energies

Article
Dynamic Analysis of a Hybrid Energy Storage System
(H-ESS) Coupled to a Photovoltaic (PV) Plant
Linda Barelli 1, * ID , Gianni Bidini 1 , Fabio Bonucci 2 , Luca Castellini 3 ID
, Simone Castellini 4 ,
Andrea Ottaviano 1 ID , Dario Pelosi 1 and Alberto Zuccari 5
1 Department of Engineering, University of Perugia, Via G. Duranti 1/A4, 06125 Perugia, Italy;
gianni.bidini@unipg.it (G.B.); panfilo.ottaviano@unipg.it (A.O.); dario.pelosi@gmail.com (D.P.)
2 VGA Srl, Via dell’Innovazione SNC, 06053 Deruta, Italy; fabio.bonucci@vgasrl.com
3 Umbra Cuscinetti SpA, Via V. Baldiccini 1, 06034 Foligno, Italy; lcastellini@umbragroup.com
4 ERA Electronic Systems Srl, Via G. Benucci 206, 06135 Perugia, Italy; simone.castellini@eraes.it
5 QFP Srl, Via Gullotti 31, 06049 Spoleto, Italy; zuccari@qfp-service.it
* Correspondence: linda.barelli@unipg.it; Tel.: +39-075-585-3740

Received: 10 January 2018; Accepted: 2 February 2018; Published: 8 February 2018

Abstract: Nowadays energy storage is strongly needed to allow grid safety and stability due to the
wide penetration of renewable plants. Mainly economic and technological issues impede a relevant
integration of conventional storage devices in the energy system. In this scenario, the hybridization
of different storage technologies can be a techno-economic solution useful to overcome these issues
and promote their diffusion. Hybridization allows multi-operation modes of the Energy Storage
System (ESS), merging the positive features of base-technologies and extending their application
ranges. This paper provides a dynamic analysis of a hybrid energy storage system (H-ESS) consisting
of a flywheel and a battery pack coupled to a photovoltaic generation plant and a residential load
up to 20 kW. A dynamic model of the overall micro-grid (MG) was developed implementing the
H-ESS preliminary sizing and a suitable management algorithm. The instantaneous behavior of each
component was evaluated. A brief summary of the MG performance at different weather and load
conditions was provided together with a characterization of the impact of power fluctuations on the
battery current and on the power exchange with the grid.

Keywords: load management; micro-grid; battery; flywheel; simulations

1. Introduction
Recent studies on renewable energy sources (RES) show complexity in controlling and offsetting
over and under production events [1,2]. As described in [3], the main limitations of RES are due to their
intermittent and fluctuating behaviour, especially for solar and wind power plants. For these reasons,
RES are defined as non-programmable energy sources that negatively affect the stability and safety
of the electric grid [4]. In fact, thermal power plants, with particular reference to combined cycles
fed by natural gas, are forced to compensate continuously their power variations avoiding network
imbalances. In particular, the conventional power plants cycling leads to a significant deterioration in
their overall performance (e.g., efficiency, wear) [5–9].
Synchronization of network reserves and ESS integration in the electric grid can be seen as
two effective and complementary solutions to overcome the above-mentioned RES technical limits.
This is consistent with the objectives of the IEC T120 work program [10], where Energy Storage
Systems (ESSs) are identified as a solution to efficiently deliver sustainable, economic and secure
electricity supplies. Indeed, as discussed in [11–13], thanks to their excellent skills in terms of scalability
and efficiency, ESSs represent one of the best solutions available for a better RES exploitation and
penetration. Specifically, the most interesting energy storage technologies suitable to achieve RES

Energies 2018, 11, 396; doi:10.3390/en11020396 www.mdpi.com/journal/energies

Energies 2018, 11, 396 2 of 23

penetration targets set by the EU (European Union) are presented in [11]. An interesting review is
provided in [14] where the current technology readiness level of various ESS technologies is analysed.
In this context, the importance of storage systems also in micro-grids (MGs) to extend the RES
penetration, especially in remote and rural areas, has to be carefully analyzed. A MG can be defined as
“a network of low voltage power generating units, storage devices and loads capable of supplying
a local area such as suburban area, an industry or any commercial area with electric power and
heat” [15]. Therefore, an innovative systemic solution can be the integration of different energy storage
systems into micro-grids that will permit to overcome the intrinsic limitations of individual storage
technologies. In fact, in order to achieve future aims about the reduction of the greenhouse gas
emissions and polluting species, future electricity network should be reliable, flexible, accessible and
economically viable [16].
Flywheels represent a specific energy storage technology [12]. They can provide power in short
time applications and are characterized by long lifetimes, high efficiency and fast response [17].
Flywheels are often used to reach energy quality and stability improvement [18–20], power
smoothing [21], renewable energies integration support [22,23].
In the technical literature, several studies using flywheels in mobile applications are presented [24,25].
In particular, an in depth discussion about ESS for electric vehicle applications is provided in [26].
Moreover, broad comparative reviews of ESS for stationary applications are presented in [27,28].
In particular, in [29] an interesting solution, for residential applications, composed by a hybrid generation
system was analysed. Specifically, it consists of a PV (photovoltaic)-wind power plant coupled with a
flywheel-based storage system, aiming to clipping grid consumption by optimizing the RES exploitation.
So dynamic power management strategy and algorithm were developed with the purpose to achieve a
controlled power output in a generation stochastic system.
The integration of flywheels in wind power plants is also discussed in [30]. Specifically, a solution
to maximize the extracted power characterized by a stochastic behavior, by coupling a flywheel with
a wind generator is presented. The same authors in [21] highlighted the smoothing function of the
flywheel in wind generation applications.
Flywheel hybridization with other technologies characterized by higher energy capacity can
extend their application range. At the same time, hybridization provides, mainly thanks to the
fast response of flywheels, beneficial effects towards the other base technologies. If batteries are
considered, the potential enhancement of their duration due to the flywheel peak shaving function
is highlighted. Currently, only one large-scale hybrid storage plant is planned thanks to a EU’s
Horizon2020 project [31], while relative to small-sized plants only in [32] is the flywheel-battery hybrid
configuration is addressed. Anyway, the management strategy is not described and the dynamic
modelling and analysis of operating modes along the year are not investigated for both battery pack
and flywheel.
For the above reasons, the present work aims to fill this gap providing a dynamic analysis of the
power flow interaction among the components of a micro-grid (MG) characterized by a flywheel-battery
hybrid storage system (H-ESS) coupled with a PV plant and a complex user load. In the investigation
of the H-ESS behaviour also the environmental conditions under which the RES plant operates are
taken into account.
In the chosen architecture, the battery pack is the main energy storage system while the flywheel
has a peak shaving function. In particular, as anticipated, in order to investigate the dynamic interaction
among the components of a micro-grid, a dynamic model was developed in the Matlab Simulink
environment. Also a power flow management strategy was configured, in order to maximize the
MG energy independence from the grid, and implemented in the Simulink model. Simulations were
carried out under different operating and weather conditions to evaluate the energy performance of
the MG and to characterize the dynamic interaction among plant devices and their working profiles.
Simulations results, which will be the basis of further developments of plant components, confirm the
Energies 2018, 11, 396 3 of 23

flywheel peak shaving function, rising the quality of the battery exchanged current and of the power
exchanged with the grid.
Energies 2018, 11, x FOR PEER REVIEW 3 of 23
2. Micro-Grid Layout
2. Micro-Grid
The electricalLayout
architecture of the micro-grid here studied was picked out by comparing several
alternatives
The electricalthose
among proposed
architecture of theinmicro-grid
the literature. In [33]
here studied was different
picked outarchitecture
by comparing models
severalof the
MG, alternatives
with their advantages
among those and disadvantages,
proposed wereInanalysed.
in the literature. Also
[33] different in [34] and
architecture [16] a
models ofperformance
the MG,
with their
comparison advantages and
of Alternating disadvantages,
Current (AC), Direct were analysed.
Current Also
(DC), in [34]
hybrid and [16]
AC-DC a performance
micro-grids and other
comparison of Alternating Current (AC), Direct Current (DC), hybrid AC-DC micro-grids
configurations was realized. In particular, in [16] authors assert how DC distribution systems present and other
configurations
less power was realized. In particular, in [16] authors assert how DC distribution systems
quality problems.
present less power quality problems.
In this context, the architecture here analysed, depicted in Figure 1, can be divided into four
In this context, the architecture here analysed, depicted in Figure 1, can be divided into four
sections: “Photovoltaic plant”, “Battery storage system”, “Flywheel storage system” and “Residential
sections: “Photovoltaic plant”, “Battery storage system”, “Flywheel storage system” and “Residential
load”.load”.
EachEach
parallel branch
parallel branchis connected
is connectedtotoaacommon
common DC DC busbusatat400
400VV DCDC
rated
rated voltage.
voltage. All branches
All branches
interact
interact with an inverter which brings the voltage conditions to those typical of residential users.users.
with an inverter which brings the voltage conditions to those typical of residential
DC/DC andand
DC/DC AC/DC
AC/DC converters,
converters,with
with their efficiencyand
their efficiency andresponse
response delay,
delay, areare
notnot implemented
implemented in this
in this
dynamic model
dynamic yet.yet.
model

Figure 1. 1.Micro-grid
Figure Micro-grid architecture. PV:photovoltaic.
architecture. PV: photovoltaic.

As said in the Introduction, in this configuration, the battery is the primary energy storage
As said in the Introduction, in this configuration, the battery is the primary energy storage system
system and the flywheel helps in the load peak-shaving, providing or absorbing power peaks and
and the flywheel helps in the load peak-shaving, providing or absorbing power peaks and reducing
reducing fluctuations of power from/to battery and grid.
fluctuations of power
Regarding thefrom/to battery
MG sizing, in a and grid. study [35], according to a multi-criteria sensitivity
previous
Regarding the MG sizing, in a previous
analysis on yearly base, PV plant, battery and study [35], according
flywheel to a multi-criteria
sizes have been sensitivity
fixed. Specifically, analysis
a PV plant
on yearly
with base,
a peakPV plant,
power ofbattery
11 kW, aand flywheel
battery of 10 sizes
kWh have beeniron
(a lithium fixed. Specifically,
phosphate pack aisPV plant with
modelled in thea peak
power of 11 kW,
present study)a battery of 10 kWh
and a mechanical (a lithium
flywheel of 11iron
kW phosphate
of maximumpack power is modelled
and 2 kWhin of the present
capacity were study)
and aidentified.
mechanical flywheel of 11 kW of maximum power and 2 kWh of capacity were identified.
The The
chosenchosen
PV PV plant
plant size,asasthe
size, theothers
others analyzed
analyzed inin[35]
[35]upuptoto2020kW,
kW,is compliant
is compliant withwith
the the
standard size of Italian plants. Indeed, at the end of 2016, about 91% of PV installed plants in
standard size of Italian plants. Indeed, at the end of 2016, about 91% of PV installed plants results results
the power range up to 20 kW [36]. Consequently, in consideration of the PV sizing a complex
in the power range up to 20 kW [36]. Consequently, in consideration of the PV sizing a complex
residential load was selected.
residential load was selected.
3. Load Profile Analysis and Micro-Grid Dynamic Modelling
3. Load Profile Analysis and Micro-Grid Dynamic Modelling
As mentioned in the Introduction, the Simulink dynamic model has allowed to simulate and,
As mentionedanalysed
subsequently, in the Introduction,
the MG operationthe Simulink
at various dynamic modelInhas
working points. allowedthanks
particular, to simulate
to the and,
subsequently, analysed thethe
dataset implemented, MGdynamic
operation at various
model workingthe
can reproduce points. In particular,
photovoltaic thanks to the
plant production fordataset
a
implemented,
chosen day,the dynamic
taking model of
into account can reproduce
seasonal the photovoltaic
and meteorological plant
effects. At production
the same time,forthe
a chosen
model day,
permits
taking to evaluate
into account the operative
of seasonal andconditions of the hybrid
meteorological storage
effects. At thesystem,
sameintime,
termthe
of battery
modelstate of to
permits
charge (SOC) and flywheel angular velocity. In the load management strategy, the above parameters
evaluate the operative conditions of the hybrid storage system, in term of battery state of charge (SOC)
(PV produced
and flywheel angularpower, batteryInand
velocity. theflywheel SOC) are used
load management in orderthe
strategy, to minimize the withdrawal
above parameters from
(PV produced
the electrical network. Figure 2 reports a global overview of the MG dynamic model.
Energies 2018, 11, 396 4 of 23

power, battery and flywheel SOC) are used in order to minimize the withdrawal from the electrical
network. Figure
Energies 2 xreports
2018, 11, a REVIEW
FOR PEER global overview of the MG dynamic model. 4 of 23

Energies 2018, 11, x FOR PEER REVIEW 4 of 23

Figure
Figure 2. Simulink
2. Simulink model
model overview.Parameters
overview. Parameters are
aredescribed
describedinin
Section 3.4.3.4.
Section

As mentioned in the description of MG layout, the model was essentially divided into four
As mentionedFigure
interconnected in the2. “Load”,
blocks:
Simulink modelofoverview.
description “Flywheel”,
Parameters
MG “Battery”
layout, the are described
andmodel“Control was inessentially
Section 3.4. divided into four
Logic” (including QB block).
interconnected
In the “Load” blocks: “Load”,
section, the “Flywheel”,
instantaneous power“Battery”
difference and “Control
between currentLogic”
PV (including
production QB block).
(referred
As mentioned in the description of MG layout, the model was essentially divided into four
In theinterconnected
“Load” section, the instantaneous
to a location in Central-Southern power
Italy) and load
blocks: “Load”, “Flywheel”,
difference
demand
“Battery” and
between
is calculated. current
“Control Logic”
PV
This time production
dependent
(including
(referred
QBsignal
block).is
to
a location
In the in
central toCentral-Southern
understanding
“Load” section, thethe Italy) and load
MG operative
instantaneous demand
functions.
power is between
When
difference calculated.
it assumes aThis
current PVtime
positive dependent
value,
production a surplus signal
(referred in is
centralenergy production occurs together with the possibility to storage it in the hybrid
to a location in Central-Southern Italy) and load demand is calculated. This time dependent signal is in
to understanding the MG operative functions. When it assumes a positive ESS, according
value, a to
surplus
energy the current
production
central energy
occurs
to understanding and power
together
the MGwithlimits.
the
operativeVice versa, a When
possibility
functions. negative
to storage value isa positive
it in the
it assumes indicative
hybridvalue,of aaccording
ESS, a lack in PV
surplus into the
currentproduction
energy
energy with
production
and powerrespect
occurs to the
together
limits. load
Vicewithdemand, which
theapossibility
versa, implies
negative to a
storage
value power demand
it in the hybrid
is indicative from the
ESS,in
of a lack hybrid
according ESS.
to
PV production
In the follow, the singular sections of the MGversa,
modelaare described.
with the current
respect energy
to the loadand power
demand, limits.
which Vice
implies a power negative
demand value is indicative
from the hybrid ofESS.
a lackIn in
thePVfollow,
production
the singular with respect to the load demand, which implies a power demand from the hybrid ESS.
3.1. Loadsections
Section of the MG model are described.
In the follow, the singular sections of the MG model are described.
The user load demand was based on a measurements based dataset presented in [37]. Experimental
3.1. Load Section
3.1.
dataLoad
wereSection
sampled continuously for two years by 20 homes, with intervals of 8 s. The implemented
The user load
residential loadtakes
demand was based
into account of on on a measurements
programmable andbased based dataset
non-programmable loads. presented in [37].
The user load demand was based a measurements dataset presented in [37]. Experimental
Experimental data
In this were
study the sampled continuously
load relative to ayears
complexfor two yearsuser by made
20 homes, with intervals of 8 s.
data were sampled continuously for two by 20residential
homes, with intervals up8of
of threeimplemented
s. The houses was
The implemented
considered. residential
Specifically, a load takes
summer and into
a account
winter of
(Figureprogrammable
3) daily profiles
residential load takes into account of programmable and non-programmable loads. and
werenon-programmable
extracted. loads.
In this
In this study the load relative to a complex residential user made up of three houses was was
study the load relative to a complex residential user made up of three houses
considered. Specifically,
considered. a summer
Specifically, a summer and
anda awinter
winter(Figure 3) daily
(Figure 3) dailyprofiles
profileswere
were extracted.
extracted.

Figure 3. Example of a complex residential user load demand. (a) Winter and (b) summer daily profile.

Figure 3. Example of a complex residential user load demand. (a) Winter and (b) summer daily profile.
Figure 3. Example of a complex residential user load demand. (a) Winter and (b) summer daily profile.
Energies 2018, 11, 396 5 of 23
Energies 2018, 11, x FOR PEER REVIEW 5 of 23

Concerning the
Concerning the renewable
renewableproduction,
production,the
thephotovoltaic
photovoltaicpower
power generation
generation profile
profile implemented
implemented in
in the
the model
model refers
refers to atolocation
a location in the
in the Central-Southern
Central-Southern Italy,
Italy, withwith south
south facing
facing panels
panels withwith
a tiltaof
tilt30of
◦.
30°. Figure
Figure 4 shows
4 shows the average
the average monthly
monthly production
production for 11forkW11PV
kWplant.
PV plant.

Figure
Figure 4.
4. Average daily production
Average daily production of
of aa 11
11 kW
kW peak
peak photovoltaic
photovoltaic plant.
plant.

As anticipated in the introduction, also the weather conditions for the evaluation of the PV
As anticipated in the introduction, also the weather conditions for the evaluation of the PV
production were taken into account. In particular four different meteorological conditions were
production were taken into account. In particular four different meteorological conditions were
considered through the implementation of the following mitigation coefficients [38]:
considered through the implementation of the following mitigation coefficients [38]:
•• 0%: no
0%: no production,
production, day
day with
with total
total cloud
cloud cover.
cover.
•• 30%: very cloudy with temporary clear
30%: very cloudy with temporary clear up.up.
• 50%: day with variable cloudiness.
• 50%: day with variable cloudiness.
• 100%: maximum production, clear sky.
• 100%: maximum production, clear sky.
3.2. Battery Model
3.2. Battery Model
In order to implement the most MG’s capabilities, an integrated dynamic model of the battery
In order to implement the most MG’s capabilities, an integrated dynamic model of the battery
pack was developed to analyse its transient operation. In this model it was also evaluated how much
pack was developed to analyse its transient operation. In this model it was also evaluated how much
power is possible to store or deliver to the MG.
power is possible to store or deliver to the MG.
Authors in a previous research works [39] have tuned the model for a Q (Ah) capacity battery
Authors in a previous research works [39] have tuned the model for a Q (Ah) capacity battery
with which the battery SOC (state of charge) was calculated taking into account battery datasheet
with which the battery SOC (state of charge) was calculated taking into account battery datasheet int
characteristics. Specifically, the open circuit voltage (Vocv), the change of internal resistance ( Rbat )
characteristics. Specifically, the open circuit voltage (Vocv ), the change of internal resistance (Rint
bat )
during charging
during charging (R
(Rch
ch)) or
or discharge
discharge (R (Rdis) were used to characterize the battery behaviour.
dis ) were used to characterize the battery behaviour.
Usually battery
Usually battery current
current (I
(Ibat
bat)) and
and voltage
voltage (V
(Vbat) can be characterized as:
bat ) can be characterized as:
2
V ocv− qVocv
2 − 4 R int
intP
I bat = Vocv − Vocv − 4Rbat
bat P (1)
int
Ibat = 2 Rint
bat
(1)
2Rbat
int
VVbat =V
bat = Vocv −R
ocv − bat I
Rint
bat Ibat
bat (2)
(2)

where PP is
where is the
the power
power required/delivered
required/delivered to the battery and:

= f=
VocvcVocvc 1 (SOC ) ) charge
charge 
( )
f 1 (SOC
Vocv==
Vocv  (3)
(3)
= f=
VocvdVocvd 2 ( SOC ) ) discharge
f 2 (SOC discharge 
Energies 2018, 11, x FOR PEER REVIEW 6 of 23

int  Rch = f 3 (SOC ) charge 

Rbat =  (4)
Energies 2018, 11, 396  Rdis = f 4 ( SOC ) discharge  6 of 23

The Equations (3) and (4) were implemented by using look-up tables based on experimental
( )
data. In [39] the Vocv trends of the
int
LFP (Lithium
Rch = f 3 (iron
SOCphosphate)
) charge battery together with its internal
Rbat =
resistance are shown. The determination (4)
RofdisSOC
= f 4was
(SOCcarried out as follows:
) discharge
η I bat tables based on experimental data.

The Equations (3) and (4) were implemented by using
SOC = SOCini − look-up (5)
Q
In [39] the Vocv trends of the LFP (Lithium iron phosphate) battery together with its internal resistance
are shown. The determination of SOC was carried out as follows:
and:
η Ibat
Z
SOC = SOCini − (5)
 V Q 
η ch = V − I R
ocv
charge 
 ocv bat ch 
and: η = Vocv
 (6)
 − −I bat
 ηchV=ocvVocv IbatR charge
η =η dis = discharge 
Rchdis

(6)
  η = VV − Ibat Rdis
ocvocv 
discharge 
dis Vocv

where
where SOC
SOCiniiniisisthe
theinitial
initialvaluevalueofofSOC SOCand andQQ(Ah) (Ah)represents
representsthethebattery
batterycapacity.
capacity. In Inthe
thepresent
present
study
studybattery
batterycapacity
capacitywas wasset setat at79.4
79.4Ah Ah(five
(fivemodules
modulesin inseries)
series)with
withaa nominal
nominalstorage
storagepowerpowerof of
about
about10
10kWh.
kWh.
The
TheSimulink
Simulinkbatterybatterymodel modelisisshown shownin inFigure
Figure5.5. Specifically,
Specifically,the
thecurves
curvesshown
shownin in[39]
[39]were
were
implemented
implemented in the V Vcharge
charge, , V
Vdischarge , R, charge
discharge Rchargeandand Rdischarge blocks.
Rdischarge In the
blocks. “Current
In the “Currentandandcapacity
capacitycontrol”
control”
section
section (Figure
(Figure 6) 6) the
the current
current demand demand IIbatt batt(out4
(out4ininFigure
Figure5)5) was
was compared
compared withwith the
the maximum
maximum
charge/discharge
charge/discharge(82/200 (82/200 A)A) current
current as aswell as the
well as the instantaneous
instantaneousSOC. Therefore
SOC. Thereforeif Ibatt
if exceeds the
Ibatt exceeds
power and/or
the power energy
and/or limits,
energy the exceeding
limits, the exceeding current was delivered/requested
current was delivered/requestedto thetogrid.
the grid.

Figure
Figure5.5.Battery
Batterymodel.
model.SOC:
SOC:State
StateofofCharge.
Charge.
Energies 2018, 11, 396 7 of 23
Energies 2018, 11, x FOR PEER REVIEW 7 of 23

Figure 6. Current and capacity control.

Figure 6. Current and capacity control.

3.3. Flywheel Model

3.3. Flywheel Model
According to [23], a flywheel is able to store energy in the form of kinetic energy, rotating
According to [23], a flywheel is able to store energy in the form of kinetic energy, rotating around
around its axis, and to quickly respond to energy peaks demand. Thanks to these features, in the MG
its axis, and to quickly respond to energy peaks demand. Thanks to these features, in the MG model
model here presented, flywheel works in peak-shaving mode extending the battery lifetime.
here presented, flywheel works in peak-shaving mode extending the battery lifetime.
Equations (7) and (8) show flywheel power P and shaft torque T, as reported in [25]:
Equations (7) and (8) show flywheel power P and shaft torque T, as reported in [25]:
= (7)
P = Tω (7)
= . + + (8)
T = J ω + Td + Tb (8)
where: J is the rotational inertia, Tb denotes the frictional torque due to bearings, is the
where: J is the drag
aerodynamic rotational is Tthe
inertia,
torque, b denotes
angularthe frictional torque due to bearings, Td is the aerodynamic
acceleration.
.
drag Mathematical
torque, ω is themodels
angularconcerning
acceleration.mechanical (friction) and aerodynamic losses useful for the
model Mathematical
development models
can beconcerning
found in mechanical
[40,41]. The(friction) and aerodynamic
power losses lossesthe
associated with useful for the
mechanical
model
bearingsdevelopment can beby
are quantifiable found in [40,41].Equation
the following The power(9):losses associated with the mechanical bearings
are quantifiable by the following Equation (9):
, = 1.05 × 10 (9)
where , indicates the power losses,Pb,loss =
denotes 10−4 M rotational speed and
1.05 × flywheel (9)
is the total
bearing frictional moment, as detailed in [41].
Pb,loss indicates
whereAerodynamic the power
losses can be losses, n denotes
calculated by theflywheel
followingrotational
Equationsspeed andM
(10)and is the total bearing
(11):
frictional moment, as detailed in [41].
Aerodynamic losses can be calculated by the=following Equations (10) and (11): (10)
= (11)
Pw = MC ω (10)
In Equation (10) aerodynamic power losses are linearly related to windage frictional moment
(MC) and to flywheel rotational speed (ω).MM C C=for ρπω 2 r4 L flywheel is expressed by Equation (11),
C fcylindrical (11)
known cylinder torque coefficient , fluid density ρ and the geometric features of the cylinder (i.e.,
In Equation (10) aerodynamic power losses are linearly related to windage frictional moment (MC )
radius r and length L). It is remarked that, in the investigated system, aerodynamic drag is negligible
and to flywheel rotational speed (ω). MC for cylindrical flywheel is expressed by Equation (11), known
since the system is kept under vacuum.
cylinder torque coefficient C f , fluid density ρ and the geometric features of the cylinder (i.e., radius r
The flywheel was modelled in the Simulink environment (Figure 7) implementing the following
and length L). It is remarked that, in the investigated system, aerodynamic drag is negligible since the
technical data:
system is kept under vacuum.
• Maxflywheel
The Power (kW): 11
was modelled in the Simulink environment (Figure 7) implementing the following
• Energy
technical (kWh): 2.1
data:
• ω (rad/s): max 1131, min 314
•• MaxRadiusPower
(m):(kW):
0.4 11
•• Energy
Length (kWh): 2.1
(m): 0.08
•• ω (rad/s):
Weight max
(kg): 1131, min 314
160.6
• Material: Steel
Energies 2018, 11, 396 8 of 23

• Radius (m): 0.4

• Length (m): 0.08
• Weight (kg): 160.6
• Material: Steel
Energies 2018, 11, x FOR PEER REVIEW 8 of 23
The model input is a power load signal (In1). If this signal is negative, it corresponds to a power
Theversa,
request, vice model to
input is a power
a power loadThe
supply. signal (In1).provides
model If this signal is negative,
as outputs theitexchanged
correspondspower
to a power
with the
request, vice versa, to a power supply. The model provides as outputs
grid and the battery, flywheel rotational speed and its angular acceleration. the exchanged power with the
grid and the battery, flywheel rotational speed and its angular acceleration.
On the basis of experimental data, the parameters of Equations (9)–(11) were evaluated and
On the basis of experimental data, the parameters of Equations (9)–(11) were evaluated and
implemented in “bearing losses” and “air drag” subsystems (in light blue and blue, respectively) to
implemented in “bearing losses” and “air drag” subsystems (in light blue and blue, respectively) to
determine mechanical
determine mechanicalandandaerodynamic
aerodynamiclosses.
losses.

Figure 7. Simulink flywheel model. PID: Proportional-Integral-Derivative.

Figure 7. Simulink flywheel model. PID: Proportional-Integral-Derivative.
3.4. Models Integration
3.4. Models Integration
The integration of model sections was carried out according to Figure 2. Specifically the model
The
core integration of Logic”
is the “Control modelblock
sections was
where thecarried out according
main output tosubsystems
of the other Figure 2. Specifically the
are elaborated model
and
used
core is theto determine
“Control the MG
Logic” power
block flows.
where theInmain
particular theof
output input
the parameters needed by
other subsystems arethe “Control and
elaborated
used Logic” subsystem
to determine theare:
MG power flows. In particular the input parameters needed by the “Control
Logic”• subsystem
“omega”: are:
flywheel instantaneous angular speed. It is an output of “Flywheel” block.
• “diff”: difference between the electric load required by the user and the photovoltaic production.
• “omega”: flywheel instantaneous angular speed. It is an output of “Flywheel” block.
These signal can represent a power lack or surplus. It is an output of “Load” block.
• “diff”:
• difference
“P_pv”: between
photovoltaic the electric production.
instantaneous load required by previous
As the the user one,
and itthe photovoltaic
is determined production.
in “Load”
Theseblock.
signal can represent a power lack or surplus. It is an output of “Load” block.
• • “SOC”:
“P_pv”: instantaneous
photovoltaic state of charge,production.
instantaneous determined in “Battery” subsystem.
As the previous one, it is determined in
• “qb” and
“Load” block. “qb2”: two parameters determined in “QB” block on the basis of the difference values
(the current one and the one relative to the previous calculation time step) between production
• “SOC”: instantaneous state of charge, determined in “Battery” subsystem.
and load. These parameters approximate the trend of “diff”, by excess or defect respectively
• “qb”(with
and “qb2”: two parameters determined in “QB” block on the basis of the difference values
reference to absolute values), according to step profiles characterized by a slow variation.
(the current one and
In the follow the
a brief one relative
discussion to the
on these previousiscalculation
parameters provided. time step) between production
and load. These parameters approximate the trend of “diff”, by excess or defect respectively
The above listed parameters are elaborated in the “Control Logic” subsystem to define the
(with reference to absolute values), according to step profiles characterized by a slow variation.
energy amounts processed separately by the battery and the flywheel (“batt” and “Envol” output
In the follow a brief discussion on these parameters is provided.
signals). These two values are directly sent, as input, to the “Battery” and “Flywheel” blocks.
The above listed parameters are elaborated in the “Control Logic” subsystem to define the energy
4. Micro-Grid Management
amounts processed separately by the battery and the flywheel (“batt” and “Envol” output signals).
These twoAs just anticipated
values are directlyabove, theinput,
sent, as MG management was and
to the “Battery” implemented in blocks.
“Flywheel” the Simulink model
through a Matlab code in the “Control Logic” block. The designed management strategy of the MG
(flowcharts are provided in Figures 8 and 9) starts by evaluating the instantaneous difference
between PR (production from renewable) and PL (power required by the electric load); subsequently,
the two main cases can be distinguished:
Energies 2018, 11, 396 9 of 23

4. Micro-Grid Management
As just anticipated above, the MG management was implemented in the Simulink model
through a Matlab code in the “Control Logic” block. The designed management strategy of the
MG (flowcharts are provided in Figures 8 and 9) starts by evaluating the instantaneous difference
between PR (production from renewable) and PL (power required by the electric load); subsequently,
the two main cases can be distinguished:

• CASE 1: Diff = PR (i) − PL (i) ≤ 0:

The renewable plant production is not sufficient to satisfy the electrical load and the flowchart of
Figure 8 is followed; additional conditions (mainly battery state of charge, flywheel rotational speed,
absence of PV production) lead to the identification of different sub-cases (1.1a, 1.1b, 1.2a, 1.2b, 1.2c
and 1.2d).
• CASE 2: Diff = PR (i) − PL (i) > 0:
There is a surplus of renewable production and the flowchart of Figure 9 is followed activating at
each time step one of the 2.1a, 2.1b, 2.2a and 2.2b sub-cases.
In the following a brief discussion of all operating cases is done.

- Case 1.1

The difference between production and load assumes a negative value (lack of production,
Diff < 0), so if the flywheel is capable to deliver energy (rotational speed greater than the minimum
threshold ω min ) the software enters in sub-case 1.1. Moreover, after evaluating the battery SOC, one
of the cases 1.1a (battery state of charge greater than the minimum limit, SOC > SOCL ) and 1.1b
(the battery is not capable to deliver energy) is chosen.
In case 1.1a the difference between the energy produced by the renewable plant and energy
required by the load is supplied partly by the battery, while the flywheel provides the amount to
dampen load peaks. In the case the storage system is not able to completely satisfy the load, energy
can be drawn also from the grid. In 1.1b the battery is discharged. Therefore, energy is partly provided
by the grid and the flywheel releases the energy amount to provide a peak-shaving function towards
the grid.

- Case 1.2

Also in this case, there is a lack of production, but the flywheel has already reached the minimum
rotational speed ω min . Depending on the energy renewable production and battery SOC, cases 1.2a
(PR > 0 and battery able to deliver, SOC > SOCL ), 1.2b (PR > 0 and discharged battery), 1.2c (PR = 0
and SOC > SOCL ) and 1.2d (PR = 0 and discharged battery) can be detected.
In case 1.2a the renewable plant delivers energy. The battery supplies an energy amount, necessary to
cover the electrical load increased of a share of energy relative to the oscillations absorbed by the flywheel.
Thus, the flywheel is charged and increases its rotational speed starting from ωmin . This operation mode
is aimed to guarantee a more constant profile of energy demand required to the battery.
In case 1.2b battery is discharged. Consequently, the grid supplies the energy amount to guarantee
the complete satisfaction of the electric load and to partly recharge the flywheel. So, a peak-shaving
function is provided towards the grid.
If RES production is zero the case 1.2c occurs where the electrical load request is covered by
using the battery and the grid. The flywheel, considering the energy losses, progressively reduces its
rotational speed under ω min up to its stopping.
Finally, case 1.2d is characterized by no RES production and battery SOC equal to the minimum
one. Therefore, the electrical load is satisfied by the grid, while the flywheel progressively reduces its
rotational speed below its minimum rotational speed in order to avoid energy withdrawal to cover
energy losses.
Energies 2018, 11, 396 10 of 23

- Case 2.1

In this case a PV production surplus occurs. The flywheel is not fully charged (ω < ωmax ) and,
in relation to the battery SOC, it is possible to distinguish:
Case 2.1a with battery not fully charged: the surplus energy is stored both in battery and flywheel.
Flywheel absorbs the fluctuations to provide an almost constant charging profile to the battery.
Case 2.1b with battery fully charged: flywheel absorbs energy fluctuations to deliver a constant
energy profile to the grid.
Energies 2018, 11, x FOR PEER REVIEW 10 of 23
- Case 2.2
Case 2.1b with battery fully charged: flywheel absorbs energy fluctuations to deliver a constant
energy profile to the grid.
In case 2.2, in the presence of a PV production surplus, the flywheel is saturated since it is already
-
at its maximum Case 2.2
rotational speed. In case the battery has a SOC < SOCL (2.2a) an amount of energy is
stored by the battery,
In case 2.2,since thepresence
in the flywheel
of areleases to thesurplus,
PV production batterythethe oscillations
flywheel reducing
is saturated since itits
is rotational
already at its maximum rotational speed. In case the battery has a SOC < SOC
speed. In this way the battery is operated under an almost constant charging profile. In case 2.2b, L (2.2a) an amount of

energy is stored by the battery, since the flywheel releases to the battery the oscillations reducing its
instead, because the battery is already fully charged, RES production is directed towards the grid.
rotational speed. In this way the battery is operated under an almost constant charging profile. In
Specifically, the flywheel releasesthethe
case 2.2b, instead, because oscillations
battery is alreadytowards the RES
fully charged, grid.production is directed towards
In all cases, energy fluxes processed by storage devices are the
the grid. Specifically, the flywheel releases the oscillations towards reduced
grid. if not compatible with their
In all cases,
energy and power capabilities. energy fluxes processed by storage devices are reduced if not compatible with their
energy and power capabilities.

Figure 8. Micro-grid management strategy: lack of RES (Renewable Energy Source) production.
Figure 8. Micro-grid management strategy: lack of RES (Renewable Energy Source) production.
Energies 2018, 11, 396 11 of 23
Energies 2018, 11, x FOR PEER REVIEW 11 of 23

Figure 9. Micro-grid management strategy: surplus of RES (Renewable Energy Source) production.
Figure 9. Micro-grid management strategy: surplus of RES (Renewable Energy Source) production.
5. Simulation Results
5. SimulationInResults
this section, the simulation results of the analyzed MG are presented. In particular, taking
into account of the variation of daily weather conditions and of the seasonal load demand, the
In thisbattery
section,
and the simulation
flywheel dynamic results
behaviorsofwere
the analyzed
determined.MG The are
timepresented.
step used forIn particular,
daily simulationtaking into
account ofwas theset
variation
equal to 1of s. daily weather conditions and of the seasonal load demand, the battery and
flywheel dynamic In order to improve
behaviors the clarity
were of this section,
determined. The only
timethe results
step usedrelating to thesimulation
for daily summer working was set equal
operation (in terms of load demand and irradiation) and three weather conditions (100%, 50% and
to 1 s.
30% of the nominal irradiation) are presented below. Aiming to appreciate the H-ESS daily
In order to improve
performance, thecondition
its initial clarity of wasthis section,
set as only the
“fully charged”. results
This relating
hypothesis to the
is arbitrary summer
and it is not working
operation (in terms of load demand and irradiation) and three weather conditions (100%,
representative of the actual conditions. The simulations, based on the actual load and the 50%
PV and 30%
production profiles of a generic day, aim to analyze the dynamic
of the nominal irradiation) are presented below. Aiming to appreciate the H-ESS daily performance, interactions among MG
components and with the grid (evaluating the beneficial effects of hybridization), without evaluating
its initial condition was set as “fully charged”. This hypothesis is arbitrary and it is not representative
the MG energy performance. To this last purpose, studies have already been conducted by the
of the actual conditions.
authors The basis,
[35] on a yearly simulations, based
leading to the sizingonofthe actual
the MG load and
components. thefollowing
In the PV production
figures, profiles
of a generic day, aimresults
simulations to analyze
are showntheindynamic interactions among MG components and with the grid
the order below:
(evaluating (a)the
thebeneficial effectsPV
difference between of production
hybridization),
and loadwithout
demand; evaluating the MG energy performance.
To this last(b)purpose,
flywheelstudies
angular have already
velocity (index ofbeen conducted
its state of charge);by the authors [35] on a yearly basis, leading
to the sizing(c) of
battery
the MGSOC; components. In the following figures, simulations results are shown in the
(d) flywheel instantaneous exchanged power respectively.
order below:

(a) the difference between PV production and load demand;

(b) flywheel angular velocity (index of its state of charge);
(c) battery SOC;
(d) flywheel instantaneous exchanged power respectively.

It is emphasized that seconds were chosen as time unit because the studied phenomena are
characterized by fast variations. However for clarity in each simulation 0 s stands for the day midnight,
43,200 s for midday and 86,400 for the next midnight.
 Energies 2018, 11, x FOR PEER REVIEW 12 of 23
Energies 2018, 11, 396 12 of 23
It is emphasized that seconds were chosen as time unit because the studied phenomena are
characterized by fast variations. However for clarity in each simulation 0 s stands for the day
midnight,
Figure 10 43,200
showss for
themidday and 86,400
simulation forfor
results theanext midnight.
summer load profile with a completely clear sky
Figure 10 shows the simulation results for
(maximum PV production). At the simulation beginning the a summer load profile
user loadwith a completely
demand clear
is totally sky by
covered
(maximum PV production). At the simulation beginning the user load demand is totally covered by
the H-ESS as the PV production is zero. Consequently, battery SOC and flywheel angular velocity fall
the H-ESS as the PV production is zero. Consequently, battery SOC and flywheel angular velocity
down respectively to about 89% and 200 rpm (the flywheel is left free to rotate under the minimum
fall down respectively to about 89% and 200 rpm (the flywheel is left free to rotate under the
angular velocity). With the start of PV production, the difference between production and energy
minimum angular velocity). With the start of PV production, the difference between production and
demand becomes
energy demand positive
becomesand the storage
positive and thesystems begins begins
storage systems to charge up to up
to charge 100%.
to 100%.

Figure 10. Responses to summer load profile with completely clear sky (no attenuation). (a) difference
Figure 10. Responses to summer load profile with completely clear sky (no attenuation). (a) difference
between PV production and load demand; (b) flywheel angular velocity; (c) battery SOC; (d) flywheel
between PV production and load demand; (b) flywheel angular velocity; (c) battery SOC; (d) flywheel
exchanged power.
exchanged power.

It is emphasized as in the case of extremely high demand peaks (at around 30,000 s of simulation,
Figure 10a), storage devices react in a coordinated manner (Figure 10b,c) supplying energy while
respecting the control logic, maximizing H-ESS lifetime and effectiveness.
Energies 2018, 11, 396 13 of 23
Energies 2018, 11, x FOR PEER REVIEW 13 of 23

It is emphasized as in the case of extremely high demand peaks (at around 30,000 s of
As for the initial part of the simulation, after the sunset, the load demand is covered by battery
simulation, Figure 10a), storage devices react in a coordinated manner (Figure 10b,c) supplying
and flywheel until the first reaches its minimum daily charge state (about 85%), while the second is
energy while respecting the control logic, maximizing H-ESS lifetime and effectiveness.
fully discharged.
As for the initial part of the simulation, after the sunset, the load demand is covered by battery
Itandis interesting
flywheel until tothe
make
first areaches
clarification regarding
its minimum the power
daily charge exchanged
state (about and provided
85%), while the secondby is the
flywheel. As it can
fully discharged. be seen in Figure 10d, the flywheel, even when it is fully charged (constant velocity
in Figure 10c) contrary totothe
It is interesting make battery, always needs
a clarification regardingof athesmall
power amount
exchangedof energy to balance
and provided out the
by the
bearing flywheel.
losses. As Foritthis
can reason,
be seen onlyin Figure
when10d, the flywheel,
a flywheel even discharged
its totally when it is fully and charged
it is left (constant
free to rotate
undervelocity
the minimumin Figureangular
10c) contrary to the
velocity, itsbattery,
power always
exchange needs of a small amount of energy to balance
is zero.
out the bearing losses. For this reason, only
Figures 11 and 12 show respectively the MG operating modes when a flywheel its totally
withdischarged and it is left
a RES production free to
characterized
rotate under the minimum angular velocity, its power exchange is zero.
by an irradiation equal to 50% and 30% of the nominal one. As it can be seen, lowering the photovoltaic
Figures 11 and 12 show respectively the MG operating modes with a RES production
production, a greater exploitation of the stored energy occurs. At the beginning of simulation, as
characterized by an irradiation equal to 50% and 30% of the nominal one. As it can be seen, lowering
in thethe nominal
photovoltaiccase,production,
H-ESS covers totally
a greater the users
exploitation of load demand.
the stored energy Then,
occurs.when
At the RES production
beginning of
begins, simulation, as in the nominal case, H-ESS covers totally the users load demand. Then, when (panel
the difference between photovoltaic power and load demand becomes positive RES a
in Figures 11 andbegins,
production 12). As the the irradiation
difference between decreases,
photovoltaicthispower
condition
and loadoccurs
demandmorebecomes
and more later in
positive
the time.
(panelDuring the day,
a in Figures 11 and the12).
difference surplus sent
As the irradiation to H-ESS
decreases, (maximum
this condition occurspower
more at and 40,000
more s of
later in the
simulation), time.ofDuring
in case the day, the is
30% irradiation, difference
not enough surplus sent to H-ESS
to increase (maximum
the charge levelpower
of theatflywheel
40,000 s and
of simulation), in case of 30% irradiation, is not enough to increase
battery up to the maximum SOC while it is sufficient for 50% irradiation. During daytime hours, the charge level of the flywheel
and battery
the minimum up to
state of the maximum
charge SOC while
of flywheel and itbattery
is sufficient for 50% irradiation.
corresponds to the high During
demand daytime
peaks hours,
at about
the minimum state of charge of flywheel and battery corresponds to the high demand peaks at about
30,000 s of simulation (i.e., respectively, 314 rad/s and 78% for 50% irradiation and 314 rad/s and
30,000 s of simulation (i.e., respectively, 314 rad/s and 78% for 50% irradiation and 314 rad/s and
about 52% for 30% irradiation). Indeed, as seen in Figure 12b,c), the battery and flywheel never reach
about 52% for 30% irradiation). Indeed, as seen in Figure 12b,c), the battery and flywheel never reach
their maximum
their maximum storage
storagecapacity
capacityduring
during the day,because
the day, because thethe
PVPV production
production is onlyisused
onlytousedcovertothe cover
the loadload required by residential utilities and it is not sent to the H-ESS. After the sunset, it can be noted be
required by residential utilities and it is not sent to the H-ESS. After the sunset, it can
notedthatthatH-ESS
H-ESSenergy energy falls
falls downdown quickly:
quickly: in case
in case of 30%
of 30% irradiation
irradiation batterybattery
state ofstate
chargeof reaches
charge reaches
the
the minimum
minimumallowed allowed SOC SOC (i.e., 10%) at
(i.e., 10%) atabout
about70,000
70,000s sofofsimulation
simulation and and flywheel,
flywheel, bothboth for 30%
for 30% and and
50% irradiation,
50% irradiation, is left
is left freefree to rotate
to rotate belowits
below itsminimum
minimum angularangularvelocity
velocity upupto to
thethe
endendof theof day
the day
(see the(see the flywheel
null null flywheel powerpower exchange
exchange Figures11d
Figures 11dand
and 12d).
12d).

Figure 11. Cont.

Energies 2018, 11, 396 14 of 23
Energies 2018, 11, x FOR PEER REVIEW 14 of 23
Energies 2018, 11, x FOR PEER REVIEW 14 of 23

Figure
Figure
Figure 11. Responses
11. Responses
11. Responses to summer
to summer
to summer load
load
load profilewith
profile
profile withpartially
with partially cloudy
partially cloudy sky
cloudysky (50%
sky(50%
(50% attenuation).
attenuation).
attenuation). (a) difference
(a) difference
(a) difference
between PV production and load demand; (b) flywheel angular velocity; (c) battery SOC; (d) flywheel
between
between PV production
PV production andand load
load demand;(b)
demand; (b)flywheel
flywheel angular
angularvelocity;
velocity;(c)(c)
battery SOC;
SOC;
battery (d) flywheel
(d) flywheel
exchanged
exchanged power.
power.
exchanged power.

Figure 12. Cont.

Energies 2018, 11, 396 15 of 23
Energies 2018, 11, x FOR PEER REVIEW 15 of 23

Figure 12. Responses to summer load profile with cloudy sky (30% attenuation). (a) difference between
Figure 12. Responses to summer load profile with cloudy sky (30% attenuation). (a) difference
PV production and load demand; (b) flywheel angular velocity; (c) battery SOC; (d) flywheel
between PV production and load demand; (b) flywheel angular velocity; (c) battery SOC; (d) flywheel
exchanged power.
exchanged power.
Relevant considerations can be drawn from Figure 12d. Specifically, it is substantially shown
the peak-shaving
Relevant considerationsfunction ofcanthebe
flywheel
drawntowards the battery,
from Figure 12d.inSpecifically,
order to extendit isitssubstantially
lifetime, and the shown
the peak-shaving function of the flywheel towards the battery, in order to extend itson
network. In fact, in Figures 11c and 12c it is clear how the peak-shaving function acts the charge
lifetime, and the
and discharge battery process, avoiding harmful peaks entering or leaving the battery. The latter
network. In fact, in Figures 11c and 12c it is clear how the peak-shaving function acts on the charge
consideration becomes particularly understandable verifying how the MG reacts in case of absence
and discharge battery process, avoiding harmful peaks entering or leaving the battery. The latter
or non-operation of the flywheel.
consideration becomes
To this aim, Figureparticularly
13 showsunderstandable verifying
the simulation results, how the MG
as exemplified forreacts in case of absence
30% irradiation, in the or
non-operation of the flywheel.
case of flywheel absence. It is possible to observe, with respect to Figure 12, that the battery
To this aim, Figure
discharge profile is more 13 shows the simulation results, as exemplified for 30% irradiation, in the
fluctuating.
case of flywheel absence.
This is even clearerIt by
is possible
comparing to the
observe,
profiles,with
withrespect to Figure
and without 12, that
flywheel, the current
of the batterysignal
discharge
entering
profile is more into the battery pack. In Figure 14, it is easy to notice that in case of flywheel absence, the
fluctuating.
battery
This iscurrent is characterized
even clearer by a profilethe
by comparing with higher oscillation,
profiles, with andlacking
without the peak shavingoffunction.
flywheel, the current
signal entering into the battery pack. In Figure 14, it is easy to notice that in case of both
To quantify the oscillation reduction provided by the flywheel introduction towards flywheel
battery and grid, a comparison on fluctuation indexes between the operation with and without
absence, the battery current is characterized by a profile with higher oscillation, lacking the peak
flywheel is done.
shaving function.
Firstly, the battery current profile was studied considering as example a 30% irradiation day.
To quantify
Figures 15 andthe16 oscillation
represent reduction provided by
the 1-min differential the of
form flywheel introduction
fluctuation towards both
and the accumulated rampbattery
and grid, a comparison on fluctuation indexes between the operation with and
probability function respectively. Specifically to improve the results readability, Figure 16 provides a without flywheel
is done.
zoom of the 1-min ramp probability up to 15 A, anyway covering the 96% of samples. With reference
Firstly, the battery
to this figure, current
it is noted profile
that for 90% ofwasthe studied considering
day the oscillation, as example
evaluated with aa1 30% irradiation
min time step, is day.
below
Figures 15 3and
amps 16while this value
represent the rises
1-minto 7differential
A in the absence
formofofthe flywheel. and the accumulated ramp
fluctuation
probability function respectively. Specifically to improve the results readability, Figure 16 provides a
zoom of the 1-min ramp probability up to 15 A, anyway covering the 96% of samples. With reference to
this figure, it is noted that for 90% of the day the oscillation, evaluated with a 1 min time step, is below
3 amps while this value rises to 7 A in the absence of the flywheel.
 Energies 2018, 11, 396 16 of 23
Energies 2018,
Energies 11,11,
2018, x FOR PEER
x FOR REVIEW
PEER REVIEW 16 16
of of
23 23

Figure 13.13.
Figure Battery
13. response
Battery response to to
summer
summer load profile
load (a)(a)
profile with
(a) cloudy
with
with sky
cloudy
cloudy (30%
sky
sky attenuation)
(30%
(30% without
attenuation)
attenuation) without
without
flywheel inin
flywheel
flywheel term
in ofof
term
term state
of ofof
state
state charge
of charge
charge(b).
(b).
(b).

Figure 14.14.
Figure Battery current
Battery demand
current with
demand and
with without
and flywheel.
without flywheel.
flywheel.
Energies 2018, 11, 396 17 of 23
Energies 2018,
Energies 2018, 11,
11, xx FOR
FOR PEER
PEER REVIEW
REVIEW 17 of
17 of 23
23

Figure 15.
Figure 15. 1-min
15. 1-min difference
1-min difference comparison
comparison between
comparison between battery
between battery current
battery current of
current of aa MG
MG (Micro
(Micro Grid)
Grid) with
with and
and
and
withoutflywheel.
without
without flywheel.
flywheel.

Figure 16.
Figure 16. Fluctuation
Fluctuation statistics
statistics comparison
comparison in
in 1-min
1-min between
between battery
battery current
current of
of aa MG
MG (Micro
(Micro Grid)
Grid)
Figure 16. Fluctuation statistics comparison in 1-min between battery current of a MG (Micro Grid)
with and without flywheel.
with and without flywheel.
with and without flywheel.
For completeness,
For completeness, Figures
Figures 17
17 andand 18
18 provide
provide the the ramp
ramp difference
difference and
and the
the accumulated
accumulated rampramp
For completeness,
probability for a 10 s Figures
time step. 17
In and 18 provide
confirmation of the
what ramp difference
already said, andinthe
also thisaccumulated
case, it’s ramp
clear the
probability for a 10 s time step. In confirmation of what already said, also in this case, it’s clear the
probability
advantage for ato10
due s flywheel
the time step.introduction
In confirmationin a of what
MG. In already
fact, in said,17a
Figure also in thisreduction
sensible case, it’s of
clear
the
advantage due to the flywheel introduction in a MG. In fact, in Figure 17a sensible reduction of the
the advantage
current can bebedue to the flywheel
identified. Moreover, introduction
Figure in
in Figure 18aititMG. In fact,
is noted
noted in Figure
that for 96%
96% 17a
ofsensible
day,reduction
the day, the currentof
current
current can identified. Moreover, in 18 is that for of the the
the current can
oscillation is be
is less identified.
less than
than 0.01
0.01 A Moreover,
A but
but it in
it increases Figure
increases to to 0.8 18
0.8 A it
A withoutis noted that
without aa H-ESS.
H-ESS. for 96% of the day, the current
oscillation
oscillation is less
Relevant than 0.01
benefits A but
were it increases
detected also to reference
in 0.8 A without to a H-ESS.
the power exchanged
exchanged with with grid.
grid. Power
Power
Relevant benefits were detected also in reference to the power
variation was
variation was evaluated
evaluated forfor both
both 11 min
min andand 10 10 ss time
time steps,
steps, on
on the
the basis
basis of
of the
the power
power profile
profile
Energies 2018, 11, 396 18 of 23

Relevant
Energies 2018, 11, xbenefits
FOR PEERwere detected also in reference to the power exchanged with grid. Power
REVIEW 18 of 23
Energies 2018,
variation was11,evaluated
x FOR PEERfor
REVIEW 18 of 23
both 1 min and 10 s time steps, on the basis of the power profile simulated
simulated for a 50% irradiation.
for a 50% irradiation. This is
This condition condition
needed to is needed to havecharacterized
have a profile a profile characterized
by power sent by power
to the
simulated for a 50% irradiation. This condition is needed to have a profile characterized by power
sent
grid to the grid comparable
comparable to power
to power from from
the grid. the grid. Figures
Specifically, Specifically,
19 andFigures
20 show 19 the
andramp
20 show the ramp
difference and
sent to the grid comparable to power from the grid. Specifically, Figures 19 and 20 show the ramp
difference and theramp
the accumulated accumulated ramp
probability probability
of power of power
variation for a 1variation
min timefor a 1 It
step. min time as
is clear step. It isH-ESS
in the clear
difference and the accumulated ramp probability of power variation for a 1 min time step. It is clear
as in the
case thepower
H-ESSfluctuations
case the power fluctuations
are reduced leadingareto,reduced leading to,
in the cumulated in the cumulated
distribution, distribution,
90% of samples with
as in the H-ESS case the power fluctuations are reduced leading to, in the cumulated distribution,
90% of samples
amplitude with amplitude
at maximum of 200 W atinstead
maximum of 200
of 1000 WW insteadflywheel).
(without of 1000 W (without flywheel).
90% of samples with amplitude at maximum of 200 W instead of 1000 W (without flywheel).

Figure
Figure17.
17.10-s
10-sdifference
differencecomparison
comparisonbetween
betweenbattery
batterycurrent
currentofofaaaMG
MGwith
withand
andwithout
withoutflywheel.
flywheel.
Figure 17. 10-s difference comparison between battery current of MG with and without flywheel.

Figure 18. Fluctuation statistics comparison in 10-s between battery current of a MG with and
Figure 18. Fluctuation statistics comparison in 10-s between battery current of a MG with and
without flywheel.
Figure 18. Fluctuation statistics comparison in 10-s between battery current of a MG with and
without flywheel.
without flywheel.
Figures 21 and 22 present the same analysis with 10 s time step. Referring to the accumulated
Figures 21 and 22 present the same analysis with 10 s time step. Referring to the accumulated
ramp probability plot, the hybridization allows to reduce, at 95% of samples, the power amplitude
ramp probability plot, the hybridization allows to reduce, at 95% of samples, the power amplitude
from 4000 W to 1000 W.
from 4000 W to 1000 W.
Energies 2018, 11, 396 19 of 23

Figures 21 and 22 present the same analysis with 10 s time step. Referring to the accumulated
ramp
Energiesprobability plot,
2018, 11, x FOR PEERthe hybridization allows to reduce, at 95% of samples, the power amplitude
REVIEW 19 of 23
Energies 2018, 11, x FOR PEER REVIEW 19 of 23
from 4000 W to 1000 W.

Figure 19. 1-min difference comparison between grid power exchanges of a MG with and without
Figure 19. 1-min difference comparison between grid power exchanges of a MG with and without
Figure 19. 1-min difference comparison between grid power exchanges of a MG with and without flywheel.
flywheel.
flywheel.

Figure 20.
20. Fluctuation statisticscomparison
comparison in 1-min between grid power exchanges of a MG with
Figure
Figure 20. Fluctuation statistics
statistics comparisoninin1-min
1-minbetween
betweengrid power
grid exchanges
power of aof
exchanges MG withwith
a MG and
and without
without flywheel.
flywheel.
and without flywheel.

Concluding, as it can be seen from the profiles shown, flywheel works in a collaborative way
Concluding,
Concluding, asas itit can
canbebeseen
seenfrom
fromthe
the profiles
profiles shown,
shown, flywheel
flywheel works
works in ain a collaborative
collaborative way way
with
with battery, shaving the load fluctuations towards both battery and grid, consequently with
with battery,
battery, shaving
shaving the load thefluctuations
load fluctuations
towardstowards bothand
both battery battery
grid, and grid, consequently
consequently with
with benefits in
benefits in terms of battery durability and grid safety and stability.
benefits in terms of battery durability and grid safety
terms of battery durability and grid safety and stability. and stability.
Energies 2018, 11, 396 20 of 23
Energies 2018, 11, x FOR PEER REVIEW 20 of 23
Energies 2018, 11, x FOR PEER REVIEW 20 of 23

Figure 21.
Figure 21. 10-s
21. 10-sdifference
10-s difference comparison
difference comparison between
comparison between grid
grid power
power exchanges of aa MG
exchanges of with and
MG with and without
without flywheel.
flywheel.
flywheel.

Figure 22.
Figure 22. Fluctuation
22. Fluctuation statistics
Fluctuation statistics comparison
statistics comparison in
comparison in 10-s
in 10-s between
10-s between grid
between grid power
grid power exchanges
power exchanges of
of aaa MG
exchanges of MG with
MG with and
with and
and
Figure
without
without flywheel.
flywheel.
without flywheel.
6.
6. Conclusions
Conclusions
6. Conclusions
This
This paper
paper provides
provides aa dynamic
dynamic analysis
analysis of
of aa hybrid
hybrid flywheel-battery
flywheel-battery energy
energy storage
storage system
system
This
coupled paper provides a dynamic analysis of a hybrid flywheel-battery energy storage system
coupled to a photovoltaic generation plant and a residential load. Thanks to the development of
to a photovoltaic generation plant and a residential load. Thanks to the development of aa
coupled
MG to a photovoltaic
dynamic model, the generation plant
instantaneous and a residential
behaviour of each load. Thanks
component and to
thethe development
mutual of a MG
interconnections
MG dynamic model, the instantaneous behaviour of each component and the mutual interconnections
dynamic
between model, the instantaneous behaviour of each component and the mutual interconnections
between the the singular
singular components
components are are characterized.
characterized. This
This analysis
analysis takes
takes also
also into
into account
account of
of the
the
between
effect of the singular
components components
sizing, power are
flow characterized.
control strategy This
and analysis
weather takes also
conditions. into
This account of the
hybridization
effect of components sizing, power flow control strategy and weather conditions. This hybridization
effect
has of components sizing, power flow control strategy and weather conditions. This hybridization
has allowed
allowed to to merge
merge the
the positive
positive features
features of
of base
base storage
storage technologies
technologies extending,
extending, in in this
this way,
way, their
their
has allowed
application to merge
ranges. the positive
Specifically, features
the of
flywheel base storage
peak-shavingtechnologies
function, extending,
typically in this
usable way,
in their
power
application ranges. Specifically, the flywheel peak-shaving function, typically usable in power
quality
quality applications,
applications, finds
finds in
in H-ESS
H-ESS an an implementation
implementation with with aa useful
useful effect
effect in
in energy
energy storage
storage
Energies 2018, 11, 396 21 of 23

application ranges. Specifically, the flywheel peak-shaving function, typically usable in power quality
applications, finds in H-ESS an implementation with a useful effect in energy storage applications.
Enhanced performances of the overall H-ESS, as briefly listed below, are provided with respect to the
case of battery storage systems:
- Enhanced duration. The issue related to the battery cycling is mitigated by the peak-shaving
function of the flywheel as demonstrated in terms of the oscillation reduction of the battery load profile.
In fact, with regard to the simulation conditions, it is quantified that this oscillation, valued with a
1 min time step, is maintained for 90% of the day below 3 A, while this value rises to 7 A in the case
of only installed battery. Evaluating with a reduced time step of 10 s, for 96% of the day, the battery
suffers extremely low current oscillation (0.01 A) that rises to 1 A in the absence of the flywheel. In [35]
the effect of oscillation reduction on battery lifetime is quantified by applying rainflow algorithm.
- Improvement of the quality of the power exchanged with grid. Specifically, considering 1 min and
10 s time steps, the power fluctuations are reduced respectively five times, during the 90% of the day,
and four times for 95% of daily samples.

Acknowledgments: The research activity has been carried out within the project “Sviluppo di una innovativa
Tecnologia integrata Volano-Batteria per l’accumulo efficiente di energia da rinnovabile per applicazioni di piccola
taglia”, funded by the Italian MISE (CCSEB_00201). Project partners are acknowledged.
Author Contributions: Linda Barelli, Andrea Ottaviano conceived and designed the simulations; Andrea Ottaviano,
Dario Pelosi implemented and performed the simulations; Linda Barelli, Andrea Ottaviano, Dario Pelosi wrote the
paper; Linda Barelli, Andrea Ottaviano analysed the data; Gianni Bidini reviewed the manuscript; Fabio Bonucci
designed Matlab code; Luca Castellini, Simone Castellini contributed to the architecture design; Alberto Zuccari
provided experimental data.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Cossent, R.; Gómez, T.; Olmos, L. Large-scale integration of renewable and distributed generation of
electricity in Spain: Current situation and future needs. Energy Policy 2011, 39, 8078–8087. [CrossRef]
2. Burgos-Payán, M.; Roldán-Fernández, J.M.; Trigo-García, Á.L.; Bermúdez-Ríos, J.M.; Riquelme-Santos, J.M.
Costs and benefits of the renewable production of electricity in Spain. Energy Policy 2013, 56, 259–270.
[CrossRef]
3. Pickard, W.F.; Shen, A.Q.; Hansing, N.J. Parking the power: Strategies and physical limitations for bulk
energy storage in supply-demand matching on a grid whose input power is provided by intermittent sources.
Renew. Sustain. Energy Rev. 2009, 13, 1934–1945. [CrossRef]
4. Moseley, P.T.; Garche, J. Electrochemical Energy Storage for Renewable Sources and Grid Balancing; Elsevier B.V.:
Amsterdam, The Netherlands, 2015; ISBN 9780444626103.
5. Barelli, L.; Desideri, U.; Ottaviano, A. Challenges in load balance due to renewable energy sources
penetration: The possible role of energy storage technologies relative to the Italian case. Energy 2015,
93. [CrossRef]
6. Keatley, P.; Shibli, A.; Hewitt, N.J. Estimating power plant start costs in cyclic operation. Appl. Energy 2013,
111, 550–557. [CrossRef]
7. Van Den Bergh, K.; Delarue, E. Cycling of conventional power plants: Technical limits and actual costs.
Energy Convers. Manag. 2015, 97, 70–77. [CrossRef]
8. Perez-Arriaga, I.J.; Batlle, C. Impacts of Intermittent Renewables on Electricity Generation System Operation.
Econ. Energy Environ. Policy 2012, 1, 3–18. [CrossRef]
9. Eser, P.; Singh, A.; Chokani, N.; Abhari, R.S. Effect of increased renewables generation on operation of
thermal power plants. Appl. Energy 2016, 164, 723–732. [CrossRef]
10. IEC T120. Available online: http://www.iec.ch/dyn/www/f?p=103:7:11494349307735::::FSP_ORG_ID,FSP_
LANG_ID:9463,25 (accessed on 31 January 2018).
11. Kaldellis, J.K.; Zafirakis, D. Optimum energy storage techniques for the improvement of renewable energy
sources-based electricity generation economic efficiency. Energy 2007, 32, 2295–2305. [CrossRef]
12. Evans, A.; Strezov, V.; Evans, T.J. Assessment of utility energy storage options for increased renewable energy
penetration. Renew. Sustain. Energy Rev. 2012, 16, 4141–4147. [CrossRef]
Energies 2018, 11, 396 22 of 23

13. Barelli, L.; Bidini, G.; Bonucci, F. A micro-grid operation analysis for cost-effective battery energy storage
and RES plants integration. Energy 2016, 113, 831–844. [CrossRef]
14. Luo, X.; Wang, J.; Dooner, M.; Clarke, J. Overview of current development in electrical energy storage
technologies and the application potential in power system operation. Appl. Energy 2015, 137, 511–536.
[CrossRef]
15. Mahmoud, M.S.; Azher Hussain, S.; Abido, M.A. Modeling and control of microgrid: An overview. J. Frankl. Inst.
2014, 351, 2822–2859. [CrossRef]
16. Mariam, L.; Basu, M.; Conlon, M.F. Microgrid: Architecture, policy and future trends. Renew. Sustain. Energy Rev.
2016, 64, 477–489. [CrossRef]
17. Conteh, M.A.; Nsofor, E.C. Composite flywheel material design for high-speed energy storage. J. Appl.
Res. Technol. 2016, 14, 184–190. [CrossRef]
18. Cheng, M.; Sami, S.S.; Wu, J. Benefits of using virtual energy storage system for power system frequency
response. Appl. Energy 2016, 194, 376–385. [CrossRef]
19. Aneke, M.; Wang, M. Energy storage technologies and real life applications—A state of the art review.
Appl. Energy 2016, 179, 350–377. [CrossRef]
20. Hamzaoui, I.; Bouchafaa, F.; Talha, A. Advanced control for wind energy conversion systems with flywheel
storage dedicated to improving the quality of energy. Int. J. Hydrogen Energy 2016, 41, 20832–20846.
[CrossRef]
21. Díaz-González, F.; Sumper, A.; Gomis-Bellmunt, O.; Bianchi, F.D. Energy management of flywheel-based
energy storage device for wind power smoothing. Appl. Energy 2013, 110, 207–219. [CrossRef]
22. Ren, G.; Liu, J.; Wan, J.; Guo, Y.; Yu, D. Overview of wind power intermittency: Impacts, measurements, and
mitigation solutions. Appl. Energy 2017, 204, 47–65. [CrossRef]
23. Mousavi, G.S.M.; Faraji, F.; Majazi, A.; Al-Haddad, K. A comprehensive review of Flywheel Energy Storage
System technology. Renew. Sustain. Energy Rev. 2017, 67, 477–490. [CrossRef]
24. Spiryagin, M.; Wolfs, P.; Szanto, F.; Sun, Y.Q.; Cole, C.; Nielsen, D. Application of flywheel energy storage for
heavy haul locomotives. Appl. Energy 2014, 157, 607–618. [CrossRef]
25. De Jong, H.; Williams, K. Development and Simulation of a Flywheel-Based Energy Storage System on a
Clamshell Dredge. pp. 1401–1414. Available online: https://www.westerndredging.org/phocadownload/
ConferencePresentations/2007_WODA_Florida/Session8C-DredgingProjectCaseStudies/2%20-%20de%
20Jong%20-%20Development%20%20Sim%20of%20Flywheel-Based%20Energy%20Strge%20System%
20on%20Clamshell%20Drdg.pdf (accessed on 5 January 2018).
26. Hannan, M.A.; Hoque, M.M.; Mohamed, A.; Ayob, A. Review of energy storage systems for electric vehicle
applications: Issues and challenges. Renew. Sustain. Energy Rev. 2017, 69, 771–789. [CrossRef]
27. Chen, H.; Cong, T.N.; Yang, W.; Tan, C.; Li, Y.; Ding, Y. Progress in electrical energy storage system: A critical
review. Prog. Nat. Sci. 2009, 19, 291–312. [CrossRef]
28. Nikolaidis, P.; Poullikkas, A. A comparative review of electrical energy storage systems for better
sustainability. J. Power Technol. 2017, 97, 220–245.
29. Boukettaya, G.; Krichen, L. A dynamic power management strategy of a grid connected hybrid generation
system using wind, photovoltaic and Flywheel Energy Storage System in residential applications. Energy
2014, 71, 148–159. [CrossRef]
30. Díaz-González, F.; Hau, M.; Sumper, A.; Gomis-Bellmunt, O. Coordinated operation of wind turbines and
flywheel storage for primary frequency control support. Int. J. Electr. Power Energy Syst. 2015, 68, 313–326.
[CrossRef]
31. Europe’s Largest Hybrid Flywheel Battery Project to Help Grid Respond to Energy Demand. Available online:
https://www.sheffield.ac.uk/news/nr/flywheel-europe-energy-1.704921 (accessed on 6 February 2018).
32. Prodromidis, G.N.; Coutelieris, F.A. Simulations of economical and technical feasibility of battery and
flywheel hybrid energy storage systems in autonomous projects. Renew. Energy 2012, 39, 149–153. [CrossRef]
33. Martin-Martínez, F.; Sánchez-Miralles, A.; Rivier, M. A literature review of Microgrids: A functional layer
based classification. Renew. Sustain. Energy Rev. 2016, 62, 1133–1153. [CrossRef]
34. Patrao, I.; Figueres, E.; Garcerá, G.; González-Medina, R. Microgrid architectures for low voltage distributed
generation. Renew. Sustain. Energy Rev. 2015, 43, 415–424. [CrossRef]
35. Barelli, L.; Bidini, G.; Bonucci, F.; Castellini, L.; Fratini, A.; Gallorini, F.; Zuccari, A. Flywheel hybridization to
improve battery life in energy storage systems coupled to RES plants. Energy. under review.
Energies 2018, 11, 396 23 of 23

36. GSE Rapporto Statistico 2016—Solare Fotovoltaico. Available online: https://www.gse.it/documenti_site/

DocumentiGSE/Rapportistatistici/SolareFotovoltaico-RapportoStatistico2016.PDF (accessed on 6 February 2018).
37. Murray, D.; Stankovic, L.; Stankovic, V. An electrical load measurements dataset of United Kingdom
households from a two-year longitudinal study. Sci. Data 2017, 4, 160122. [CrossRef] [PubMed]
38. La Radiazione Solare. Available online: http://www.itishertz.gov.it/minisiti/solare/radiazione.htm
(accessed on 6 February 2018).
39. Barelli, L.; Bidini, G.; Ottaviano, A. Optimization of a PEMFC/battery pack power system for a bus
application. Appl. Energy 2012, 97, 777–784. [CrossRef]
40. Liu, H.-P.; Werst, M.; Hahne, J.J.; Bogard, D. Prediction of Windage Losses of an Enclosed High Speed
Composite Rotor in Low Air Pressure Environments. Proc. ASME Summer Heat Transf. Conf. 2003, 1, 15–23.
41. SKF Rolling Bearings. Available online: http://www.skf.com/binary/77-121486/10000_3-EN-webb.pdf
(accessed on 6 February 2018).

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).