Energy Conversion and Management 91 (2015) 405–415

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

On battery-less autonomous polygeneration microgrids: Investigation

of the combined hybrid capacitors/hydrogen alternative
George Kyriakarakos a,⇑, Dimitrios D. Piromalis b, Konstantinos G. Arvanitis a, Anastasios I. Dounis b,
George Papadakis a
a
Agricultural University of Athens, Department of Natural Resources and Agricultural Engineering, 75 IeraOdos Street, GR 11855 Athens, Greece
b
Technological Education Institute of Piraeus, Department of Automation, 250, P. Ralli&Thivon Str., Egaleo 122 44, Greece

a r t i c l e i n f o a b s t r a c t

Article history: The autonomous polygeneration microgrid topology aims to cover holistically the needs in remote areas
Received 1 October 2014 as far as electrical power, potable water through desalination, fuel for transportation in the form of
Accepted 10 December 2014 hydrogen, heating and cooling are concerned. Deep discharge lead acid batteries are mostly used in such
systems, associated with specific disadvantages, both technical and environmental. This paper investi-
gated the possibility of replacing the battery bank from a polygeneration microgrid with a hybrid capac-
Keywords: itor bank and more intensive utilization of a hydrogen subsystem. Initially commercial hybrid capacitors
Hybrid capacitors
were tested under laboratory conditions and based on the respective results a case study was performed.
Polygeneration
Microgrid
The optimized combination of hybrid capacitors and higher hydrogen usage was then investigated
Hydrogen through simulations and compared to a polygeneration microgrid featuring deep discharge lead acid bat-
teries. From the results it was clear that it is technically possible to exchange the battery bank with a
hybrid capacitor bank and higher hydrogen utilization. From the economic point of view, the current cost
of the hybrid capacitors and the hydrogen components is high which leads to higher overall cost in com-
parison with deep discharge lead acid batteries. Taking into account, though, the decreasing cost pros-
pects and trends of both the hybrid capacitors and the hydrogen components it is expected that this
approach will become economically competitive in a few years.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction For an autonomous microgrid to operate, energy storage is

The autonomous polygeneration microgrid (APM) topology
aims to holistically cover the needs of communities in remote
areas [1]. The products of such polygeneration microgrids can
include power, fuel for transportation, pumped and desalinated
water, space heating and cooling, and refrigeration. This topology
has been investigated and proved to be both technically feasible
and economically viable [1]. Various approaches have been used
for their energy management [2,3]. Finally an intelligent demand
side energy management system has been created [4]. This topol-
ogy has been also deployed in a real world microgrid in the Sahara
desert in Egypt [5].
Abbreviations: APM, autonomous polygeneration microgrid; CF, cost function;
FCM, Fuzzy Cognitive Map; FM, Flow Matrix or Incidence Matrix of the Petri Net;
NPC, Net Present Cost; PSO, Particle Swarm Optimization; RO, reverse osmosis.
⇑ Corresponding author. Tel.: +30 210 5294046; fax: +30 210 5294032.
E-mail address: gk@aua.gr (G. Kyriakarakos).
essential. The most commonly used energy storage approach is
battery banks, most often in the type of deep discharge lead acid
batteries [6–9]. This type of batteries can reach efficiencies of
about 85% in market available solutions, are simple to operate
and are readily available [10]. On the other hand though, lead acid
batteries present specific disadvantages. Practice has shown that
they have low operational lifetimes [10], they cannot be fully dis-
charged and are also delicate to handle after their operational life-
time has ended because they contain toxic elements [11]. Because
of these specific disadvantages, efforts have been made in order to
minimize lead acid battery usage in renewable energy systems in
favor of hybridized storage approaches [12].
In an effort to minimize battery usage there have been many
examples were ultracapacitors (double layer electrochemical
capacitors) are introduced in a hybridized energy storage system
both for mobile and stationary applications [12–15]. Ultracapaci-
tors have specific advantages like round-trip efficiencies of about
or above 95% and are optimally used in applications where higher
power density is needed. On the other hand, though, they present

http://dx.doi.org/10.1016/j.enconman.2014.12.027
0196-8904/Ó 2014 Elsevier Ltd. All rights reserved.
406 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415
Nomenclature
Cn n concept of the Fuzzy Cognitive Map (–)
eta efficiency of the hybrid capacitor (%)
HTANK volume of stored hydrogen in the hydrogen storage tank
(N m3 H2)
Ic capacitor current (A)
PES energy storage penalty (€)
PH2 hydrogen penalty (€)
PTANKS tanks penalty (€)
Pw water penalty (€)
SOC State of Charge (%)
low energy densities [16]. Another type of capacitors is the hybrid
capacitors, which are manufactured with one electrode being a
double layer material and the other electrode being of a pseudo-
capacitance material [16], which in many cases is a lithium ion
electrode [17,18]. Their performance lies between an ultracapaci-
tor and a battery. These devices present higher energy densities
in comparison with ultracapacitors and could look attractive in
completely replacing batteries in specific applications like the
autonomous polygeneration microgrid, since ultracapacitors can
supply power bursts with ease, but cannot store much energy
[19]. In a polygeneration microgrid the maximum total power
can be observed for constant periods of time up to hours – there
are no loads presenting power peaks for some seconds, if any load
is activated, the needed power is going to be constant until they are
deactivated. In literature hybrid capacitors are also named hybrid
supercapacitors and hybrid ultracapacitors.
In the autonomous polygeneration microgrid topology a hydro-
gen subsystem is used both for the production of on-site fuel, and
for medium to long term energy storage. Also long-term energy
storage can take place in the form of desalinated water [1]. Various
devices used in this topology are not ‘‘instant on’’. For example a
reverse osmosis (RO) desalination unit starts producing water,
after about a minute or in some cases even more until the needed
pressure is reached in the system and acceptable salinity water is
produced. Also it is not advisable to turn on and off an RO desali-
nation unit in a matter of minutes. In the same manner hydrogen
components like electrolyzers and fuel cells need a time period
of some seconds up to minutes until they have completed their
self-checks and are operating. This means than an energy buffer
with adequate capacity to cover the various needs for at least some
minutes is needed. Hybrid capacitors have the potential to fit into
this scheme by substituting the battery bank and being able to
cover the needed energy buffer for the time period needed until
a consumer like the desalination unit or the electrolyzer is acti-
vated when it becomes full and the fuel cell can be activated in
the case that the buffer is getting empty. Apart from this, the
hybrid capacitors have much extended expected lifetime. Cur-
rently commercially available hybrid capacitors are advertised
with lifetime from 10,000 charging/discharging cycles up to
50,000 cycles depending on the company. In comparison deep dis-
charge batteries can only present usually about 1800 charging/dis-
charging cycles at 50% depth of discharge before the capacity drops
below 80% and the battery needs to be replaced [20]. It has to be
noted that the lifetime of deep discharge batteries can considerably
decrease when operating above 25 °C; at 45 °C the flooded electro-
lyte batteries can present severe levels of corrosion and the gelled
electrolyte batteries can present considerable lifetime decrease
[21]. On the other hand commercially available hybrid capacitors
have temperature operating ranges of 40–60 °C allowing for bet-
ter system integration in most remote areas around the world.
SOCL set point of battery SOC from which and below the fuel
cell should be turned on (%)
SOCM set point of battery SOC from which and above one or
both the consumptions should be turned on (%)
Vc capacitor voltage (V)
Wij weight of FCM concept (–)
W3D desalinated water needed in order to cover the needs for
3 days (m3)
WTANK volume of stored desalinated water in the desalinated
water tank (m3)
Finally, the round trip efficiency of the hybrid capacitors in litera-
ture is found to exceed 98% in some cases [17], in comparison to
efficiencies of about 86% for deep discharge batteries [20].
Another important aspect in order to accomplish a combined
hybrid capacitors/hydrogen alternative is the needed energy man-
agement and control algorithms [22,23]. Without an intelligent
energy management system the benefits of a hybrid energy storage
system cannot be harvested. Different approaches have been pro-
posed and investigated in literature for the management of hybrid
energy storage systems [24–27]. The approaches that have already
been developed for the energy management in a polygeneration
microgrid can also be utilized [2,3], after their operational vari-
ables have been optimized for this configuration.
This paper presents the investigation of the possibility of
exchanging the battery bank in the APM topology with the combi-
nation of a hybrid capacitor bank and higher utilization of the
hydrogen subsystem. The investigation was planned to be realized
in two steps. The first step would include the experimental testing
of market available hybrid capacitors and the second step was
planned to comprise of a techno-economic comparison of an
APM based on deep discharge batteries and an APM based on the
combined hybrid capacitors/hydrogen approach through a case
study based on simulations.
Initially a market research took place and three manufacturers
that offer hybrid capacitors were identified. Samples for each of
these manufacturers were procured and laboratory testing took
place in order to investigate their ability to operate as energy buf-
fers in the polygeneration microgrid topology. Further experiments
took place for the chosen hybrid capacitor in order to measure its
real world energy storage capabilities and round-trip efficiency.
Using these experimental results, a comparison between an opti-
mized polygeneration microgrid utilizing the new approach and
an optimized polygeneration microgrid utilizing deep discharge
batteries took place through simulation.
2. Experimental investigation of the hybrid capacitors
2.1. Selection of hybrid capacitors for testing and choice of tests
A market investigation took place world-wide trying to identify
commercial hybrid capacitors. The result of this investigation was
three companies that all offer hybrid capacitors at different capac-
ities. Samples of the model of the hybrid capacitor with the highest
capacity offered by each manufacturer were gathered in the labo-
ratory. Fig. 1 presents these hybrid ultracapacitors.
The technical characteristics of the considered capacitors are
given in Table 1. The most notable differences between them are
in their capacity, rated voltage, and physical dimensions. Compa-
nies A and C state that they use lithium ion material for the pseudo
capacitance electrode.
 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415 407

hybrid capacitors passed this first test then the one with the larger
capacity was going to be chosen for further testing. Since the work
presented in this paper comprises the first step in designing and
building a hybrid capacitor bank, it is understandable that less
pieces for the same capacity means less complexity and easier
bank preparation.

2.2. Experimental setup

Regarding the experiments of the capacitors self-discharging,

the electronic circuit depicted in Fig. 2 has been implemented. In
this circuit, three relays, controlled by three digital outputs of the
NI-USB-6008 device, connect the terminals of one capacitor at a
time to a single-ended analog input channel of the device. This
way, the self-discharging operation of the three capacitors under
testing can be performed automatically.
Fig. 1. The three hybrid capacitors tested (from left to right 1000 F, 800 F and
For the charging and discharging experiments of the hybrid
200 F).
capacitors the electronic circuit of Fig. 3 has been designed and
implemented. In particular, the capacitor under test was connected
Table 1 in series with two resistors, namely R and Rshunt. The value of the
Electrical and physical characteristics of three hybrid capacitors. Rshunt is 0.1 X, while the value of R can be modified in order to
Parameter Company A Company B Company C
allow different charge and discharge currents. For example, for
the case of testing the 1000 F which has a rated voltage of
Capacity (F) 1000 800 200
Voltage rated (V) 2.3 2.3 3.8
2.3 Vdc, three resistances where chosen, a 1.1 X, a 2.2 X and a
Voltage rated min (V) 1.0 1.0 2.2 4.4 X. These resistances allow, following Ohm’s law, a discharge
Capacitance tolerance +10%/5% +30%/10% +20%/20% current equal to Vrated/Rs, where Rs is the sum of R and Rshunt.
Operational temperature 40 to +60 25 to +60 25 to +60 For R equal to 1.1 X the charge/discharge current is 1.92 A, for
range (°C)
R equal to 2.2 X the charge/discharge current is 1 A and for R equal
ESR at DC (mX) 14 15 50
Diameter (mm) 35 35 25 to 4.4 X the charge/discharge current is 0.51 A.
Length (mm) 88 70 40 With the electronic circuit of Fig. 3, the transfer between the
charging and discharging operations can be performed automati-
cally using two single-pole single-through relays (Rel_1 and
Rel_2). When Rel_1 is enabled then the contact K1 closes and the
One of the most important technical characteristics that needs circuit operates in charging mode. On the other hand, when the
to be known when designing a high energy storage at the kW h Rel_2 is enabled and Rel_1 is disabled, then the contact K2 closes
range hybrid capacitor bank of a battery-less polygeneration and the circuit operates in discharging mode. The Vdc voltage is
microgrid is the self-discharge of the capacitors. If the capacitors set to 2.3 V DC and is coming from a DC voltage power supply unit.
self-discharge in a matter of minutes or even hours, significant On the other hand, when the Rel_2 is enabled and Rel_1 is disabled,
energy is lost. After the evaluation of the self-discharge properties then the contact K2 closes and the circuit operates in discharging
of each sample were assessed, one capacitor was going to be cho- mode. In both cases, the Rs resistance determines the loop current.
sen to undergo further testing so as to measure its real world To measure the charging and discharging currents, the voltage
energy storage capacity, as well as its efficiency. If more than one across the capacitor’s terminals, but also to control the status of

C1 C2 C3
1000 F 800 F 200 F
12 Vdc 12 Vdc 12 Vdc

Rel_1 Rel_2 Rel_3

D1 D2 D3
1N4148 1N4148 1N4148

AnIn_1

R1 R2 R3
Q1 Q1 Q3
DOut_1
BC547 BC547 BC547
1 kΩ 1 kΩ 1 kΩ
DOut_2

DOut_3

GND

Fig. 2. The electronic circuit for the automatic self-discharging measurements of three hybrid capacitors.
408 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415
Fig. 3. The theoretical circuit of the experimental setup.
the two relays, the NI-USB-6008 multifunctional data acquisition
device from National Instruments has been employed. This is a
microcontroller-based device providing up to four differential ana-
log input channels, 8 single-ended analog input channels, and sev-
eral digital inputs and outputs. It has measurements resolution of
12 bits while its sampling rate reaches 10 kS/s. The voltage mea-
surement range is programmatically configured and extends from
10 to +10 Vdc. For the charge/discharge circuit, a differential ana-
log input in parallel with the shunt resistor has been used (AnIn_1
and AnIn_2). For the measurement of the capacitors voltage, a sin-
gle-ended analog input has been used (AnIn_3). The two relays
have been driven using two NPN transistors triggered by two dig-
ital outputs (DOut_1 and DOut_2) of the NI-USB-6008 respectively.
In Fig. 4, the implementation of the charge/discharge circuit is
illustrated. The NI-USB-6008 multifunctional data acquisition and
control device was connected to a personal computer via a USB
port. In the personal computer, a data acquisition and control soft-
ware application has been developed based on the graphical lan-
guage platform of National Instruments LabVIEW (ver. 2013)
developer’s suite.
Fig. 4. The electronic panel for the charge–discharge experiments of three hybrid
capacitors.
LabVIEW suite includes all the necessary functions to communi-
cate, configure and control the NI-USB-6008 device. The applica-
tion developed for experiments is illustrated in Fig. 5. As it is
shown there are two graphs which display in real-time fashion
the capacitor’s voltage and current. This application was used for
both the charge/discharge and self-discharge experiments circuits
described above.
Through this software application, all the acquired measure-
ment data were saved into measurement data files for post analy-
sis. In addition, the software allowed for events triggering, a
function that is being used to alter the operating mode of testing
according to either particular upper and lower voltage limits, or
elapsed time.
2.3. Self-discharge experiment
The self-discharge experiment was performed using the circuit
of Fig. 3 and the LabVIEW-based software application. In this
experiment, the three capacitors were monitored by the NI-USB-
6008 device almost simultaneously every twelve hours. A log file
of twenty-four days was generated. The results are presented in
Fig. 6. All the capacitors were initially fully charged at their rated
voltage. For both 1000 F and 800 F capacitors the rated voltage is
2.3 Vdc, while for the 200 F capacitor the rated voltage is 3.8 Vdc.
At the end of the 24th day of the experiment, the voltage remained
in each one of the three capacitors is given in Table 2.
All three capacitors performed well in the test and proved to be
able to be used in a bank designed for operation in a battery-less
polygeneration microgrid. For a time period equal to 3–5 days all
three capacitors showed very little self-discharge. The 800 F capac-
itor presented higher energy losses in the 24 day period than the
other two, but still, all are suited well for energy storage applica-
tions. The 1000 F hybrid capacitor was chosen to be used in the
next series of tests, since its higher capacity would allow easier
bank design and assembly in the future due to the fewer pieces
needed for the same energy storage capacity.
2.4. Energy storage capacity and efficiency experiment
The following tests were performed in order to evaluate the
charging and discharging behavior of the hybrid capacitor, calcu-
late the hybrid capacitor’s usable energy storage capacity and the
relevant round-trip efficiency. The circuit used for this purpose is
 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415 409

Fig. 5. A view of the LabVIEW-based software application for the experiments.

Table 3
Parameters of the charge/discharge tests.

Test (#) R (X) Rshunt (X) Rs (X) Iloop_max (A)

1 1.1 0.1 1.2 2
2 2.2 0.1 2.3 1
3 4.4 0.1 4.5 0.5

charged to its rated voltage and then discharged to the minimum

rated voltage. In all cases during charging, the charging current is
initially high and decreases as the capacitor is charged. During dis-
charge, the current was again initially high and decreased as the
minimum rated voltage was reached. The negative values of the
discharge current are due to the fact that the data acquisition
device has kept the same polarity for its analog input channel as
Fig. 6. Self-discharge of the hybrid capacitors.
in the charge mode.
The efficiency (eta) for each series resistance circuit was calcu-
lated using Eq. (1) and presented in Table 4. The usable energy
Table 2 capacity that was measured is the maximum amount of energy
Remaining capacitors’ voltage after 24 days of experiment. that can be drawn from the hybrid capacitor starting from the
Capacitor (F) Rated voltage (V) Remaining voltage (V) DV (V) DV (%)
rated voltage after being fully charged, down to the minimum
allowed voltage defined by the manufacturer and is presented in
200 3.8 3.78 0.02 0.53
Table 4. The manufacturer gives as nominal energy capacity of each
800 2.3 2.03 0.27 11.74
1000 2.3 2.21 0.09 3.91 hybrid capacitor the value of 0.60 W h. The efficiency in all cases

the one depicted in Fig. 2. The charge was terminated when the
rated voltage, namely the 2.3 Vdc, was reached. On the other hand,
the discharging was terminated when the voltage measured at the
terminals of the capacitor reached the minimum permitted voltage
of the 1 Vdc. Three tests were made with different charge/dis-
charge currents. Instead of using random current values, values
of 2’s multiples were used. Specifically, currents of 0.5 A, 1 A, and
2 A were regulated using multiples of 1.1 X resistance. Table 3 pre-
sents the details for the resistance values and the associated cur-
rents related to the three charge/discharge tests.
The sampling rate was 250 ms. The results are illustrated in the
graphs of Figs. 7 and 8. It can be seen that the capacitor was Fig. 7. Charge of the 1000 F hybrid capacitor.
410 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415

Fig. 10. Cycling of the 1000 F hybrid capacitor with Rs = 4.5 X.

Fig. 8. Disharge of the 1000 F hybrid capacitor.

charging cycle multiplied by the efficiency of the capacitor. It can

be observed that after charging stopped, there was a small drop
Table 4 to the voltage of the capacitor. The opposite behavior was observed
Efficiency and usable energy of the 1000 F hybrid capacitor. when discharging stopped, where a small rise to the voltage of the
Circuit Rs (X) Efficiency (%) Usable energy capacity (W h) capacitor occurred. While such behavior is not observed in capac-
1.15 92.7 0.578
itors or ultracapacitors, a behavior with similarities in different
2.30 92.7 0.614 magnitude can be observed in lithium ion batteries [29,30]. Taking
4.60 92.7 0.617 in consideration that a hybrid capacitor has one electrode typical of
an ultracapacitor and one typical of a lithium-ion battery, this
behavior can be regarded as an inherited behavior from the lithium
remained practically the same, but the capacity was dependent on ion electrode.
the current being drawn from the hybrid capacitor. As the current A greater series of charging and discharging cycles is presented
increases, the available energy that can be drawn from the hybrid in Fig. 10. This time the capacitor was connected to the 4.5 X series
capacitor decreases. Comparable behavior is typical of deep dis- resistance circuit. The cycles begin with the voltage being equal to
charge batteries and has also been observed in lithium ion batter- 1 V. After about 200 min the same behavior that was observed in
ies [28]. It can be regarded as an inherited behavior due to the Fig. 9 is viewed in Fig. 10 as well.
lithium ion electrode of the hybrid capacitor.
R tdisch 3. Design and sizing through optimization
IC ðtÞ  V C ðtÞdt
eta ¼ R0 tch  100% ð1Þ
0
IC ðtÞ  V C ðtÞdt 3.1. Method employed

The autonomous polygeneration microgrid is presented in

2.5. Cycling tests
The operation of the 1000 F hybrid capacitor was also tested
under cycling conditions of continuous charging and discharging.
In Fig. 9 the capacitor was connected to the 1.2 X series resistance
circuit. Each test cycle included four discrete functional states,
namely charging for 60 s, no current for 60 s, discharging for 60 s,
and no current for 60 s. Negative current values shown in Fig. 9
during the discharge state was due to the current flow direction
reversal.
For this specific circuit and range voltage presented in Fig. 9 the
energy that was drawn from the hybrid capacitor in the 60 s of the
capacitor was equal to the energy that is fed to the capacitor during
Fig. 9. Cycling of the 1000 F hybrid capacitor with Rs = 1.2 X.
Fig. 11. The Energy Storage Bank can either be a hybrid capacitors
bank or a typical deep discharge batteries bank. The optimization
method used for the design and sizing of the microgrids was based
on Particle Swarm Optimization (PSO) and has been presented in
detail in [1]. PSO is an effective global optimizer than can be used
for continuous variables, discrete variables or combinations of
both. Through the use of a penalty method it can accommodate
constraints easily. A disadvantage of PSO is the sensitivity it has
concerning its parameter settings, since a small change in the
parameters could lead to a considerable effect. On the other hand
though it has very little parameters to fine tune [31]. Research
has been performed in the fine tuning of PSO parameters and effec-
tive parameter sets have been proposed [32,33]. Most of the clas-
sical optimization techniques, which need derivative information
of the objective function in order to determine the search direction,
cannot be used when the problem is complex, multi-dimensional,
non-convex and non-differentiable. Modern heuristic techniques
like PSO, genetic algorithms, tabu search and ant colony optimiza-
tion can be used effectively to solve multidimensional optimiza-
tion problems [34]. Moreover, improper selection of the initial
condition usually leads to convergence problems when applying
conventional optimization techniques [35]. Finally, one major
advantage of PSO is that it can handle the whole optimization
problem easily and naturally, without having to divide it to sub-
problems and apply various methods for solving each sub-prob-
lem, which is the case with the use of conventional optimization
methods [36]. The PSO has been investigated as an optimization
approach for energy systems extensively and proved to provide
good results [1–3,32,37–41]. The PSO settings used for this
 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415
Fig. 11. Autonomous polygeneration microgrid.
Table 5
PSO settings.
Topology Lbest
Neighborhood size 3
Particles 20
Generations 100
Seed 0 by the respective company. The interest rate was considered to be
Constriction gain 0.729
Cognitive acceleration constant 2.05
Social acceleration constant 2.05
optimization are presented in Table 5. It was realized with a com-
bination of software packages; TRNSYS 17, GenOpt 3 and TRNOPT.
The combination of these versions of the software packages allow
the simultaneous optimization of up to 100 variables – it is noted
that earlier versions allowed less than 10 variables, which meant
that the optimization problem had to be broken down in discrete
steps. This method would identify in the end the system with the
lowest Net Present Cost for a 20 year operation, which at the same
time managed to cover the needs in all products at 100%.
3.2. Case study parameters
The microgrid was considered to be installed on an island in the
Aegean Sea, Greece. Typical meteorological data of this area were
used. Two households were considered to be occupied by 8 people
in total. The heating and cooling of the households was covered by
an air to air heat pump (Daikin FTXR28EV1B9) with a nominal COP
equal to 5.14. Two StecaPF 166 refrigerators were used in each
house, one in fridge mode and one in freezer mode for covering
the refrigeration needs. The daily desalinated water consumption
was assumed to be 1.92 m3 and this includes the drinkable water,
laundry, kitchen, sanitary and bathing use. The two hybrid hydro-
gen battery scooters [42] used were assumed to cover 50 km and
their daily consumption of hydrogen to be 2.4 N m3 H2. For lighting
LED lamps were considered each of 8 W rated power. All electrical
411
devices were considered to have an energy rating of A/A+. For
cooking an induction cooker and a grill were considered.
Typical market prices were used for all the components. The
wind turbine utilized in both microgrids was a typical 7.5 kW DC
one. As far as the hybrid capacitors were concerned the price of
25 € for each 1000 F capacitor was used, which is a real quote given
equal to 6%.
The available routines present in TRNSYS were used in combi-
nation with the routines presented in [3] for the hydrogen compo-
nents, the desalination unit and the rest of the microgrid
components needed for the simulation of the microgrid. The simu-
lation time-step for the microgrid based on batteries was 1 h, while
for the microgrid based on hybrid capacitors was 3 min.
3.3. Hybrid capacitor bank
In order to simulate the hybrid capacitor bank the experimental
results presented in Section 2 of this paper were used for the cho-
sen hybrid capacitor. A capacitor bank is comprised by the capac-
itors and the relevant electronic circuits for their interconnection.
In order to calculate the efficiency of the hybrid capacitors bank,
the efficiency of the hybrid capacitors has to be multiplied by the
efficiency of the electronic circuits. The efficiency of the hybrid
capacitors was measured to be 92.7% and the efficiency of the elec-
tronic circuits is considered to be 97%. Thus, the overall bank effi-
ciency is considered to be equal to 90% (97%  92.7%  90%). The
results of the experimental testing showed that the remaining
available energy of each hybrid capacitor depends on the magni-
tude of the current that is being drawn from it. When considering
a hybrid capacitor bank that is able to cover the energy needs of a
polygeneration microgrid at least for a few minutes, until the fuel
cell can be activated in order to supply the energy needed at the
worst case scenario, when there is no power production, a big
number of hybrid capacitors ought to be used. For example a
350 W h bank needs about 500 pieces of the 1000 F hybrid capac-
itor. It is understandable that some of these hybrid capacitors have
412 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415

to be connected in series and a number of these series of hybrid t7: (SOC < SOCM) OR (WTANK > W3D).
capacitors have to be connected in parallel. With an effective bank
design, taking into consideration the maximum power than will be The flow matrix of the Petri Net is presented below in Eq. (2):
drawn from the bank, the maximum current being drawn from 0 1
1 1 0 0 0
each capacitor can be kept at a low value, in order to ensure the B 1 0
highest obtainable energy capacity of each hybrid capacitor. For B 0 0 1 C
C
B C
the purposes of this case study, taking into consideration the B 0 0 1 0 1 C
B C
experimental results presented in Table 4, the nominal energy B C
FM ¼ B 0 0 0 1 1 C ð2Þ
B C
capacity that is given by the manufacturer was used, which is B 1 1 0 0 0 C
B C
equal to 0.60 W h. B C
@ 1 0 1 0 0 A
3.4. Energy management system 1 0 0 1 0
By definition, concepts (Cn) of the Fuzzy Cognitive Maps (FCMs)
The energy management system used in the polygeneration present different aspects of the modeled system such as inputs, out-
microgrid with the batteries along with the explanation of all rel- puts, rules or intermediate states [3]. The value of each concept is
evant parameters has been presented in detail in [3], and is based fuzzified in the space [0, 1]. Five concepts are used in the relevant
on a combined Petri Nets – Fuzzy Cognitive Maps approach. For the Fuzzy Cognitive Maps presented in Fig. 12, namely:
microgrid featuring the hybrid capacitors bank the above energy C1: frequency.
management system was adapted accordingly and presented in C2: hybrid capacitor bank State of Charge (SOC).
Fig. 12 [3]. The Petri Net in this energy management system was C3: electrolyzer operation point.
used in order to decide the operational mode of the microgrid. C4: desalination unit operation point.
There are four discrete states which are presented in Table 6 and C5: fuel cell operation point.
the parameters depicted in Fig. 12, which determine the operation
of the Petri Net, are analyzed below: The various weights of the FCMs (Wij) depicted in Fig. 12 define
their relationships. The sigmoid function with a steepness param-
t1: SOC < SOCL, where SOCL (Hybrid Capacitors Bank State of eter equal to 1 was used in the FCMs [3].
Charge Low) is a set point below of which the fuel cell is turned
on. 3.5. Optimization parameters and results
t2: SOC > SOCM, where SOCM (Hybrid Capacitors Bank State of
Charge Medium) is a set point of SOC above of which one or The optimal microgrid of each configuration investigated is the
both the consumptions (electrolyzer, desalination unit) are one that presents the lowest Net Present Cost (NPC) for a 20 year
turned on. investment period and at the same time fulfils all the set technical
t3: WTANK > W3D, where WTANK is the desalinated water cur- constraints. The technical constraints used were:
rently stored in the desalinated water tank and W3D is the desa-
linated water needed to cover the needs for 3 days. – The Energy Storage Banks ought not to be deep discharged. For
t4: WTANK < W3D. the deep discharge batteries the limit of the fractional State of
t5: SOC > SOCL. Charge is set at 20% and for the hybrid capacitors bank is equal
t6: (SOC < SOCM) OR (WTANK < W3D). to 1%.

Fig. 12. Adapted combined Petri Nets Fuzzy Cognitive Maps energy management system.
 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415 413

Table 6
Petri Net modes for the adapted combined Petri Nets Fuzzy Cognitive Maps energy management system.

Mode Fuel cell Desalination unit Electrolyzer Remarks

1 OFF OFF OFF –
2 Decided by the energy management OFF OFF –
system
3 OFF Decided by the energy management Decided by the energy management Priority: electrolyzer
system system
4 OFF Decided by the energy management Decided by the energy management Priority: desalination
system system unit

– The desalinated water tank ought never to get empty storage tank was also enlarged, from 25 N m3 H2 to 47.5 N m3 H2.
(WTANK > 0). This was expected because the hydrogen subsystem is used in this
– The low pressure hydrogen tank used for hydrogen storage configuration also for short to medium term energy storage and its
ought never to get empty (HTANK > 0). overall efficiency is lower than the efficiency of the battery bank
– The stored desalinated water and hydrogen in the beginning of utilized in the other configuration. The desalination unit and desa-
the year are less or equal than the stored quantities in the end of linated water tank were comparable in the two polygeneration
the year. microgrid approaches. Water tanks of this size are available com-
mercially usually in cylindrical shapes and have approval for drink-
For each timestep when one or more of the above technical con- ing water storage. The parameters of the energy management
straints were not met, a penalty equal to a big amount of money for system were different, as expected, because they were optimized
each constraint was added to the NPC. for each specific polygeneration microgrid configuration. It has to
The cost function (CF) used is presented in Eq. (3). be pointed out though, that the combined Petri Nets – Fuzzy Cog-
nitive Maps energy management system can accommodate a range
X
8760 X
8760 X
8760
CF ¼ NPC þ PES ðtÞ þ PH2 ðtÞ þ PW ðtÞ þ PTANKS ð3Þ of energy buffer combinations and through its parameters’ optimi-
t¼1 t¼1 t¼1 zation it can operate a polygeneration microgrid effectively.

where NPC is the Net Present Cost for a 20 year period; PES the
energy storage penalty; PH2 the hydrogen penalty; Pw the water
penalty; and PTANKS is the tanks penalty. If the stored water and/
or stored hydrogen are less in the end of the year than its beginning
another 100,000 € are added.
The optimization variables along with the optimal values for the
polygeneration microgrid with the hybrid capacitors are presented
in Table 7 and the optimization variables along with the optimal
values for the polygeneration microgrid with the conventional
deep discharge batteries in Table 8.
The Net Present Cost for 20 year operation of the optimal poly-
generation microgrid with the hybrid capacitors is 139,295 € and
for the one with the batteries is 95,179 €. It is clear that the utiliza-
tion of typical deep discharge batteries led to a decreased NPC by
32%. The actual cost of the hybrid capacitors is equal to
27,800 €. It can also be seen that the polygeneration microgrid with
the hybrid capacitors needed higher installed photovoltaic power
(1.26 kWp) and the hydrogen components were also of higher
power rating – the fuel cell was 1800 W instead of 800 W and
the electrolyzer was 1900 W instead of 1200 W. The hydrogen
4. Discussion
It is clear from the results that it is technically feasible to create
an autonomous polygeneration microgrid without conventional
batteries today. This is made possible due to the combination of
hybrid capacitors, which can provide the needed energy to form
the microgrid, and the fuel cell which can be activated in the cases
where there is an energy deficit in the microgrid. From an econom-
ical point of view though, it can be clearly seen that the investment
in typical deep discharge batteries is more profitable, even taking
into consideration the needed replacements throughout the oper-
ational life time of the microgrid, than investing in hybrid capaci-
tors and enlarged hydrogen components.
On the other hand, though, hydrogen components and hybrid
capacitors possess a clear potential as far as expected price drops
are concerned. It has to be mentioned that ultracapacitor prices
have declined 99% in the last 10 years, compared with 30–40%
reduction for batteries [43] and the potential for further consider-
able reductions is still very strong. Two market reports published

Table 7
Polygeneration microgrid with hybrid capacitors.

Variable Lowest value Highest value Step Optimal value

System components
Hybrid ultracapacitors 1000 F (pieces) 500 2000 Continuous 1112
Typical multicrystalline modules rated at 180 Wp each in series 25 30 1 28
Rated power of the fuel cell (W) 1500 2800 100 1800
Rated power of the electrolyzer unit (W) 1500 2500 100 1900
Low pressure hydrogen tank storage capacity (N m3 of H2) 30 60 2.5 47.5
Desalinated water tank volume (m3) 10 60 5 30
Rated power of the desalination unit (W) 800 1700 100 1100
Energy management system variables
SOCL 35 50 1 45
SOCM 70 100 1 94
W13 1 0 Continuous 0.67
W14 1 0 Continuous 0.78
W15 0 1 Continuous 0.78
W23 0 1 Continuous 0.91
W24 0 1 Continuous 0.70
W25 0 1 Continuous 0.69
414 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415

Table 8
Polygeneration microgrid with batteries.

Variable Lowest value Highest value Step Optimal value

System components
Capacity rating of each of the 2 V batteries. 24 are used for a 48 V DC bus (W h) 400, 500, 600, 800, 1000, 1200 500
Typical multicrystalline modules rated at 180 Wp each in series 20 28 1 21
Rated power of the fuel cell (W) 700 1800 100 800
Rated power of the electrolyzer unit (W) 700 2000 100 1200
Low pressure hydrogen tank storage capacity (N m3 of H2) 10 40 2.5 25
Desalinated water tank volume (m3) 20 60 5 45
Rated power of the desalination unit (W) 800 1500 50 1000
Energy management system variables
SOCL 15 40 1 38
SOCM 80 100 1 93
W13 1 0 Continuous 0.70
W14 1 0 Continuous 0.71
W15 0 1 Continuous 0.85
W23 0 1 Continuous 0.74
W24 0 1 Continuous 0.60
W25 0 1 Continuous 0.67

in 2013 were very optimistic concerning hybrid capacitors and ult-
racapacitors in general. The first report forecasted a 3.5 bn US$
energy storage revenues from the ultracapacitor market, which
will account for 5% of the energy storage market in 2020 [44].
The second report predicted a steady price decline. For 2018 a
15% fall in prices was forecasted to come upon due to materials
innovation decreasing the pricing to 0.0082 US$/F. It stated that a
high voltage operation at 3.5 V instead of 2.7 V could lower prices
a further 40% [45]. Research in graphene and nanostructured car-
bons could lead to higher capacity also in the near future, with cor-
responding impact to pricing [45]. Currently, based on real quotes
given by companies, hybrid capacitors present similar pricing as
the pricing of ultracapacitors rated at the same capacity.
As far as hydrogen is concerned the driving force behind
expected price decreases is the automotive industry. Many major
automobile manufacturers have gone forward and produced fuel
cell vehicles [46]. One of the biggest barriers in hydrogen vehicles
market growth is the lack of relevant infrastructure. Currently in
Germany 85 hydrogen refueling stations are planned to be build
focusing on expansion of H2 regions and covering their connections
including autobahns. Comparable activities have started in Eng-
land, France, Switzerland, California and four metropolitan areas
of Japan [47]. Toyota presented in the 2013 Tokyo Motor Show a
new fuel cell concept based on which a fuel cell production car will
be launched in 2015 [48]. A partnership between Daimler, Ford and
Nissan is planning to start the production of 100,000 fuel cell vehi-
cles by 2017 [47]. The costs of fuel cells can be reduced considerably
because of technology advances and economy of scale and even the
reduction of the fuel cell costs on the level of conventional drive
trains in the automotive industry is also considered possible [47].
The momentum and dynamics of both the hybrid/ultracapacitor
and hydrogen market seem significant. Their decreasing costs will
allow the proposed battery-less polygeneration microgrid concept
presented in this paper to become competitive economically-wise
in the coming years.
5. Conclusion
In this paper the possibility of exchanging the battery bank in
the autonomous polygeneration microgrid topology, with hybrid
capacitors was investigated. Commercial hybrid capacitors were
tested under laboratory conditions. All samples presented techni-
cal characteristics capable of energy storage use in an autonomous
battery-less polygeneration microgrid. Based on the results of the
tests and simulations it was found that it is technically feasible
today to create a battery-less polygeneration microgrid. The energy
storage is shifted from deep discharge batteries to short-term
energy storage in the hybrid capacitors bank and the hydrogen
subsystem for the medium and longer term energy storage. The
combined Petri Nets – Fuzzy Cognitive Maps energy management
system proved to be able to manage a range of energy buffer com-
binations effectively through its parameters’ optimization.
Due to the high cost of hybrid capacitors and hydrogen compo-
nents this approach is not competitive economically-wise cur-
rently (end of 2014). On the other hand a considerable drop in
the pricing of both capacitors and hydrogen components is
expected to occur in the short term and before 2020. This will
make this approach economically competitive.
Future work includes the design and realization of a 600 kF
hybrid capacitor bank rated at 48 V and controlled by computa-
tional intelligence algorithms. This bank is to be tested both at lab-
oratory scale and at the experimental polygeneration microgrid
installed at the Agricultural University of Athens.
Acknowledgments
Part of the presented work was conducted within the frame-
work of the research project ARISTEIA - Smart Desalination - 529,
co-funded by EU and the General Secretariat for Research and
Technology (GSRT).
References
[1] Kyriakarakos G, Dounis AI, Rozakis S, Arvanitis KG, Papadakis G.
Polygeneration microgrids: a viable solution in remote areas for supplying
power, potable water and hydrogen as transportation fuel. Appl Energy
2011;88:4517–26.
[2] Kyriakarakos G, Dounis AI, Arvanitis KG, Papadakis G. A fuzzy logic energy
management system for polygeneration microgrids. Renew Energy
2012;41:315–27.
[3] Kyriakarakos G, Dounis AI, Arvanitis KG, Papadakis G. A fuzzy cognitive maps–
petri nets energy management system for autonomous polygeneration
microgrids. Appl Soft Comput 2012;12:3785–97.
[4] Kyriakarakos G, Piromalis DD, Dounis AI, Arvanitis KG, Papadakis G. Intelligent
demand side energy management system for autonomous polygeneration
microgrids. Appl Energy 2013;103:39–51.
[5] Papadakis G, Mohamed ES, Kyriakarakos G, Kassem A-WS. Hybrid renewable
energy systems for the supply of services in rural settlements of
Mediterranean partner countries. The HYRESS project – the case study of the
hybrid system in Egypt. In: 24th European photovoltaic solar energy
conference. Hamburg, Germany; 2009. p. 4019–24.
[6] Tan X, Li Q, Wang H. Advances and trends of energy storage technology in
microgrid. Int J Electr Power Energy Syst 2013;44:179–91.
[7] Divya KC, Østergaard J. Battery energy storage technology for power systems—
an overview. Electr Power Syst Res 2009;79:511–20.
[8] Zhu WH, Zhu Y, Tatarchuk BJ. A simplified equivalent circuit model for
simulation of Pb–acid batteries at load for energy storage application. Energy
Convers Manage 2011;52:2794–9.
 G. Kyriakarakos et al. / Energy Conversion and Management 91 (2015) 405–415
[9] Ayad M-Y, Pierfederici S, Raël S, Davat B. Voltage regulated hybrid DC power
source using supercapacitors as energy storage device. Energy Convers Manage
2007;48:2196–202.
[10] Jossen A, Garche J, Sauer DU. Operation conditions of batteries in PV
applications. Sol Energy 2004;76:759–69.
[11] Poullikkas A. A comparative overview of large-scale battery systems for
electricity storage. Renew Sustain Energy Rev 2013;27:778–88.
[12] Glavin ME, Hurley WG. Optimisation of a photovoltaic battery ultracapacitor
hybrid energy storage system. Sol Energy 2012;86:3009–20.
[13] Catherino HA, Burgel JF, Shi PL, Rusek A, Zou X. Hybrid power supplies: a
capacitor-assisted battery. J Power Sources 2006;162:965–70.
[14] Sikha G, Popov BN. Performance optimization of a battery–capacitor hybrid
system. J Power Sources 2004;134:130–8.
[15] Sharma P, Bhatti TS. A review on electrochemical double-layer capacitors.
Energy Convers Manage 2010;51:2901–12.
[16] Burke A. Ultracapacitors: why, how, and where is the technology. J Power
Sources 2000;91:37–50.
[17] Sun X, Zhang X, Zhang H, Xu N, Wang K, Ma Y. High performance lithium-ion
hybrid capacitors with pre-lithiated hard carbon anodes and bifunctional
cathode electrodes. J Power Sources 2014;270:318–25.
[18] Choi HS, Park CR. Theoretical guidelines to designing high performance energy
storage device based on hybridization of lithium-ion battery and
supercapacitor. J Power Sources 2014;259:1–14.
[19] Uzunoglu M, Alam MS. Dynamic modeling, design and simulation of a PEM
fuel cell/ultra-capacitor hybrid system for vehicular applications. Energy
Convers Manage 2007;48:1544–53.
[20] Hittinger E, Wiley T, Kluza J, Whitacre J. Evaluating the value of batteries in
microgrid electricity systems using an improved Energy Systems Model.
Energy Convers Manage 2015;89:458–72.
[21] Newnham RH, Baldsing WGA. Performance of flooded- and gelled-electrolyte
lead/acid batteries under remote-area power-supply duty. J Power Sources
1997;66:27–39.
[22] Ayad MY, Becherif M, Henni A, Aboubou A, Wack M, Laghrouche S. Passivity-
based control applied to DC hybrid power source using fuel cell and
supercapacitors. Energy Convers Manage 2010;51:1468–75.
[23] Hannan MA, Azidin FA, Mohamed A. Multi-sources model and control
algorithm of an energy management system for light electric vehicles.
Energy Convers Manage 2012;62:123–30.
[24] Dursun E, Kilic O. Comparative evaluation of different power management
strategies of a stand-alone PV/Wind/PEMFC hybrid power system. Int J Electr
Power Energy Syst 2012;34:81–9.
[25] Hajizadeh A, Golkar MA. Control of hybrid fuel cell/energy storage distributed
generation system against voltage sag. Int J Electr Power Energy Syst
2010;32:488–97.
[26] Etxeberria A, Vechiu I, Camblong H, Vinassa JM. Comparison of three
topologies and controls of a hybrid energy storage system for microgrids.
Energy Convers Manage 2012;54:113–21.
[27] Mohamed FA, Koivo HN. System modelling and online optimal management of
microgrid using mesh adaptive direct search. Int J Electr Power Energy Syst
2010;32:398–407.
[28] Peng R, Pedram M. An analytical model for predicting the remaining battery
capacity of lithium-ion batteries. IEEE transactions on very large scale
integration (VLSI) systems, vol. 14; 2006. p. 441–51.
415
[29] Xing Y, He W, Pecht M, Tsui KL. State of charge estimation of lithium-ion
batteries using the open-circuit voltage at various ambient temperatures. Appl
Energy 2014;113:106–15.
[30] Abu-Sharkh S, Doerffel D. Rapid test and non-linear model characterisation of
solid-state lithium-ion batteries. J Power Sources 2004;130:266–74.
[31] Lovberg M, Krink T. Extending particle swarm optimization with self-
organized critically. In: 4th Congress evolutionary computation. IEEE,
Honolulu, Hawaii, USA; 2002. p. 1588–93.
[32] Boonbumroong U, Pratinthong N, Thepa S, Jivacate C, Pridasawas W. Particle
swarm optimization for AC-coupling stand alone hybrid power systems. Sol
Energy 2011;85:560–9.
[33] Papageorgiou EI, Parsopoulos KE, Stylios CS, Groumpos PP, Vrahatis MN. Fuzzy
cognitive maps learning using particle swarm optimization. J Intell Inform Syst
2005;25:95–121.
[34] Hazra J, Sihha AK. Environmental constrained economic dispatch using
bacteria foraging optimization. In: Joint international conference on power
systems technology and IEEE power India conference 2008. New Delhi, India;
2008. p. 1–6.
[35] Yorino N, El-Araby EE, Sasaki H, Harada S. A new formulation for FACTS
allocation for security enhancement against voltage collapse. Power Eng Rev,
IEEE 2002;22:68.
[36] Yamin HY. Dynamic optimal power flow using interior point method and
benders decomposition considering active and reactive constraints. Electr
Power Compon Syst 2006;34:1377–93.
[37] Lee W-S, Chen YT, Wu T-H. Optimization for ice-storage air-conditioning
system using particle swarm algorithm. Appl Energy 2009;86:1589–95.
[38] Kornelakis A. Multiobjective Particle Swarm Optimization for the optimal
design of photovoltaic grid-connected systems. Sol Energy 2010;84:2022–33.
[39] Kornelakis A, Marinakis Y. Contribution for optimal sizing of grid-connected
PV-systems using PSO. Renew Energy 2010;35:1333–41.
[40] El-Zonkoly AM. Optimal placement of multi-distributed generation units
including different load models using particle swarm optimization. Swarm
Evol Comput 2011;1:50–9.
[41] Avril S, Arnaud G, Florentin A, Vinard M. Multi-objective optimization of
batteries and hydrogen storage technologies for remote photovoltaic systems.
Energy 2010;35:5300–8.
[42] Lin B. Conceptual design and modeling of a fuel cell scooter for urban Asia. J
Power Sources 2000;86:202–13.
[43] Werkstetter S. Existing and future ultracapacitor-applications in the
renewable energy market. PCIM Europe 2014. In: Proceedings of
international exhibition and conference for power electronics, intelligent
motion, renewable energy and energy management; 2014. p. 1–7.
[44] Media MR. Ultracapacitor market forecast 2015–2020; 2013.
[45] Lux Research I. Power play: supercapacitor innovation for growth in
transportation and electronics; 2013.
[46] U.S.D. Partnership. Hydrogen Storage Technical Team Roadmap; 2013.
[47] Wind J. Progress on FCEV development. IEA Hydrogen Roadmap Workshop,
Paris, France; 2013.
[48] Roussel J. Fuel cell vehicles by Toyota. IEA Hydrogen Roadmap Workshop,
Paris, France; 2013.