can be interpreted in (4). In this equation, Tbank cannot
be made infinitively long using infinitely large battery banks
because there is a limit to the energy that batteries can receive
during charging due to limited power output from the renew-
able energy sources. This limitation in the charging energy
that batteries can receive from renewable energy sources is
dependent on the capacity of these sources and the weather
and astronomical conditions at the microgrid site. Evidently, as
shown in [12], [17], and [19], higher resiliency can be achieved
by planning more power generation capacity from renewable
energy sources. However, this solution involves a higher cost.
Such a cost could be reduced or even avoided by, instead, man-
aging the load power consumption through the discussed traffic
shaping and energy sharing algorithms. In order to support
1169
Fig. 11. CDF for traffic shaping in Configuration 1.
Fig. 12. CDF for traffic shaping in Configuration 2.
further discussion, a battery bank of 8.1 kWh is considered
for C1 and the one of 9.45 kWh is considered for C2 yielding
availabilities of 0.86 for C1 and 0.69 for C2. Fig. 10 also
shows the advantage of diversifying power sources, already
identified in [12] and [19].
In order to evaluate the real-time battery bank SOC manage-
ment, two cell sites are initially operated independently based
on the strategies in Table I. Then the results of independent
operation are compared with the results when the energy
sharing algorithm in Table II is added to the traffic shaping
strategies in Table I. Figs. 11 and 12 show the SOC CDFs
obtained when three energy management strategies are con-
sidered for each of the configurations: no traffic shaping, with
traffic shaping based on Table I, and maximum traffic shaping
( [n +1] = 0.3 all the time). In general, Figs. 11 and 12 show
that the probability of batteries having a given SOC or lower
decreases as traffic shaping is implemented more aggressively.
Thus, Figs. 11 and 12 make use of the performance metrics
indicated throughout this paper, such as resilience metrics (and
its relationship with the SOC of the system considering loss of
service when batteries are discharged and there is no sufficient
power generation) and QoS changes through traffic shaping,
in order to provide an indication of improvements provided
by RWCMs. Since fragility can be associated with the prob-
ability that batteries are fully discharged, then the resiliency
corresponding to the 8-month operation from August 2013 to
1170 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
TABLE III
C OMPARISON B ETWEEN N O -T RAFFIC S HAPING AND
T RAFFIC S HAPING IN T WO C ONFIGURATIONS
March 2014 without any traffic shaping equals 0.83 for C1 and
0.64 for C2. These values are similar to those obtained when
planning the RWTS using weather data from January 2013
to March 2014. When traffic shaping is implemented based
on the algorithm in Table I, then the resiliency increases to
0.94 for C1 and 0.79 for C2. If the maximum traffic shaping
is applied for the whole test period, the resiliency increases
to 0.97 for C1 and 0.83 for C2, but this aggressive setting
results in maximum reduction in video quality and increase in
data delay. These results for C1 can be compared with those
presented in [20] with the same configuration but using only
the logistic function for the traffic shaping algorithm. Addi-
tionally, [20] was evaluated under more favorable weather and
astronomical conditions because data in [20] were measured
during the summer period at 12 higher in latitude than data
in Fig. 9, which mostly spans fall and winter conditions. The
comparison of the proposed approach here with [20] shows a
significant resiliency improvement from 0.79 in [20, Fig. 5.a]
to 0.94 here confirming the validity of the algorithm in Table I
balancing QoE, resiliency, and risk.
Table III summarizes the results for the no-traffic shaping
approach and for traffic shaping strategy following Table I
and also indicates the expected downtime during the 8-month
evaluation period. As Table III shows, with traffic shaping
based on the algorithm in Table I, the downtime is reduced
from 42 days over an 8-month period to 14 days for C1
and from 86 days over the same period to 50 days for C2.
In terms of resilience and considering a typical electric power
outage lasting a week after a hurricane or an earthquake and
general weather conditions similar to those observed during
the 8-month evaluation period, traffic shaping allows extending
the uptime from 139.4 to 157.9 h for C1 and from 107.5 to
132.72 h for C2. This additional uptime not only simplifies
logistical efforts of network operators and reduces service
recovery time, but also provides operators with additional
time to implement mitigating solutions, such as deployment
of portable diesel gensets, to prevent any loss of service. Such
measures may not be necessary if the proposed configurations
include additional power generation sources from renewable
energy and increased energy storage capacity in batteries. Still,
as indicated, the exemplified configuration was selected in
order to show more clearly the effect that the proposed energy
management controller has on the RWCM. Nevertheless, as
exemplified above, additional requirements in terms of power
Fig. 13. CDF results for energy sharing in RWCM.
TABLE IV
C OMPARISON B ETWEEN N O -T RAFFIC S HAPING AND
T RAFFIC S HAPING IN T WO C ONFIGURATIONS
generation capacity and energy storage are also reduced with
the implementation of the proposed controller.
Table III also shows the battery life estimations for both
configurations when no traffic shaping is applied and when
traffic shaping is applied based on the algorithm in Table I.
As Table III indicates, traffic shaping yields higher CTF,
throughput, and life for both C1 and C2. Other battery
configurations were also evaluated and the results are shown in
Table IV. As Table IV shows, with traffic the shaping battery
life is extended by about 10% irrespective of the configuration.
This result indicates that in the case of sealed lead acid
batteries, which usually have a nominal life of 10 years,
the nominal life is extended by about 1 year or 10% more.
Considering the relatively high cost of batteries compared with
that of other components in the RWCM, a 10% battery life
extension is another important result of the proposed control
approach. Additionally, in all cases when traffic shaping is
applied, discharges are shallower effectively extending Tbank
in (4) and, thus, improving resiliency. It is also observed that
one advantage of diversifying power sources, like in C1, is to
yield shallower discharges than when only one type of source,
such as PV modules in C2, is used.
Fig. 13 and Table V summarize the main observations
when the energy sharing algorithm in Table II is used in
combination with the traffic shaping algorithm in Table I.
As Fig. 13 and Table V indicate that although for C1 improve-
ments when adding the energy sharing algorithm is marginal,
improvements for C2 are substantial. In particular, the energy
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
TABLE V
C OMPARISON B ETWEEN L OCAL T RAFFIC S HAPING AND
E NERGY S HARING IN T WO C ONFIGURATIONS
Fig. 14. Summary of resilience results for configurations C1 and C2.
sharing algorithm allows the improving resiliency for C2 from
0.79 to 0.91, while discharges are shallower, with the average
DOD being reduced from 96.32% to 80.42%. Moreover,
the downtime over an 8-month period improves from 50 to
21 days. In terms of resilience and considering the same
aforementioned equivalent natural disaster scenario in which
an electric power grid outage lasts at least a week, the use of
the energy sharing algorithm and traffic shaping algorithm in
C2 allows improving the uptime from 132.72 to 154 h and thus
improving resiliency metrics. This control strategy applied to
C2 also achieves important improvements in terms of number
of cycles, CTF, throughput, and life of batteries. As indicated,
improvements when the energy sharing algorithm is applied
to C1 are not as significant because average discharges when
implementing traffic shaping to C1 are already relatively
shallow. Hence, improvements with energy sharing with C1 are
relatively minor compared with those with C2 in which dis-
charges without energy sharing are relatively deep, even when
traffic shaping is implemented. It is important to highlight the
role of APDNs in achieving these significant improvements
to C2, because, as explained, APDNs are the key RWCM
components necessary for implementing an effective energy
sharing algorithm.
As indicated, both the traffic shaping and energy sharing
algorithms are implemented by the high-level controller, which
sends the necessary commands to each node to regulate their
1171
operation as given by these algorithms. In case of failure of the
high-level controller or on the communication links between
the nodes and the higher level controller, such algorithms
would not operate. Still, the RWCM will maintain operation
powering the BSs but at the suboptimal regime in which there
is no traffic shaping or energy sharing.
V. C ONCLUSION
This paper has discussed the realization of a dc microgrid
in order to provide a highly resilient power supply to wireless
communication network BSs. In this way, this microgrid
addresses the main sources of communication network dis-
ruptions during natural disasters. The focus of this paper is
on wireless communication networks applications because the
proposed system can be readily used by this industry. The
reason for this direct application is that its use is not dependent
on modifying regulatory or safety codes, because in this
application, the proposed microgrid is a private industrial-type
power distribution grid accessible only by trained personnel.
The RWCM presented here is intended to facilitate the use
of renewable sources in wireless communication networks
and, thus, reduce vulnerabilities affecting resilience by avoid-
ing lifeline dependencies. The microgrid power distribution
architecture uses APDNs, which are power electronic cir-
cuits placed at the nodes of the power distribution grid.
These APDNs are essential components enabling the control
approach proposed in this paper.
In addition to proposing a novel architecture to power
wireless communication cell sites, another fundamental con-
tribution of this paper is addressing issues in the application of
renewable energy sources that limit their application to power
BSs. These limitations of renewable energy sources are large
footprint and variable output. These issues are addressed by
a high-level controller presented in this paper that operates
power generation, energy storage, and load power consump-
tion in an integrated way. This controller uses a combination
of short term, 1-step-aheadlook-ahead strategies, and longer
term, M-time-step look ahead strategy, in order to control
the load so that batteries have their autonomy extended, and
thus, resiliency is increased. The load is controlled by limiting
traffic on the BSs while still maintaining acceptable user QoE.
An M-time-step look-ahead strategy based on Markov chains
using NB weather estimates is also discussed in detail.
The analysis supported by quantitative resilience metrics
and actual BS and weather data indicates that the proposed
control approach improves power supply resilience by extend-
ing battery autonomy and increasing uptimes. Moreover, since
the analysis shows that installed power generation capacity
limits how much resilience can be improved by adding bat-
teries, the proposed control approach allows for improved
resilience without adding more batteries. In the same way,
the analysis explains how the proposed approach allows
achieving the same levels of resiliency than conventional
approaches while reducing the footprint of PV arrays. Addi-
tionally, as Fig. 14 summarizing the analytic results show, the
inclusion of an energy sharing algorithm in the controller also
contributes to improve resiliency and to extend battery life.
1172 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 4, DECEMBER 2016
Future work includes experimental validation using the
Vienna MATLAB LTE-A System Level Simulator, which
allows simulating and measuring performance for a multicell
LTE network. The simulation is detailed up to the point
to be possible to measure the resource allocation for the
transmission of each frame in each BS (eNB).
R EFERENCES
[1] Presidential Policy DirectiveCritical Infrastructure Security
and Resilience, accessed on Dec. 25, 2015. [Online]. Available:
https://www.whitehouse.gov/the-press-office/2013/02/12/presidential-
policy-directive-critical-infrastructure-security-and-resil
[2] A. Kwasinski, Telecommunications outside plant power infrastructure:
Past performance and technological alternatives for improved resilience
to hurricanes, in Proc. INTELEC, 2009, pp. 121.
[3] A. Kwasinski, W. W. Weaver, P. L. Chapman, and P. T. Krein, Telecom-
munications power plant damage assessment for Hurricane KatrinaSite
survey and follow-up results, IEEE Syst. J., vol. 3, no. 3, pp. 277287,
Sep. 2009.
[4] A. Tang, A. Kwasinski, J. Eidinger, C. Foster, and P. Anderson,
Telecommunication systems performance: Christchurch earthquakes,
Earthquake Spectra, vol. 30, no. 1, pp. 231252, Feb. 2014.
[5] A. Kwasinski, Effects of notable natural disasters from 2005 to 2011
on telecommunications infrastructure: Lessons from on-site damage
assessments, in Proc. INTELEC, Oct. 2011, pp. 19.
[6] A. Kwasinski, Effects of hurricanes Isaac and sandy on data and com-
munications power infrastructure, in Proc. IEEE INTELEC, Oct. 2013,
pp. 16.
[7] L. Dueas-Osorio and A. Kwasinski, Quantification of lifeline system
interdependencies after the 27 February 2010 Mw 8.8 offshore Maule,
Chile, earthquake, Earthquake Spectra, vol. 28, no. S1, pp. S581S603,
Jun. 2012.
[8] A. Kwasinski, Lessons from Field Damage Assessments about Commu-
nication Networks Power Supply and Infrastructure Performance During
Natural Disasters With a Focus on Hurricane Sandy, document 11-60,
FCC, Feb. 2013.
[9] D. Robinson et al., Impact of distributed energy resources on the
reliability of a critical telecommunications facility, Sandia Nat. Lab.
Nat. Renew. Energy Lab., Tech. Rep. SAND2006-1277, Mar. 2006.
[10] G. M. Grant et al., Emergency diesel engine generator power system
reliability 19871993, Idaho Nat. Eng. Lab., Idaho Falls, ID, USA,
Tech. Rep. INEL-9510035, Feb. 1996.
[11] A. Kwasinski and P. T. Krein, Telecom power planning for natural
and man-made disasters, in Proc. Int. Telecommun. Energy Conf.
(INTELEC), Sep./Oct. 2007, pp. 216222.
[12] A. Kwasinski, V. Krishnamurthy, J. Song, and R. Sharma, Availability
evaluation of micro-grids for resistant power supply during natural
disasters, IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 20072018,
Dec. 2012.
[13] A. Kwasinski, Quantitative evaluation of DC microgrids availabil-
ity: Effects of system architecture and converter topology design
choices, IEEE Trans. Power Electron., vol. 26, no. 3, pp. 835851,
Mar. 2011.
[14] A. Kwasinski and A. Kwasinski, Role of energy storage in a microgrid
for increased use of photovoltaic systems in wireless communication
networks, in Proc. IEEE INTELEC, Sep./Oct. 2014, pp. 18.
[15] A. Kwasinski and A. Kwasinski, Operational aspects and power archi-
tecture design for a microgrid to increase the use of renewable energy in
wireless communication networks, in Proc. IEEE Int. Power Electron.
Conf. (IPEC ECCE-Asia), May 2014, pp. 26492655.
[16] A. Kwasinski, Examination of power supply options for communication
sites operating in grid-islanded environments, in Proc. IEEE INTELEC,
Sep./Oct. 2012, pp. 18.
[17] A. Kwasinski, Numerical evaluation of communication networks
resilience with a focus on power supply performance during natural
disasters, in Proc. INTELEC, Oct. 2015, pp. 16.
[18] Y. Kwon, A. Kwasinski, and A. Kwasinski, Microgrids for base
stations: Renewable energy prediction and battery bank management
for effective state of charge control, in Proc. INTELEC, Oct. 2015,
pp. 16.
[19] J. Song, V. Krishnamurthy, A. Kwasinski, and R. Sharma, Develop-
ment of a Markov-chain-based energy storage model for power supply
availability assessment of photovoltaic generation plants, IEEE Trans.
Sustain. Energy, vol. 4, no. 2, pp. 491500, Apr. 2013.
[20] A. Kwasinski and A. Kwasinski, Increasing sustainability and resiliency
of cellular network infrastructure by harvesting renewable energy, IEEE
Commun. Mag., vol. 53, no. 4, pp. 110116, Apr. 2015.
[21] E. Altman, M. K. Hanawal, R. El-Azouzi, and S. Shamai (Shitz),
Tradeoffs in green cellular networks, ACM SIGMETRICS Perform.
Eval., vol. 39, no. 3, pp. 6771, Dec. 2011.
[22] Z. Niu, Y. Wu, J. Gong, and Z. Yang, Cell zooming for cost-
efficient green cellular networks, IEEE Commun. Mag., vol. 48, no. 11,
pp. 7479, Nov. 2010.
[23] K. Son, H. Kim, Y. Yi, and B. Krishnamachari, Base station oper-
ation and user association mechanisms for energy-delay tradeoffs in
green cellular networks, IEEE J. Sel. Areas Commun., vol. 29, no. 8,
pp. 15251536, Sep. 2011.
[24] D. Brubaker, Optimizing performance and efficiency of PAs in wireless
base stations: Digital pre-distortion reduces signal distortion at high
power levels, Texas Instrum. Incorporated, Dallas, TX, USA, White
Paper SLWY002A, Feb. 2008.
[25] M. Varela, L. Skorin-Kapov, and T. Ebrahimi, Quality of service
versus quality of experience, in Quality of Experience, S. Moller and
A. Raake, Eds. Switzerland, Springer, 2014, pp. 8596.
[26] M. H. Albadi and E. F. El-Saadany, A summary of demand response
in electricity markets, Electr. Power Syst. Res., vol. 78, no. 11,
pp. 19891996, Nov. 2011.
[27] L. Zhao and B. Zeng, Robust unit commitment problem with demand
response and wind energy, in Proc. IEEE PES General Meeting,
Jul. 2012, pp. 18.
[28] S. Shao, M. Pipattanasomporn, and S. Rahman, Grid integration of
electric vehicles and demand response with customer choice, IEEE
Trans. Smart Grid, vol. 3, no. 1, pp. 543550, Mar. 2012.
[29] Z. Tan, P. Yang, and A. Nehorai, An optimal and distributed demand
response strategy with electric vehicles in the smart grid, IEEE Trans.
Smart Grid, vol. 5, no. 2, pp. 861869, Mar. 2014.
[30] J. M. Lujano-Rojas, C. Monteiro, R. Dufo-Lpez, and
J. L. Bernal-Agustn, Optimum residential load management strategy
for real time pricing (RTP) demand response programs, Energy Policy,
vol. 45, pp. 671679, Jun. 2012.
[31] H. A. Aalami, M. P. Moghaddam, and G. R. Yousefi, Demand response
modeling considering interruptible/curtailable loads and capacity market
programs, Appl. Energy, vol. 87, no. 1, pp. 243250, Jan. 2010.
[32] V. L. Erickson and E. E. Cerpa, Occupancy based demand response
HVAC control strategy, in Proc. 2nd ACM Workshop Embedded Sens.
Syst. Energy-Efficiency Building (BuildSys), 2010, pp. 712.
[33] M. Houwing, R. R. Negenborn, and B. De Schutter, Demand response
with micro-CHP systems, Proc. IEEE, vol. 99, no. 1, pp. 200213,
Jan. 2011.
[34] P. Li, P. Degobert, B. Robyns, and B. Franois, Participation in the
frequency regulation control of a resilient microgrid for a distribu-
tion network, Int. J. Integr. Energy Syst., vol. 1, no. 1, pp. 15,
Jan./Jun. 2009.
[35] C. M. Colson, M. H. Nehrir, and R. W. Gunderson, Distributed multi-
agent microgrids: A decentralized approach to resilient power system
self-healing, in Proc. 4th Int. Symp. Resilient Control Syst. (ISRCS),
Aug. 2011, pp. 8388.
[36] C. M. Colson, M. H. Nehrir, and R. W. Gunderson, Multi-agent
microgrid power management, in Proc. 18th IFAC World Congr., Milan,
Italy, Sep. 2011, pp. 36783683.
[37] L. Che and M. Shahidehpour, DC microgrids: Economic operation and
enhancement of resilience by hierarchical control, IEEE Trans. Smart
Grid, vol. 5, no. 5, pp. 25172526, Sep. 2014.
[38] V. Meyer, C. Myres, and N. Bakshi, The vulnerabilities of the power-
grid system: Renewable microgrids as an alternative source of energy,
J. Bus. Continuity Emergency Planning, vol. 4, no. 2, pp. 142153,
Mar. 2010.
[39] H. Kanchev, D. Lu, F. Colas, V. Lazarov, and B. Francois, Energy
management and operational planning of a microgrid with a PV-based
active generator for smart grid applications, IEEE Trans. Ind. Electron.,
vol. 58, no. 10, pp. 45834592, Oct. 2011.
[40] J. M. Guerrero, J. C. Vasquez, J. Matas, L. G. de Vicua, and M. Castilla,
Hierarchical control of droop-controlled AC and DC microgrids
A general approach toward standardization, IEEE Trans. Ind. Electron.,
vol. 58, no. 1, pp. 158172, Jan. 2011.
[41] V. Nasirian, S. Moayedi, A. Davoudi, and F. L. Lewis, Distributed
cooperative control of dc microgrids, IEEE Trans. Power Electron.,
vol. 30, no. 4, pp. 22882303, Apr. 2015.
[42] B. Patterson, DC, come home: DC microgrids and the birth of the
Enernet, IEEE Power Energy Mag., vol. 10, no. 6, pp. 6069,
Nov./Dec. 2012.
KWON et al.: COORDINATED ENERGY MANAGEMENT IN RESILIENT MICROGRIDS FOR WIRELESS COMMUNICATION NETWORKS
[43] H. Kakigano, Y. Miura, and T. Ise, Configuration and control of a
DC microgrid for residential houses, in Proc. Transmiss. Distrib. Conf.
Expo., Asia Pacific, Oct. 2009, pp. 14.
[44] I. Aharon and A. Kuperman, Topological overview of powertrains for
battery-powered vehicles with range extenders, IEEE Trans. Power
Electron., vol. 26, no. 3, pp. 868876, Mar. 2011.
[45] L. Solero, A. Lidozzi, and J. A. Pomilio, Design of multiple-input
power converter for hybrid vehicles, IEEE Trans. Power Electron.,
vol. 20, no. 5, pp. 10071016, Sep. 2005.
[46] P. Rodatz, G. Paganelli, A. Sciarretta, and L. Guzzella, Optimal
power management of an experimental fuel cell/supercapacitor-powered
hybrid vehicle, Control Eng. Pract., vol. 13, no. 1, pp. 4153,
Jan. 2005.
[47] R. E. Hebner et al., Technical cross-fertilization between terrestrial
microgrids and ship power systems, J. Modern Power Syst. Clean
Energy, vol. 4, no. 2, pp. 161179, Apr. 2016.
[48] M. Kim, A. Kwasinski, and V. Krishnamurthy, A storage integrated
modular power electronic interface for higher power distribution avail-
ability, IEEE Trans. Power Electron., vol. 30, no. 5, pp. 26452659,
May 2015.
[49] M. Kim and A. Kwasinski, Decentralized hierarchical control of active
power distribution nodes, IEEE Trans. Energy Convers., vol. 29, no. 4,
pp. 934943, Dec. 2014.
[50] R. Zhao and A. Kwasinski, Controller analysis for active distribution
nodes in advanced dc power systems, in Proc. IEEE COMPEL,
Jun. 2012, pp. 17.
[51] M. Kim and A. Kwasinski, Distribution interface for microgrid oper-
ation and expansion with local energy management, in Proc. IEEE
PEDG, Jun. 2014, pp. 18.
[52] National Oceanic and Atmospheric Administration (NOAA),
accessed on Mar. 1, 2014. [Online]. Available: http://www.
forecast.weather.gov/MapClick.php?lat=30.19452690400044&lon=-97.
6698695449997#.Vn_wNxV96Cr
[53] Measurement Frameworks and Metrics for Resilient Networks and
Services: Challenges and Recommendations. Heraklion, Greece, Eur.
Netw. Inf. Secur. Agency (ENISA), 2010.
[54] A. Kwasinski, Field technical surveys: An essential tool for improving
critical infrastructure and lifeline systems resiliency to disasters, in
Proc. IEEE GHTC, Oct. 2014, pp. 7885.
[55] M. Panteli and P. Mancarella, The grid: Stronger, bigger, smarter?:
Presenting a conceptual framework of power system resilience, IEEE
Power Energy Mag., vol. 13, no. 3, pp. 5866, May/Jun. 2015.
[56] A. Kwasinski. Quantitative Model and Metrics of Electrical
Grids Resilience Evaluated at a Power Distribution Level, Ener-
gies, vol. 9, no. 2, pp. 127, Feb. 2016. [Online]. Available:
http://dx.doi.org/10.3390/en9020093
[57] H. Bindner, T. Cronin, P. Lundsager, J. F. Manwell,
U. Abdulwahid, and I. Baring-Gould, Lifetime modelling of lead
acid batteries, Ris Nat. Lab., Roskilde, Denmark, Tech. Rep.,
Apr. 2005.
[58] T. Lambert, P. Gilman, and P. Lilienthal, Micropower System
Modeling With Homer. New York, NY, USA: Wiley, 2006,
pp. 379418.
[59] U.S. Battery, accessed on Mar. 1, 2015. [Online]. Available:
http://www.usbattery.com/wp-content/uploads/2015/02/usb_250_group_
web_2015b.pdf
[60] Bergey Wind Power, accessed on Mar. 1, 2015. [Online]. Available:
http://bergey.com/documents/2012/05/excel-10-swcc-summary-report.
pdf
[61] G. Auer et al., Energy Efficiency Analysis of the Reference Sys-
tems, Areas of Improvements and Target Breakdown, document
INFSO-ICT-247733 EARTH, Energy Aware Radio and network
tecHnologies (EARTH) Project deliverable D2.3, 2010.
[62] A. Kwasinski and A. Kwasinski, Tradeoff between quality-of-
service and resiliency: A mathematical framework applied to LTE
networks, in Proc. IEEE Int. Conf. Commun. (ICC), May 2016,
pp. 348353.
1173
Youngsung Kwon (S11) received the B.S. degree
(Hons.) in electrical engineering from Soongsil Uni-
versity, Seoul, South Korea, in 2009, and the M.S.E.
and Ph.D. degrees in electrical and computer engi-
neering from the University of Texas at Austin,
Austin, TX, USA, in 2012 and 2015, respectively.
He is currently with LG Electronics, Seoul, where
he is involved in designing and developing electric
vehicle charging system.
Alexis Kwasinski (S03M07SM14) received
the B.S. degree in electrical engineering from
the Buenos Aires Institute of Technology (ITBA),
Buenos Aires, Argentina, in 1993, the Graduate
Specialization degree in telecommunications from
the University of Buenos Aires, Buenos Aires, in
1997, and the M.S. and Ph.D. degrees in electri-
cal engineering from the University of Illinois at
UrbanaChampaign, Champaign, IL, USA, in 2005
and 2007, respectively.
From 1993 to 1997, he was with the Telefonica
de Argentina, Buenos Aires, designing and planning telephony outside plant
networks. Then he was with Lucent Technologies Power Systems (later Tyco
Electronics Power Systems) for five years as a Technical Support Engineer and
a Sales Technical Consultant in Latin America in Mesquite, TX. He was also
a Part-Time Instructor and was in charge of the ITBA Telecommunications
Laboratory for three years Buenos Aires, Argentia. From 2007 to 2014,
he was a Faculty Member with the University of Texas at Austin, Austin,
TX, USA, where he reached the rank of Tenured Associate Professor. He
is currently an Associate Professor and a R. K. Mellon Faculty Fellow
with the University of Pittsburgh, Pittsburgh, PA, USA. His current research
interests include power electronic systems, distributed generation (microgrids),
renewable and alternative energy, smart grids, and the analysis of the impact of
natural disasters on critical power infrastructure, which included site damage
assessments after several natural disasters in Japan, New Zealand, the U.S.,
and Chile.
Dr. Kwasinski was a member of the Executive Committee of the Argentine
Electrotechnical Association in 1994 and 1995. He was a recipient of
the Joseph J. Suozzi INTELEC Fellowship in 2005, the Best Technical
Paper Award at International Communications Energy Conference in 2007,
the National Science Foundation CAREER Award in 2009, and an IBM
Faculty Innovation Award in 2011. He is an Associate Editor of the IEEE
T RANSACTIONS ON E NERGY C ONVERSION.
Andres Kwasinski (S98M04SM13) received
the Diploma degree in electrical engineering from
the Buenos Aires Institute of Technology, Buenos
Aires, Argentina, in 1992, and the M.S. and Ph.D.
degrees in electrical and computer engineering from
the University of Maryland at College Park, College
Park, MD, USA, in 2000 and 2004, respectively.
He is currently an Associate Professor with the
Department of Computer Engineering, Rochester
Institute of Technology, Rochester, NY, USA. He has
co-authored over 60 publications in peer-reviewed
journals and international conferences, and the books Cooperative Com-
munications and Networking (Cambridge University Press, 2009) and 3-D
Visual Communications (Wiley, 2013). His current research interests include
smart infrastructures, with an emphasis on the use of renewable energy,
cognitive radios and wireless networks, cross-layer techniques in wireless
communications, multimedia communications, and networking and operation
of the Internet-of-Things.
Dr. Kwasinski is an Area Editor of the IEEE Signal Processing Magazine,
and an Editor of the IEEE T RANSACTIONS ON W IRELESS C OMMUNICA -
TIONS and the IEEE W IRELESS C OMMUNICATIONS L ETTERS .