This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

Utilization of Supercapacitors in Protection

Schemes for Resiliency against Communication Outages:
A Case Study on Size and Cost Optimization
Hany Fawzy Habib, Mohamad El Hariri, Ahmed Elsayed, Student members, IEEE and Osama Mohammed, Fellow IEEE

Abstract— Adaptive protection techniques used for a microgrid In [13]-[15], central protection approaches are proposed to
rely on a stable communication link to and from protective devices, calculate the fault current of the distributed energy resources
at the point of common coupling (PCC), to adjust the settings (DERs). This system can be used to monitor a microgrid over
within these corresponding devices for either a grid-connected or
communication lines and react to dynamic changes of the sys-
islanded mode of operation. However, during communication out-
ages or in the event of a cyberattack, relays’ settings are not tem. The authors in [16] change the time-current characteristics
changed. Thus adaptive protection schemes are rendered unsuc- for short circuit and overload conditions by observing the differ-
cessful. Due to their fast response, supercapacitors, which are pre- ence in voltage drops during these two events respectively.
sent in the microgrid to feed pulsed loads, could also be utilized to Nonetheless, a key disadvantage of the formal is the fact that
enhance the resiliency of adaptive protection schemes against com- these approaches are completely reliant on exchange or transfer
munication outages. Proper sizing of the supercapacitors is there- of data/information in the shape of measured system parameters
fore important in order to maintain a stable system operation and
between different protection devices via some kind of commu-
the cost of the protection scheme. This paper presents a two-level
optimization scheme for minimizing the supercapacitor size along nication link [17], [18]. The challenge is in the fact that in an
with optimizing its controllers’ parameters. The latter will lead to extremely interconnected cyber-physical system, such as adap-
a reduction of the supercapacitor fault current contribution and an tive protection in microgrids, minor unconventionalities in the
increase in that of other AC resources in the microgrid in the ex- cyber domain can have catastrophic impacts in the physical do-
treme case of having a fault occurring simultaneously with a pulsed main [19]-[21]. Therefore, the goal of a resilient adaptive pro-
load. It was also shown that the size of the supercapacitor can be tection algorithm is to ride through communication failures
reduced if the pulsed load is temporary disconnected during the without serious damage to hardware assets, or financial losses.
transient fault period. Simulation results showed that the resulting Considering the above factors an adaptive protection scheme
supercapacitor size and the optimized controller parameters were
for AC microgrids, which is capable of surviving communica-
feeding enough fault currents for several types of faults in different
locations and minimizing the cost of the protection scheme. Exper- tion failures by the aid of supercapacitor, is a viable solution for
imental results on a hardware microgrid setup validated the simu- enhancing the resiliency of adaptive protection to communica-
lation results. tion outages. In the normal case, the PCC senses a microgrid’s
shift to islanded mode and thus issues a control command to the
Index Terms— control; microgrid; optimization; protection; su-
corresponding relays to shift their settings from high (which is
percapacitor.
between 6-7 times of the rated current value) to low (which is
I. INTRODUCTION between 1.5-2.5 times of the rated current value). However, dur-
ing communication outages, this process is not possible and thus
icrogrids are gaining increasing attention as an important
M part of the smart grid due to their numerous benefits and
relay settings remain high in islanded mode of operation [22].
Therefore in such situations, due to their fast responses, su-
their ability to operate both in islanded and in grid-connected
percapacitor can be utilized to contribute to the fault current and
modes [1]-[3]. While such a diverse deployment of microgrids raise this current value to a level which is sensed by the high
provide some important advantages, they introduce key chal- relay settings. However, attention must be paid as not to over-
lenges in terms of protection. Microgrids are dynamic entities size the utilized supercapacitor and therefore incur additional
where distributed generation (DG), loads, and Energy Storage costs on the overall system. As such, this paper extends our pre-
Devices (ESDs) are constantly connected and disconnected [4]- vious work [23] by proposing a two-level optimization scheme
[6]. Operating conditions vary rapidly, since the inertia of the for minimizing the supercapacitor size along with optimizing its
system is much less than that of conventional networks. Weather controllers’ parameters to reduce its fault current contribution
conditions, resource availability and consumption affect the op- and increase that of other AC resources in the microgrid. In or-
erating conditions of the equipment [7]-[9]. For safe operation der to ensure that the optimization process will converge to a
of the system, all elements must be monitored and necessary sufficient supercapacitor size, this paper addresses the extreme
changes must be made to their settings as the operating condi- case where a three-phase-to-ground fault occurs during the turn-
tions of the microgrid change. In order to protect AC microgrids on period of a pulsed load. The work in this paper also shows,
in both grid-connected and islanded modes of operation, an in a case study, that the size of the supercapacitor can be reduced
adaptive protection scheme is required [10]-[12]. if the pulsed load is temporary disconnected during the transient
fault period. All investigated cases, in simulation and in experi-
This work was partially supported by grants from the Department of En- mental verification, showed that the optimization process did in-
ergy (DoE). The authors are with the Energy Systems Research Laboratory,
Department of Electrical and Computer Engineering, Florida International deed reduce both the size and the cost of the supercapacitor bank
University, Miami, FL 33174 (e-mail: mohammed@fiu.edu). as well as maintain stable system operation.

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

Grid
R :- Relay
TL :- Transmission Line
CB :- Circuit Breaker
Transformer
F :- Fault
:- Fault Symbol Bus 1

R1 / CB1
F1 TL1
R2 / CB2 Bus 2
R5 / CB5 R3 / CB3

TL3 TL2
R6 / CB6 R4/ CB4
Bus 5 Bus 3

CB 15 Circuit R9 / CB9 Circuit R7 / CB7

CB 12 CB 11 CB 13
breaker breaker

F3 TL5 F2 TL4
AC filter 2
AC filter 1
Load 3 Load 1
R10 / CB10 R8 / CB8
Bus 6 Bus 4
Converter 2 Converter 1
CB 16 CB 14
Pulse
G Load G
Super
DER Capacitor
Load 4 Generator 2 Generator 1 Load 2

Fig.1. Configuration of the microgrid under study. III. PROPOSED PROTECTION ALGORITHM
II. HYBRID AC/DC MICROGRID DESCRIPTION The logic algorithm of the relays during the different modes
of operation under study, namely grid connected mode, islanded
Fig.1 shows the topology of the hybrid AC/DC microgrid un- mode with communication, and islanded mode with loss of com-
der study. The system has grid connection capabilities and con- munication is shown in Fig. 2 and can be described in the fol-
tains two ac generators of 7.5 KVA, 60 HZ, 208 V and 1800 lowing equations.
RPM synchronous machines coupled to individual induction Equation (1) gives the operation of the logic circuit of the re-
motors as prime movers.
lay at trigger high (𝐼𝑇𝐻 ) that represents high settings, when the
A three phase 7.5 KVA Δ/Yg transformer is implemented to
system operates at grid connected or at islanded mode with com-
feed the AC microgrid with the required rated voltage. The sys-
tem has a filter located between the transformer and the AC mi- munication failure.
crogrid to filter out the harmonics of the AC grid with induct- (𝐼𝑓 ≥ 𝐼𝐺 + 𝐼𝑀 ) + (𝐼𝑓 ≥ 𝐼𝑆𝐶 + 𝐼𝑀 )
1
𝐼𝑇𝐻 = { (1)
ance of 4 mH. 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Another AC filter is added between the AC and DC parts of where 𝐼𝑓 is the fault current, 𝐼𝐺 , IM and 𝐼𝑆𝐶 are the current con-
the microgrid to improve the performance of the converter 1 and
tributions from grid, microgrid, and supercapacitor, respec-
2 and reduce the harmonics of the AC microgrid as well.
The supercapacitor bank is 2.9 F with rated voltage of 320 V tively.
to feed a 12-Ω DC resistive pulsed load. Equation (2) describes the logic of the relay at islanded mode
A DER is also used to inject current to the AC side and help of operation when the communication is available in the system.
the generators feed different loads that are connected with circuit
If Grid & Islanded
breakers at this side. Mode with Loss of
Converter 1 that is connected between the supercapacitor and Communication
IG
the AC side is used for charging the supercapacitor in the normal
Isc
operation and the supercapacitor injects current to the AC side
during fault condition in islanded mode of operation when the ITH
Trip Signal
communication is not available. ITL
Converter 2 is present to allow the DER to support the AC IM
sources in feeding various loads in the microgrid. A relay and a Ic
Islanded Mode with
Communication
circuit breaker is connected at each end of all transmission lines.
𝐶𝐵11 and 𝐶𝐵12 are connected at the terminals of the DC sources
and provide the ability to connect and disconnect these sources. Fig.2. The logic diagram of the relay.

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

In this case, the relay will produce (𝐼𝑇𝐿 ) and will be adjusted LA Battery LI Battery
Energy Energy
to low settings.
1 (𝐼𝑓 ≥ 𝐼𝑚 ) . 𝐼𝑐
𝐼𝑇𝐿 = { (2)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
where 𝐼𝐶 is the communication signal issued from the PCC to
the relay, adjusting its settings.
As indicated in equation (3), the relay will send a trip signal
Power Power
to the circuit breaker when either (𝐼𝑇𝐻 ) or (𝐼𝑇𝐿 ) is activated:
FW SC
Energy Energy
Trip signal = 𝐼𝑇𝐻 + 𝐼𝑇𝐿 (3)

IV. SUPERCAPACITORS FOR RESILIENCY ENHANCEMENT

A. Supercapacitor Energy Storage
For a microgrid with only one type of energy storage, tasking
it to assist in feeding the fault is more complicated. First, a sys-
Power Power
tem which only contains a lead acid or lithium ion batteries
would be limited to their lifespans [24]. In addition, these bat- Fig.3. Energy Storage Performance Snapshot.
teries are limited by excessive operational currents and particu-
larly temperature. The two most suitable devices, if adequately B. The Supercapacitor’s Sizing Process
sized, are the supercapacitor and flywheel, as they provide the A two-level optimization process for proper sizing of the su-
greatest balance which can be visualized in Fig. 3. The flywheel percapacitor in the system under study is proposed, as shown in
energy storage provides the highest power density, but has a Fig. 4. In Level 1, the main objective is to reduce the cost of the
very high self-discharge rate as a result of friction losses and low proposed protection scheme by selecting the value of the super-
energy density. The response time is very fast, limited only by capacitor. In level 2, the main of objective of the optimization
the initial inertia required to start moving the rotating mass. For process is to minimize the supercapacitor’s fault current contri-
the maintenance issue that is represented a big problem for the bution and maximize that of the distributed generators. This is
microgrid system, supercapacitor and flywheel are the primarily achieved by optimizing the associated frequency, AC voltage,
devices and do not need periodic maintained. They offer a sig- and DC voltage controllers. Fig. 5 shows the block diagram of
nificant improvement in their capability to source high current the three PI controllers regulating the frequency, DC voltage and
as their lifespans are much longer and they have low suscepti- AC voltage of the system. As can be seen in Fig. 5, the error
bility to lifespan issues. However, the supercapacitor has the signals for the frequency and DC voltages (𝑓𝑒𝑟 and 𝑉𝑑𝑐_𝑒𝑟 ) are
best response time of the energy storage types and their con- fed into the first and second PI controllers with parameters 𝑘𝑝𝑓 ,
struction is chemical in nature, no reaction takes place. This 𝑘𝑖𝑓 and 𝑘𝑝_𝑣𝑑𝑐 , 𝑘𝑖_𝑣𝑑𝑐 , respectively.
helps to respond extremely fast to a demand heavy current. The
capacitance and energy density of a supercapacitor are thou- These parameters are used to regulate the direct current com-
sands of times larger than a conventional electrolytic capacitor. ponent of the system (𝐼𝑑_𝑟 ), which in its turn regulates the fre-
A supercapacitor is known to have high power density, low in- quency and DC voltage of the supercapacitor bank. Using the
ternal resistance, and high life cycle [25]-[27]. High power en- AC side frequency as a signature to the active power flow allows
ergy storage is commonly used in many systems such as tele- the supercapacitor to supply large currents during the fault and
communications, shipboard and spacecraft power systems. In contribute to the fault current during loss of communication sit-
these applications, there are various types of loads that show uations. The AC voltage error signal (𝑉𝑎𝑐_𝑒𝑟 ) is given as input to
high instantaneous power requirement but require a relatively the third PI controller with 𝑘𝑝_𝑣𝑎𝑐 and 𝑘𝑖_𝑣𝑎𝑐 parameters that used
low average power. These types of loads are referred to as pulsed to regulate the quadratic current component of the system (𝐼𝑞_𝑟 ),
loads. The duration of such loads ranges typically from hundreds and regulates the AC voltage.
of milliseconds to seconds with different power levels. These Based on the inverse Park transformation the reference three
pulsed loads could be present on ship power systems, which can phase currents (𝐼𝑎_𝑟 , 𝐼𝑏_𝑟 , and 𝐼𝑐_𝑟 ) are estimated from the 𝑑𝑞0
be electrically treated as islanded microgrids as indicated in rotating reference frame currents. The angle between the two
[28], [29]. Additionally, pulsed loads do exist in regular power reference frames and the system frequency are estimated from
system microgrids (other than ships). These pulsed loads could three-phase Phase Locked Loop block.
be large cranes, radar systems, welding machines or an electric The abc reference currents are compared with the actual
vehicle park with simultaneous fast charging of large number of measured currents and applied to the pulsed width modulation
vehicles [30]-[33]. In microgrid applications, the high instanta- scheme to generate the switching signals of the inverter. The
neous power of a pulsed load can cause considerable power dis- hysteresis band current control is considered in this work be-
turbances and thermal problems. Introducing a fast response cause of its simplicity of implementation and fast response cur-
high power storage system, such as a supercapacitor bank, with rent loop. It also doesn’t need any knowledge of the load param-
the appropriate power electronic and control, will improve the eters.
performance of the system drastically [34], [35].

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

The proposed algorithm is an iterative one, which is also

combined with a Particle Swarm Optimization (PSO) routine. [Cmin C2 Ci Cmax]
On the first level, the process first starts by selecting a set of
possible values of capacitances, Ci’s, such that 𝐶𝑚𝑖𝑛 < 𝐶𝑖 <
𝐶𝑚𝑎𝑥 , based on physical constraints of the microgrid under 1. Define Lower (LB) and Upper (UB) Bounds for Kp’s and Ki’s
study. A predefined 𝛥𝐶 = 0.2 𝐹 was selected in our analysis. 2. Set constraints on system frequency, AC, and DC voltages to
It’s worth noting that selecting 𝛥𝐶 requires a tradeoff between ensure stable operation
the optimizations accuracy and the processing time. It was found 3. Generate random swarm
that 𝛥𝐶 = 0.2 𝐹 was an adequate value that preserves the opti- 4. Evaluate value of Ic for entire population
mization’s accuracy and does not incur excessive processing 4.1 If (minimum Ic is reached & constraints are satisfied)||
time. On the second level, for every capacitance value, 𝐶𝑖 the max iterations reached, then exit
PSO routine is started in order to optimize the associated fre- 4.2 Else, update swarm velocities and positions and go to
quency, AC, and DC voltage controllers. step 4
5. Save Kp’s and Ki’s
As mentioned earlier, the controllers in this work are PI con-
trollers, and thus, the parameters to be optimized are the 𝐾𝑝 ′𝑠
and 𝐾𝑖 ′𝑠 for each controller. For the PSO routine, a search space
is randomly generated by defining a population of varying com-
binations of candidate 𝐾𝑝 ′𝑠 and 𝐾𝑖 ′𝑠 for the aforementioned con-
trollers. In order to evaluate the fitness function, the PSO routine Repeat the process for every value of C and select minimum
value of C from the predefined interval
was interfaced with a Simulink model of the power microgrid
with its controllers to evaluate their response to the different par- Fig. 4. The Combined iterative optimization process for supercapacitor sizing
ticles in the generated swarm and thus retrieve the capacitor’s
current fault contribution.
The population generation process is bounded by a vector of Super
lower bounds (LB) and upper bounds (UB) for each controller. AC System Capacitor_
+
Proper definition of the lower and upper bounds is imperative
for the success of the optimization process. Va
Table I, shows the limits of the search space for the parame- Vb Bi-directional
ters to be optimized. During the entire simulation, the solution Vc Converter
of the optimization processes is bounded by constraints on the PLL
system frequency, AC voltage, and DC voltage in order to en- Sa Sb Sc
fm Ia
sure stable operation of the microgrid. _ fer Ib Switching
fr + PI
PI11 + logic
In fact, two set of constraints were imposed: one during Ic
steady state (ss) operation and the other during the pulsed load Vdc_m _ θm Ia_r _ Ia_er
V Id_r
and fault condition, since the acceptable limits for these con- Vdc_r + dc_er PI
PI21 + +
straints are not the same at normal and abnormal conditions as
dq0 Ib_r _ Ib_er
Vrms_m _ V +
indicated in equations (4-9). rms_er Iq_r _
Vrms_r + PI 3 abc Ic_r Ic_er
+
𝑓𝑚𝑖𝑛1 < 𝑓𝑠𝑠 < 𝑓𝑚𝑎𝑥1 (4)
𝑘𝑝_𝑓 𝑠+𝑘𝑖_𝑓 𝑘𝑝_𝑉𝑑𝑐 𝑠+𝑘𝑖_𝑉𝑑𝑐 𝑘𝑝_𝑉𝑎𝑐 𝑠+𝑘𝑖_𝑉𝑎𝑐
PI 1: PI 2: PI 3:
𝑓𝑚𝑖𝑛2 < 𝑓𝑓𝑎𝑢𝑙𝑡 < 𝑓𝑚𝑎𝑥12 (5) 𝑠 𝑠 𝑠
Fig. 5. Converter control logic block diagram
𝑉𝑎𝑐𝑚𝑖𝑛1 < 𝑉𝑎𝑐𝑠𝑠 < 𝑉𝑎𝑐𝑚𝑎𝑥1 (6)

𝑉𝑎𝑐𝑚𝑖𝑛2 < 𝑉𝑎𝑐𝑓𝑎𝑢𝑙𝑡 < 𝑉𝑎𝑐𝑚𝑎𝑥2 (7) V. RESULTS AND DISCUSSION

In this section, multiple case studies were performed in order
𝑉𝑑𝑐𝑚𝑖𝑛1 < 𝑉𝑑𝑐𝑠𝑠 < 𝑉𝑑𝑐𝑚𝑎𝑥1 (8) to investigate the performance of the proposed supercapacitor
sizing algorithm. The system was first simulated in islanded
𝑉𝑑𝑐𝑚𝑖𝑛2 < 𝑉𝑑𝑐𝑓𝑎𝑢𝑙𝑡 < 𝑉𝑑𝑐𝑚𝑎𝑥2 (9) mode without communication during the extreme case of having
Table I. Lower and Upper Bounds of Optimization Variables
a three phase to ground fault and a pulsed load peak period.
Next, the system was simulated under the same conditions but
Parameter Lower Limit Upper Limit using the optimized controller parameters that limits the super-
𝐶 1.5 2.7 capacitors fault current contribution and the obtained superca-
Frequency 𝐾𝑝 10 80 pacitor size.
Frequency 𝐾𝑖 80 140 Finally, the case where the pulsed load was temporarily dis-
AC Voltage 𝐾𝑝 10 50 connected from the system during the transient fault period was
investigated. The supercapacitor’s fault current contribution
AC Voltage 𝐾𝑖 100 150
along with the overall system stability was recorded and ana-
DC Voltage 𝐾𝑝 10 40 lyzed in what follows.
DC Voltage 𝐾𝑖 80 160

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

Case I: Islanded Mode of Operation without Communication Due to the assumption of the communication failure, relay 𝑅7
and During Supercapacitor Discharging Period will not be able to switch to the lower settings and the superca-
pacitor quickly contributed to the fault current while still cover-
A three-phase to ground fault (𝐹1 in Fig. 1) has been applied
ing the pulsed load.
in the transmission line (𝑇𝐿1 ) at time t = 1.5 seconds, while the
Fig 6(a) shows the microgrid frequency recovery after the
microgrid was in grid connected mode of operation. As a result
fault incident. As it can be noticed, the system is showing stable
relay 𝑅1 will send a trip signal to circuit breaker 𝐶𝐵1 to isolate
performance during and after the fault with disturbances within
the microgrid. The microgrid successfully shifted to a stable is-
the specified limits of microgrids operation. The sources volt-
landed mode by adjusting its overall frequency back to the nor-
ages 𝑉𝑔2 , 𝑉𝑖𝑛𝑣1 and 𝑉𝑖𝑛𝑣2 dropped at the fault incident and recov-
mal condition after the fault, as shown in Fig. 6(a). A small dis-
ered after clearing the fault, as shown in Fig. 6(b). As indicated
turbance, within acceptable limits, in the output voltages of the
in Fig. 6(d), the maximum current drawn from the supercapaci-
sources (𝑉𝐺 , 𝑉𝑔1 , 𝑉𝑔2 , 𝑉𝑖𝑛𝑣1 and 𝑉𝑖𝑛𝑣2 representing the voltages
tor in this case is 43 Amps which is required to feed the fault and
at busses 1, 4, 6, 3 and 5, respectively), is noticed in Fig. 6(b).
the pulsed load simultaneously. As shown in Fig. 6(d), the cur-
Fig. 6(c) and (d) show the supercapacitor-pulsed load microgrid
rent passing through the converter 1 during the normal operation
performance which exhibit stable performance during the is-
was 4 Apms. However, this current increased to 20 Amps during
landing instant. It can be seen from Fig. 6(d) that during the
the fault period. Fig. 6(e) shows that the fault current drastically
charging of the supercapacitor, the converter 1 allowed 4 Amps
increased from 5 Amps to 32 Amps. This resulted in relay 𝑅7
of current for the AC side to support the pulsed load. Finally,
sensing the fault and sending trip signal to 𝐶𝐵7 to isolate the
Fig. 6(e) and (f) indicate the minor change in the AC current at
fault. Fig. 6(f) shows the setting of 𝑅8 that tripped with the help
both sides of transmission line 𝑇𝐿4 after fault F1. During the is-
of 𝐺1 to isolate the fault completely.
landed operation, another three-phase-to-ground fault (F2) oc-
curred at t = 4.5 seconds during the on-time of the pulsed load Case II: System Performance with Optimized Supercapacitor
(i.e. discharging of the supercapacitor) in the middle of trans- Size
mission line 𝑇𝐿4 .
In this case, the optimized frequency, AC, and DC voltages
Frequency (Hz)

62 controllers were utilized along with the obtained supercapacitor

Frequency dip size of 2.5F (compared with the original 2.9F). Fig. 7(a) shows
60
a stable system performance, with the new parameters, in terms
Frequency recovery after fault
58 of its frequency during the first fault (𝐹1 ) to isolate the microgrid
1 2 3 4 5 6 7 8 9 10
(a) as well as during the second fault (𝐹2 ). It can be noted that the
300
Voltage (V)

VG Vg1 Vg2 Vinv1 Vinv2 drop of frequency during the second fault is less than the drop
200
without optimized parameters. Fig. 7(b) shows that the drop in
100
the sources’ voltages (𝑉𝐺 , 𝑉𝑔1 , 𝑉𝑔2 , 𝑉𝑖𝑛𝑣1 and 𝑉𝑖𝑛𝑣2 ) is within
1 2 3 4 5
(b)
6 7 8 9 10 accepted limits during both faults. Fig. 7(d) shows that the
Voltage (V)

320 amount of current that was injected by the supercapacitor is re-

310 duced to 38 Amps compared to 43 Amps in the previous case.
This plays a major role in reducing the size of the supercapac-
300
1 2 3 4 5 (c) 6 7 8 9 10 itor and thus it’s cost as it will be shown in the next section. Fig.
150
7(e) shows that 𝑅7 reached its high setting value as a result of
Current (A)

Supercapacitor Inverter1 Pulsed Load

100 increasing the amount of current from the AC side even during
Supercapacitor
50 feeding fault & pulsed load communication outages. It can be noted that 𝑅8 was tripped to
0 isolate the fault from the other terminal of 𝑇𝐿4 and removed the
1 2 3 4 5 (d) 6 7 8 9 10 fault completely from the system.
Current (A)

40
I=5A Fault Case III: Disconnecting the Pulsed Load during the Fault Pe-
20 I=32A cleared riod
0
1 2 3 4 5 6 7 8 9 10
Similar to the previous case, fault 𝐹2 occurred during the on-
(e)
time period of the pulsed load. It is noticed in Fig. 8(a) that, due
Current (A)

20
10
I=4.6A I=10A After to communication failure, relay 𝑅7 will not switch to the lower
fault settings and the supercapacitor quickly contributed to the fault
0 current. However, in this case, the pulsed load was disconnected
1 2 3 4 5 6 7 8 9 10
(f) for a short period of time during the fault in order to reduce the
Time (s) amount of current drawn from the supercapacitor which reached
Fig. 6. System performance during fault at microgrid operation without commu- a maximum of 27 Amps in this case. To disconnect the pulsed
nication (a) frequency, (b) output voltage of each source, (c) supercapacitor DC
voltage, (d) supercapacitor current, and pulsed load current, and inverter1 cur-
load, the rate of change of the supercapacitor’s voltage was mon-
rent (e) RMS current for R7, (f) RMS current for R8. itored to detect a small notch in the DC voltage indicating the
fault as shown in Fig. 8(b) and (c).

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

62
Frequency (Hz)
40

Current (A)
Fault Instant to isolate microgrid Supercapacitor
60 20 feeding fault
only
0
Fault Instant during islanded Effect of pulse load
58
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
(a) (a)
320
Voltage (V)

200 VG Vg1 Vg2 Vinv1 Vinv2

Voltage (V)
100
310
0 Voltage Variation During Fault
1 2 3 4 5 (b) 6 7 8 9 10 300
1 2 3 4 5 6 7 8 9 10
Voltage (V)

320 (b)
310 40

Current (A)
Fault
300 Period
1 2 3 4 5 6 7 8 9 10 20
(c)
80 After Fault
Current (A)

60 Supercapacitor Pulse Load 0

40 Supercapacitor 1 2 3 4 5 6 7 8 9 10
feeding pulse (c)
20 Time (s)
load& fault
0 Fig. 8. System performance during fault at microgrid operation without commu-
1 2 3 4 5 6 7 8 9 10 nication (a) supercapacitor current, (b) supercapacitor DC voltage, (c) pulsed
(d) load current.
Current (A) Current (A)

40
Fault
20 Instant ing 2.9F is 43 Amps which is required to feed the fault and the
I=32A
0 pulsed load simultaneously. In this case, for complete isolation
1 2 3 4 5 6 7 8 9 10
(e) of the fault, G1 helped 𝑅8 to reach its setting which is 10 Amps.
20 Time (s)
I=4.6A The length of the transmission lines are shown in Table III and
10 I=10A After
fault selected such that the sources in the microgrid are enough to help
0 the relays to isolate the fault taking into consideration the current
1 2 3 4 5 6 7 8 9 10
(f) losses due to the voltage drop. However, under the same fault
Time (s) condition, the optimized supercapacitor value (2.5F) and opti-
Fig. 7. System performance during fault at microgrid operation without com- mized controller parameters, the supercapacitor fault current
munication with optimized supercapacitor size (a) frequency, (b) output volt- contribution was 38 Amps. This current is less than the initial
age of each source, (c) supercapacitor DC voltage, (d) supercapacitor current, supercapacitor case (43 Amps) by 5 Amps, while the contribu-
and pulsed load current, (e) RMS current for R7. (f) RMS current for R8.
tion from AC microgrid increased to 19 Amps instead of 14
Table II shows the operation of the system under several types Amps in order to reach to the settings of relay R7 and eventually
of faults, applied on different locations (𝑇𝐿4 and 𝑇𝐿5 ), for the clear the fault. Thus, the optimized supercapacitor and control-
initial supercapacitor value of 2.9F and the optimized value ler’s values can compensate the amount of current that is re-
2.5F. For a three phase to ground fault (𝐹2 ) applied on 𝑇𝐿4 with quired to reach to the setting of the relay during discharging the
fault resistance 2Ω, it can be noted that at grid connected mode, supercapacitor and feed the pulsed load. Similarly, in the case of
the fault current reached 32 Amps that is more than 6 times the a double-line-to-ground fault (BCG), the initial supercapacitor
rated current which is 5 Amps. For islanded mode of operation, and the optimized supercapacitor and controller’s values can be
when the communication is available and the relays can adjust sufficient to compensate the amount of current that is needed in
its settings from grid mode (6-7 times rated current) to the is- islanded mode without communication mode. The AC mi-
landed mode (1.5-2.5 times rated current), the relay 𝑅7 can send crogrid generating units were capable of feeing sufficient fault
a trip signal to the circuit breaker and isolate the fault. During current for fault detection in the case of a single-line-to-ground
islanded mode without communication case, the maximum cur- fault without the need of the supercapacitor. In this case, the su-
rent drawn from the supercapacitor for the initial case with rat- percapacitor can be used to feed only pulsed load.

Table II. Several types of faults at different locations with the rating of Supercapacitor with and without optimization
Relay Islanded Islanded mode without Communication During Islanded mode without Communication During
Setting mode Discharging Supercapacitor at Case I Discharging Supercapacitor at Case II
Fault During with RS RS
Rf
Type Grid Com-
(Ω) Ic Im Ic Im Ipl
Mode munica- Ipl TL4 TL5 TL4 TL5
(A) tion (A)
R7 R9 R7 R9 R7 R9 R7 R9 R7 R8 R9 R10 R7 R9 R7 R9 R7 R8 R9 R10
ABCG 2 32 31 12 11 43 45 14 15 23 32 10 31 11 38 41 19 20 23 32 10 31 11
BCG 1.8 30 25 10 9 41 42 15 13 23 30 9 25 10 33 37 23 18 23 30 9 25 10
AG 0.1 18 16 8 7 23 23 19 17 23 18 7 16 8 23 23 19 17 23 18 7 16 8
𝑅𝑓 = Fault Resistance; RS = Relay Settings; 𝐼𝑐 = Supercapacitor current contribution; 𝐼𝑚 = Microgrid current contribution; 𝐼𝑝𝑙 = Current drawn from superca-
pacitor to feed the pulsed load

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

Also Table II shows the same fault conditions (𝐹3 ) that oc- Table IV. Cost estimation of the supercapacitor bank
curred on 𝑇𝐿5 to investigate the ability of the supercapacitor of ∆𝑽 E1 E2 E1/E2 Price (k$)
TL TL TL TL
feeding the fault current for faults which occur distant from the C 4 5 4 5 --- 4 5 4 5
location of the supercapacitor. During the grid connected and 2.9 12 12.4 208.8 237.5 2.1 99.42 113.1 12.1 13.8
islanded mode with communication, the fault current setting of 2.5 12 12.6 180 198.4 2.1 85.71 94.5 10.4 11.5
relays 𝑅9 and 𝑅10 were met without interference from the super- 1.8 11 11.1 108.9 110.8 2.1 51.85 52.8 6.3 6.4
capacitor. For the islanded mode without communication, the
initial supercapacitor and the optimized one were able to feed Henceforth, the proposed sizing algorithm along with opti-
the fault current and the pulsed load simultaneously for the three mizing the corresponding controller parameters is suitable fea-
phase to ground and the double line to ground faults. The mi- sible and produced excellent results. It is shown that the size and
crogrid resources were enough to feed the single line to ground cost of the supercapacitor were minimized, while the system is
faults. Finally, the maximum contribution from the supercapac- still meeting the protection and safe operation requirements.
itor to feed the pulsed load during a three-phase-to-ground fault This is very beneficial when the system has pulsed loads as these
is 45A in 𝑇𝐿5 . It can be noted that the pulsed load at the DC side types of loads do not tolerate any interruption during their turned
is supported by little contribution from AC side equivalent to 4 on period. It is important to mention here, that the other pro-
Amps, and from supercapacitor by 23A. The output of superca- posed solution of temporarily disconnecting the pulsed load is
pacitor which is injected to the AC side is 22A. Therefore, 𝐶𝐵11 suitable for noncritical pulsed loads.
is adjusted at 24A.
VII. EXPERIMENTAL VERIFICATION
VI. IMPACT OF REDUCED SIZING OF THE COST OF THE
SUPERCAPACITOR BANK A. Hardware Setup Description
During energy utilization of a supercapacitor bank, its termi- The hybrid AC/DC microgrid, used for experimental verifi-
cation, consists of several types of sources that supply loads with
nal voltage will vary with time. So, the total energy (𝐸1 ) that can
different characteristics, including: AC load, a normal steady-
be delivered by the supercapacitor can be expressed by:
state load, and a heavy pulsed load at DC side. Fig. 9 shows the
𝐸1 = 1⁄2 𝐶 (𝑉𝑚𝑎𝑥 − 𝑉𝑚𝑖𝑛 )2 (10) experimental test setups of the microgrid. The hardware setup
Where 𝑉𝑚𝑎𝑥 is the maximum voltage of a supercapacitor bank and the control structure for this system and the interconnected
during a pulsed load and 𝑉𝑚𝑖𝑛 is its minimum voltage. In the AC grid are explained in the following sections.
proposed hybrid DC microgrid, the maximum acceptable varia- The two AC generators are mainly supplying different loads
tion of the DC bus voltage is limited to ±5%. which composed of ten levels of parallel resistive loads from 0
In order to evaluate the economic benefits of the proposed op- to 3 kW in steps of 300 W at nominal voltage. [2 ×72 Ω + 4 ×
timization method, a financial estimation of the supercapacitor 144 Ω in each phase] resistance models, which can be controlled
cost is performed. The commercial price of the supercapacitor by PLC, to emulate various load patterns.
Model BMOD0058-E016-B02 manufactured by Maxwell [36], This system also includes supervisory control and data acqui-
is 122.25$. Table IV shows the optimization results of the cost sition system to monitor the entire system. More details about
of the supercapacitor bank, where (𝐸2 ) is the stored energy per the AC grid testbed configuration and control can be found in
module, (𝐸1 /𝐸2 ) represents the approximate number of required [37], [38]. In addition, a supercapacitor bank was implemented
modules. For the case of the fault on 𝑇𝐿4 , it can be noted from as an energy buffer in our hybrid AC/DC power system labora-
the table that the sizing and cost of supercapacitor bank is re- tory test-bed.
duced from 2.9F to 2.5F and the total cost is reduced from An analog hysteresis protection is designed to ensure that un-
12.155k$ to 10.47k$ as a result of the optimization process. even charge distribution, particularly during very fast charges,
Also, in the case where pulsed load was temporarily discon- does not cause the supercapacitor to exceed its maximum volt-
nected during the fault, the supercapacitor’s cost was reduced age limit. This protection unit also provides a backup for the en-
from 12.155 k$ to 6.339 k$. However, a slightly larger superca- ergy control system. If the voltage on any of the supercapacitor
pacitor bank was needed for the case of a fault on 𝑇𝐿5 . The ini- arrays exceeds the preset limit, the analog control circuit will
tial cost was 13.8k$ and was reduced after the optimization pro- open the output of the insulated-gate bipolar transistor (IGBT)
cess to 11.5 k$. For the case of the temporary disconnection of switch (sw), as shown in Fig. 9. The charging path remains open
the supercapacitor, the price increase was negligible. until the supercapacitor is discharged through the bypass diode
to the point that its voltage is reduced to the predefined safe
TABLE III. Lines Parameters threshold.
TL Length R (Ω/km) L (mH/km) The detailed parameters of the system are summarized in Ta-
Line 1 10 km 2.24 0.430
ble V. For dynamic operation and fault study, the detailed model
of the microgrid shown in Fig. 9 is implemented in our virtual
Line 2 8 km 1.41 0.357
test-bed platform using the MATLAB/Simulink software and
Line 3 8 km 1.41 0.357 executed with the dSPACE 1104 real-time interface to control
Line 4 6 km 2.24 0.430 and adjust the pulsed load and the steady-state load. In order to
Line 5 6 km 2.24 0.430 obtain the optimal values of the PI controllers, the Simulink Re-
sponse Optimization toolbox was employed [39].

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

G1 R/CB1 TL1 R/CB2 Inv

SW SC
Pulsed Load
Generator Station

R/CB5 Real Time

PL CL Steady State Load Interface
L1 L2

TL3
R : Relay
dSPACE DS1104
SW : Switch AC Filter
CB : Circuit Breaker DC Measurement
R/CB6
F : Fault
TL2 G : Generator
PL : Pulse Load
CL : Constant Load
G2 SC : Supercapacitor
R/CB3 F R/CB4
TL : Transmission Line Supercapacitor
Inverter Bank
L : AC Load
L3 Inv : Inverter

Fig. 9. General view of the experimental test setup

B. Experimental Test Results

Table V. System Component Parameters
The performance of the microgrid at islanded mode of opera-
Component Parameter Specification
tion was evaluated in three test cases: First, utilizing the original
supercapacitor size and controller paramenters, then, utilizing Apparent Power 13.8, 10 KVA
the values obtained from the two-level optimization process. Fi- Nominal Voltage 208, 208 V
nally, the case in which the pulsed load is disconnected at fault G1, G2
Stator Leakage Reactance 0.09 0.09 pu
instant was tested for the proposed protection technique. d-axis Synchronous Reactance 2.21, 2.248 pu
q-axis Synchronous Reactance 1.1, 1.1 pu
Fig. 10 shows the performance of the microgrid using the d-axis Transient Time Constant 0.014, 0.012 s
original supercapacitor value and controllers parameters (C =
2.9 F, 𝑘𝑝_𝑓 =50, 𝑘𝑖_𝑓 =100, 𝑘𝑝_𝑣𝑎𝑐 =40, 𝑘𝑖_𝑣𝑎𝑐 =110, 𝑘𝑝_𝑣𝑑𝑐 =30, and Number of cells 20
𝑘𝑖_𝑣𝑑𝑐 = 100). During the supercapacitor’s discharging period be- Supercapacitor
Rated voltage 320
Rated capacity 2.9 F
tween t = 0.94 sec and t = 2.97 sec, the frequency of the system bank
Surge voltage 340
remained at 59.81 Hz and the DC voltage dropped a little from Leakage current 5.2 mA
320 V to 311 V, as shown in Fig. 10 (a) and (b) respectively.
Also, during the supercapacitor charging period between t = 𝑅𝑠 1e5 Ω
2.98 sec and t = 5.05 sec, the system frequency was stable at 60 Inverter 𝑅𝑜𝑛 1e-3 Ω
Switching frequency 5 KHz
Hz and the DV voltage at 320 V. A three phase to ground fault
occurred at t = 6 seconds. This was accompanied by a drop to 𝑟1, 0.0015 Ω/Km,
59.32 Hz. Fig. 10 (c) shows that during the fault period, the su- 𝑟0 0.03 Ω/Km
percapacitor’s current increased from 5.21 A to 9.86 A, which Transmission
𝑙1, 0.03 mH/Km,
indicates that the fault current contribution from the supercapac- Line 𝑙0 0.1 mH/Km
𝑐1, 3 nF/Km,
itor was 4.65 A. Fig. 10 (d) and (e) show that the pulsed load 2nF/Km
𝑐0
and steady state currents remained stable throughout the experi-
ment. Fig 10 (f) indicates that the total fault current of 𝑅3 is 7.2
Filter 𝐿𝐴𝐹 12 mH
A, thus reaching the high relay settings, which is assumed to be
3 times the normal current (2.4 A). Note that current values in Fault Re-
the hardware experimentation were scaled down, compared to 𝑅𝑓 18 Ω
sistance
those in the simulation, to operate in safe experimental limits in 300 Ω, 900 Ω,
a laboratory environment and the relay settings were adjusted Loads 1,2,3 𝐿1, 𝐿2 , 𝐿3
1200 Ω
accordingly. It is important to note that the system maintained
it’s a stable operation in the experiment. That is, the frequency Constant Load 𝐶𝐿 550 W
remained between the acceptable ranges of 59.7 Hz and 60.2 Hz
Pulsed Load 𝑃𝐿 1.66 kW
and the AC and DC voltages have a maximum acceptable vari-
ation of ±5% of their rated values.

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

Frequency (Hz) 61 61

Frequency (Hz)
60 (a) 60 (a)

59 59
2 4 6 8 10 12 14 16 18 20 22 24 26
330
Voltage (V)

320 10

Current (A)
(b)
310 5 (b)

300
2 4 6 8 10 12 0
14 16 18 20 22 24 26
10
Current (A)

5 (c) 6

Current (A)
4
0 (c)
2
2 4 6 8 10 12
0
6
Current (A)

14 16 18 20 22 24 26
4
(d)
2 3

Current (A)
0 2
2 4 6 8 10 12 1
3
(d)
0
Current (A)

2
-1
1 (e) 14 16 18 20 22 24 26
0
10
Current (A)

-1
2 4 6 8 10 12
10 5 (e)
Current (A)

5 (f) 0
14 16 18 20 22 24 26

Time (s)
0
2 4 6 8 10 12 Fig. 11. Microgrid performance during fault at islanded mode of operation with
Time (s) optimized supercapacitor size (a) frequency, (b) supercapacitor current, (c)
pulsed load current, (d) steady state load current, (e) RMS current in the faulted
Fig. 10. Experimental test results during fault at islanded mode of operation transmission line.
without optimized supercapacitor size (a) frequency, (b) supercapacitor DC volt-
age, (c) supercapacitor current, (d) pulsed load current, (e) steady state load cur- Fig. 11 (c) and (d) show that the pulsed load and steady state
rent, (f) RMS current in the faulted transmission line. currents remained stable during the experiment. These results
Fig. 11 shows the performance of the microgrid using the op- show that the optimization enhanced the operation of the system
timized supercapacitor value and controllers parameters (C = 2.5 during normal and fault operations and decreased the size of the
F, 𝑘𝑝_𝑓 =65, 𝑘𝑖_𝑓 =110 , 𝑘𝑝_𝑣𝑎𝑐 =30, 𝑘𝑖_𝑣𝑎𝑐 =115, 𝑘𝑝_𝑣𝑑𝑐 =40, and supercapacitor (thus reducing its cost) while maintaining
enough fault current contribution to trigger the relays in the
𝑘𝑖_𝑣𝑑𝑐 =110 ). Similar to the previous case, the system operating
event of communication loss as explained in the previous sec-
under the optimized values was capable to maintain its fre-
tion. Finally, Fig. 12 presents the performance of the system
quency within the acceptable limits. In fact, during the fault in-
when uncritical pulsed loads are removed during the fault pe-
stant at t= 18.8 sec, the frequency dropped to a minimum of
riod. As mentioned earlier, in this case, a notch in the DC voltage
59.61 Hz as shown in Fig 10 (a), which is better than the drop to
profile indicates a fault and is therefore used as the trigger to
59.32 Hz of the previous case. Fig 10 (b) shows that during the
disconnect the pulsed load. Fig 12 (b) shows a notch in the DC
fault, the supecapictor’s current increased from 5.21 A to 7.8 A,
voltage at t = 10.2 sec. At this instant, the pulsed load is discon-
which indicates that the fault current contribution of the super-
nected. This is reflected in the pulsed load current in Fig. 12 (c),
capcitor in this case is 2.59 A.
which drops from 5.2 A to 0 A during the fault period. Fig. 12
This corresponds to a 44.3% reduction in the supercapacitor’s
(a) shows a decrease in the supercapacitor’s current during that
fault current contribution compared to the previous case. How-
period from 5.21 A to 1.03 A, corresponding to a 4.18 A drop.
ever, Fig. 11 (e) shows that the fault current still reached 7.2 A
It is noticeable from Fig. 10 (b) that the variation of the DC volt-
and the relays were activated. This is because the remaining fault
age was changed to be between 320 V and 306 V. This is directly
current was supplied by the AC generators as intended from the
reflected in significant cost reduction of the supercapacitor bank
optimization process.
as indicated in equation 10.

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

10 [5] M. A. Haj-ahmed and M. S. Illindala, "The Influence of Inverter-Based

DGs and Their Controllers on Distribution Network Protection," in IEEE
Current (A)
5 Transactions on Industry Applications., vol. 50, no. 4, pp. 2928-2937,
(a) July-Aug. 2014.
0 [6] H. Moradi Sizkoohi, J. Milimonfared, M. Taheri and S. Salehi, "High step-
up soft-switched dual-boost coupled-inductor-based converter integrating
-5
4 6 8 10 12 14 16 multipurpose coupled inductors with capacitor-diode stages," in IET
330 Power Electronics, vol. 8, no. 9, pp. 1786-1797, 9 2015.
[7] Hany F. Habib; T. Yossef; O. Mohammed, "A Multi-Agent Based Tech-
Voltage (V)

320 nique for Fault Location, Isolation, and Service Restoration," in IEEE
Transactions on Industry Applications , vol. no.99, pp.1-1, Feb. 2017.
(b) [8] H Moradisizkoohi, J Milimonfared, M Taheri, S Salehi, “Duty-cycle-con-
310
trolled resonant dual-half-bridge converter with multifunctional capacitors
300 for distributed generation applications”, IET Power Electronics 9 (9),
4 6 8 10 12 14 16 1873-1884.
[9] Ahmed. F. Ebrahim, T. Youssef, S. M. W. Ahmed, S. E. Elmasry and O.
10
A. Mohammed, "Fault detection and compensation for a PV system grid
Current (A)

tie inverter," North American Power Symposium (NAPS), 2014, Pullman,

5
(c) WA, 2014, pp. 1-6.
0 [10] M. Mahmoudian Esfahani, A. Sheikh and O. Mohammed, “Adaptive real-
time congestion management in smart power systems using a real-time hy-
-5 brid optimization algorithm” Electric Power Systems Research, vol. 150,
4 6 8 10 12 14 16 pp. 118-128, 2017.
Time (s) [11] H. F. Habib, C. R. Lashway and O. A. Mohammed, "On the adaptive pro-
tection of microgrids: A review on how to mitigate cyber attacks and com-
Fig. 12. Operation test result of the microgrid during fault at islanded mode (a) munication failures," 2017 IEEE Industry Applications Society Annual
supercapacitor current, (b) supercapacitor DC voltage, (c) pulsed load current. Meeting, Cincinnati, OH, 2017, pp. 1-8.
[12] M. Mahmoudian, G.R. Yousefi, “ATC Improvement and Losses Estima-
VIII. CONCLUSION tion Considering Dynamic Transmission Line Ratings”, ICEE, 20th Ira-
nian Conference, Tehran, Iran, pp. 404-409, 2012.
This paper proposes two solutions for reducing the size and [13] Laaksonen, H.J., "Protection Principles for Future Microgrids”, in IEEE
cost of supercapacitors that are used to compensate for the low Trans. Power Electronics, vol.25, no.12, pp.2910,2918, Dec. 2010.
islanded microgrid fault currents during communication outages [14] Najy, W.K.A; Zeineldin, H.H., "Optimal Protection Coordination for Mi-
in adaptive protection schemes. First, for critical pulsed loads, a crogrids With Grid Connected and Islanded Capability,", in IEEE Trans,
Industrial Electronics , vol.60, no.4, pp.1668,1677, April 2013.
two-level optimization scheme is presented. In the first level, the [15] Laaksonen, H.; Ishchenko, D.; Oudalov, A, "Adaptive Protection and Mi-
goal is to reduce the proposed protection scheme’s cost by se- crogrid Control Design for Hailuoto Island", in IEEE Trans. On Smart
lecting the minimal value of the supercapacitor that maintains Grid, vol.5, no.3, pp.1486,1493, May 2014.
the system’s stable operation. In level 2, the main of objective is [16] R.M. Tumilty, M. Brucolit, G.M. Burt and T.C. Greent , “ Approaches to
Network Protection for Inverter Dominated Electrical Distribution Sys-
to minimize the supercapacitor’s fault current contribution and tems’, in the 3rd IET, Machines and Drives, pp. 622-626, 2006.
maximize that of the distributed generators. The optimized size [17] M. Saleh, Y. Esa and A. Mohamed, "Communication Based Control for
of the supercapacitor and the associated controllers parameters DC Microgrids," in IEEE Transactions on Smart Grid, vol. PP, no. 99, pp.
were investigated under several types of faults in different loca- 1-1.
[18] M. Mahmoudian Esfahani, Abla Hariri and O. Mohammed, “A Multia-
tion in simulation and were verified experimentally. The results gent-based Game-Theoretic and Optimization Approach for Market Oper-
show the ability of the optimized supercapacitor rating to feed ation of Multi-Microgrid Systems” IEEE Transactions on Industrial Infor-
different types of faults in order to raise the fault current to the matics, DOI: 10.1109/TII.2018.2808183.
high relay settings and maintain stable system operation during [19] M. E. Hariri, E. Harmon, H. F. Habib, T. Youssef and O. A. Mohammed,
"A targeted attack for enhancing resiliency of intelligent intrusion detec-
communication outages. Also, the supercapacitor proved effec- tion modules in energy cyber physical systems," 2017 19th International
tive for faults occurring in distant locations in the microgrid. Fi- Conference on Intelligent System Application to Power Systems (ISAP),
nally, temporary disconnection of pulsed loads during the tran- San Antonio, TX, 2017, pp. 1-6.
sient fault period proved effective for non-critical pulsed loads [20] R. Akella, H. Tang, and M.M. Bruce, “Analysis of Information Flow Se-
curity in Cyber-Physical System”, in International Journal of Critical In-
as a solution to avoid utilization of oversized supercapacitors. frastructure Protection, Vol. 3, pp: 157-173, 2010.
[21] M. E. Hariri, T. A. Youssef, H. F. Habib and O. Mohammed, "Online false
REFERENCE data detection and lost packet forecasting system using time series neural
[1] J. Tang, Y. Gong, N. Schulz, M. Steurer and P. G. McLaren, "Implemen- networks for IEC 61850 sampled measured values," 2017 IEEE Power &
tation of a Ship-Wide Area Differential Protection Scheme," in IEEE Energy Society Innovative Smart Grid Technologies Conference (ISGT),
Transactions on Industry Applications, vol. 44, no. 6, pp. 1864-1871, Washington, DC, 2017, pp. 1-5.
Nov.-Dec. 2008. [22] H. F. Habib, M. E. Hariri, A. Elsayed and O. Mohammed, "Utilization of
[2] M. Mahmoudian Esfahani, M.H. Cintuglu, O.A. Mohammed, “Optimal supercapacitors in adaptive protection applications for resiliency against
real-time congestion management in power markets based on particle communication failures: A size and cost optimization case study," in IEEE
swarm optimization”, IEEE Power and Energy Society General Meeting Industry Applications Society Annual Meeting, Cincinnati, OH, 2017, pp.
2017, DOI: 10.1109/PESGM.2017.8274117. 1-8.
[3] M. Shojaie and H. Mokhtari, "A method for determination of harmonics [23] Hany. F. Habib, A. A. S. Mohamed, Mohamad. El Hariri and Osama. A.
responsibilities at the point of common coupling using data correlation Mohammed, “Utilizing Supercapacitors for Resiliency Enhancements and
analysis," in IET Generation, Transmission & Distribution, vol. 8, no. 1, Adaptive Microgrid Protection against Communication Failures”, in Elec-
pp. 142-150, Jan. 2014. tric power systems Research, vol.145, pp.223-233, Apr., 2017.
[4] H Moradisizkoohi, J Milimonfared, M Taheri, S Salehi, “Duty-cycle-con- [24] B. Sood, M. Osterman, and M. Pecht, “Health monitoring of lithium-ion
trolled resonant dual-half-bridge converter with multifunctional capacitors batteries,” in IEEE Symposium on Product Compliance Engineering
for distributed generation applications”, in IET Power Electronics 9 (9), (ISPCE), pp. 1–6, 2013.
1873-1884.

0093-9994 (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2819620, IEEE
Transactions on Industry Applications

[25] H. F. Habib, C. R. Lashway and O. A. Mohammed, "A Review of Hany Fawzy Habib received the B.S. and M.S.
Communication Failure Impacts on Adaptive Microgrid Protection degrees in electrical engineering from Helwan
Schemes and the Use of Energy Storage as a Contingency," in IEEE University, Cairo, Egypt, in 2007 and 2014, re-
Transactions on Industry Applications, vol. PP, no. 99, pp. 1-1. spectively. He is currently pursuing the Ph.D. de-
[26] P. Bajpai and V. Dash, “Hybrid renewable energy systems for power gen- gree from the Energy Systems Research Labora-
eration in stand-alone applications: A review,” in Renewable and Sustain- tory, Electrical and Computer Engineering Depart-
able Energy Reviews, vol. 16, no. 5, pp. 2926–2939, Jun. 2012. ment, College of Engineering and Computing,
[27] “Supercapacitor Information – Battery University.” [Online]. Available: Florida International University, Miami, FL, USA.
http://batteryuniversity.com/learn/article/whats_the_role_of_the_superca- His current research interests include, power sys-
pacitor. [Accessed: 10-Feb-2017]. tem protection, multiagent systems, cyber physical
[28] Z. Jin, G. Sulligoi, R. Cuzner, L. Meng, J. C. Vasquez and J. M. Guerrero, systems for active distribution networks, energy
"Next-Generation Shipboard DC Power System: Introduction Smart Grid storage devices, and microgrids
and dc Microgrid Technologies into Maritime Electrical Netowrks," in
IEEE Electrification Magazine, vol. 4, no. 2, pp. 45-57, June 2016.
[29] Ahmadi, A., Masouleh, M. S., Janghorbani, M., Manjili, N. Y. G., Sharaf, Mohamad El Hariri is currently a Ph.D. candi-
A. M., & Nezhad, A. E. (2015). Short term multi-objective hydrothermal date and Research Assistant at the Energy Systems
scheduling. Electric Power Systems Research, 121, 357-367. Research Laboratory at Florida International Uni-
[30] Samy Faddel, A. Elsayed and Osama A. Mohammed, "Bi-Layer Multi- versity. He received his B.Ss. and M.Sc. degree in
Objective Optimal Allocation and Sizing of Electric Vehicle Parking Gar- Mechatronics Engineering with distinction from
age," in IEEE Transactions on Industry Applications, vol. PP, no. 99, pp. Rafik Hariri University, Mechref, Lebanon, in
1-1, Early Access. 2012 and 2014 respectively. Between 2014 and
[31] H. F. Habib, A. O. Hariri, A. ElSayed and O. A. Mohammed, "Deployment 2015, Mohamad was a technical projects engineer
of electric vehicles in an adaptive protection technique for riding through at National Oilwell Varco, Dubai. Mohamad’s re-
cyber attack threats in microgrids," in IEEE International Conference on search includes cyber security for cyber-physical
Environment and Electrical Engineering and 2017 IEEE Industrial and systems, computational electromagnetics, and ar-
Commercial Power Systems Europe (EEEIC / I&CPS Europe), Milan, pp. tificial intelligence techniques in power system ap-
1-6, 2017. plications.
[32] H. H. Eldeeb, A. O. Hariri and O. A. Mohammed, "Coordinated control
for the integration of a large scale electric vehicle park with PV into the
MV grid," 2017 IEEE International Conference on Environment and Elec- Osama A. Mohammed is a Professor of Electrical
trical Engineering and 2017 IEEE Industrial and Commercial Power Sys- Engineering and is the Director of the Energy Sys-
tems Europe (EEEIC / I&CPS Europe), Milan, 2017, pp. 1-6. tems Research Laboratory at Florida International
[33] H Moradisizkoohi, OA Mohammed, “A quasi-resonant bi-directional University, Miami, Florida. He received his Mas-
buck-boost converter for Electric Vehicle applications”, in Transportation ter and Doctoral degrees in Electrical Engineering
Electrification Conference and Expo (ITEC), 2017 IEEE, 621-625 from Virginia Tech in 1981 and 1983, respec-
[34] R.A.Dougal, Shengyi Liu, R.E. White, "Power and life extension of bat- tively. He has performed research on various top-
tery-ultracapacitor hybrids,", in IEEE Transactions on Components and ics in power and energy systems in addition to de-
Packaging Technologies, vol.25, no.1, pp.120,131, Mar 2002. sign optimization and physics passed modeling in
[35] H. H. Eldeeb, A. O. Hariri, C. R. Lashway and O. A. Mohammed, "Opti- electric drive systems and other low frequency en-
mal sizing of inverters and energy storage for power oscillation limiting in vironments. Professor Mohammed is a world re-
grid connected large scale Electric Vehicle park with renewable energy," nowned leader in electrical energy systems. He has
2017 IEEE Transportation Electrification Conference and Expo (ITEC), performed research in the area of electromagnetic signature, wideband gap de-
Chicago, IL, 2017, pp. 288-293. vices and switching, and ship power systems modeling and analysis.
[36] Maxwell, 16V small cell module, Model: BMOD0058 E016 B02 (16.2V- He has current active research projects for several Federal agencies dealing with;
58 F), San Diego, CA. Available Online at: http://www.maxwell.com/im- power system analysis and operation, smart grid distributed control and interop-
ages/documents/datasheet_16v_small_cell_module.pdf. erability, cyber physical systems, and co-design of cyber and physical compo-
[37] H. F. Habib and O. Mohammed, "Decentralized Multi-Agent System for nents for future energy systems applications.
Protection and the Power Restoration Process in Microgrids," 2017 Ninth Professor Mohammed has published more than 450 articles in refereed journals
Annual IEEE Green Technologies Conference (GreenTech), Denver, CO, and other IEEE refereed International conference records. He also authored a
2017, pp. 358-364. book and several book chapters. Professor Mohammed is an elected Fellow of
[38] V. Salehi, A. Mohamed, A. Mazloomzadeh, and O. A. Mohammed, “La- IEEE and is an elected Fellow of the Applied Computational Electromagnetic
boratory-based smart power system, Part I: Design and system develop- Society. Professor Mohammed is the recipient of the prestigious IEEE Power
ment,” in IEEE Trans. Smart Grid, vol. 3, no. 3, pp. 1394–1404, Sep. 2012. and Energy Society Cyril Veinott electromechanical energy conversion award
[39] Simulink Response Optimization User’s Guide. Natick, MA, USA: The and the 2012 outstanding research award from Florida International University.
MathWorks Inc., 2004.

0093-9994 (c) 2018 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.