Applied Energy 137 (2015) 854866

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Multi-apartment residential microgrid with electrical and thermal

storage devices: Experimental analysis and simulation of energy
management strategies
Gabriele Comodi a,, Andrea Giantomassi b, Marco Severini b, Stefano Squartini b, Francesco Ferracuti b,
Alessandro Fonti a, Davide Nardi Cesarini c, Matteo Morodo c, Fabio Polonara a
a

Dipartimento di Ingegneria Industriale e Scienze Matematiche, Universit Politecnica delle Marche, Via Brecce Bianche, 60131 Ancona, Italy
Dipartimento di Ingegneria dellInformazione, Universit Politecnica delle Marche, Via Brecce Bianche, 60131 Ancona, Italy
c
Loccioni Group, via Fiume 16, 60030 Angeli di Rosora, Italy
b

h i g h l i g h t s
 Thermal & electric storage help to match consumption and renewable production profiles.
 Storage increase renewable energy self-consumption and independency from the grid.
 Storage reduce bidirectionality and intermittency of PV production.
 Computational framework for

l grid simulation under different operating conditions.

 The framework can evaluate diverse storage management solutions for future upgrades.

a r t i c l e

i n f o

Article history:
Received 5 March 2014
Received in revised form 18 June 2014
Accepted 18 July 2014
Available online 16 August 2014
Keywords:
Smart grid
Electrical and thermal storage
Distributed generation
Near-zero energy buildings
Solar power forecasting
Renewable energy

a b s t r a c t
The paper presents the operational results of a real life residential microgrid which includes six
apartments, a 20 kWp photovoltaic plant, a solar based thermal energy plant, a geothermal heat pump,
a thermal energy storage, in the form of a 1300 l water tank and two 5.8 kW h batteries supplying, each,
a couple of apartments. Thanks to the thermal energy storage, the solar based thermal energy plant is
able to satisfy the 100% of the hot water summer demand. Therefore the thermal energy storage represents a fundamental element in the management of the residential demand of thermal energy. It collects
renewable thermal energy during day-time to release it during night-time, effectively shaving the peak of
the thermal energy demand. The two electric storages, on the other hand, provide the hosted electrical
subsystems with the ability to effectively increase the self-consumption of the local energy production,
thus lowering the amount of energy surplus to be sold back to the grid, and increasing the self-sufficiency
of the microgrid. For instance, the storage has supported self-consumption up to the 58.1% of local energy
production with regard to the first battery, and up to the 63.5% with regard to the second one. Also, 3165
and 3365 yearly hours of fully autonomous activity have been recorded thanks to the first, and the second
battery respectively. On the other hand, the yearly average efficiency amounts to 63.7%, and 65.3%
respectively, for the first and second battery. In the second part of the paper we propose a computational
framework to evaluate the overall performance of the microgrid system, while accounting different operating conditions and energy management policies. From this perspective, the framework acts as a useful
modeling and design tool, to assess the opportunity of employing alternative energy management system
topologies and strategies. Eight different configurations, with growing complexity, have been derived
from the original system on purpose. The simulations, carried out based on real data related to one-year
time period, have provided results showing that, the higher the integration level of electrical and thermal
storage is, the higher degree of self-sufficiency can be achieved by the microgrid, and, in turn, the more
consistent the yearly energy saving become. Nevertheless, despite the energy cost reduction achievable

Corresponding author. Tel.: +39 071 220 4761; fax: +39 071 220 4770.
E-mail addresses: g.comodi@univpm.it (G. Comodi), a.giantomassi@univpm.it (A. Giantomassi), m.severini@univpm.it (M. Severini), s.squartini@univpm.it (S. Squartini),
f.ferracuti@univpm.it (F. Ferracuti), a.fonti@univpm.it (A. Fonti), d.nardicesarini@loccioni.com (D. Nardi Cesarini), m.morodo@loccioni.com (M. Morodo), f.polonara@univpm.it
(F. Polonara).
http://dx.doi.org/10.1016/j.apenergy.2014.07.068
0306-2619/ 2014 Elsevier Ltd. All rights reserved.

G. Comodi et al. / Applied Energy 137 (2015) 854866

855

with the availability of storage systems in the Leaf House, their high investment cost made them not
really profitable at the current price conditions for devices and energy purchase.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Power grids are going to face several challenges, such as the
increasing diffusion of distributed generation technologies [13],
many including renewable energy sources [47]. Other challenges
come from the integration and connection, at local scale, between
electric and thermal networks (but also electric mobility in the
near future) [812]. Moreover, with the adoption of demand side
management (DSM) strategies, final consumers are going to play
an active role in the grid activity [1319]. In this context, energy
storage technologies have a key role for several reasons. In the first
place, they represent the means to match energy production from
renewable sources and energy demand [20,21]. Even more so in
residential scenarios, since the load profile often describes the typical energy demand of an employee, who needs hot water mainly
early in the morning and late in the afternoon, with a minimum
contemporary factor with solar production. Thermal energy storage allows to collect renewable energy during day-time and to
use it during night-time. In addition, energy storage devices
enhance the energy self-consumption level achievable by final
users. Thermal energy storages, in particular, help in reducing
the burden on service utilities (natural gas or electricity). Electrical
energy storages double the benefit. On the one hand, it lowers the
burden on the power grid. On the other hand, it reduces, or prevents altogether, depending on circumstances, the bidirectional
flux of energy, from and towards the grid. It compensates for the
main obstacle against renewable energy source widespread, that
is the aleatory nature of renewable energy source availability
[2224]. The third reason, which is related to thermal storage, if
tailored on purpose, is the ability to increase the energy yield of
solar based thermal energy plants [25]. In conclusion, energy storages improve the flexibility of DSM strategies, enhance the final
user energy demand profile, and also minimize the overall energy
bill [19,26].
In the transition from nowadays power grid technology to the
smart grid technology, microgrids play a fundamental role as small
scale test bench of DSM strategies [2729].
In this paper we present a residential microgrid, the Leaf House,
which accounts six apartments, a photovoltaic (PV) energy production plant, a solar based thermal energy production plant, a geothermal heat pump, a thermal energy storage in the form of a
water tank of 1300 l and two batteries of 5.8 kW h each. The Leaf
House hosts a building automation and monitoring system which
makes it an ideal test field for energy storage systems applications.
By recording and collecting the data resulting from the everyday
life of its lodgers, the Leaf House is a living lab that records real life
energy demand profiles. Also the performance of both electrical
and thermal storages are evaluated on real life operating conditions, rather than in simulated ones.
A relevant contribution of this work is the computational
framework aimed at micro-grid design, which serves as a tool to
model and simulate the energy management occurring within
the Leaf House electrical system. It has been used to simulate the
environment behavior over a one-year time horizon, accounting
different storage management strategies and various system
configurations. The suitability of computational tools to monitor,
control and simulate the smart grid behavior in different operating
conditions and at different abstraction levels, has been extensively
shown and commented in literature [3033]. The proposed

framework is based on the Mixed-Integer Linear Programming

(MILP) paradigm, successfully used for energy management purposes by some of the authors in recent publications [34,35], but
not yet proposed as a design tool in the evaluation of case studies
based on a real life environment such as the Leaf House. The MILP
approach has shown its effectiveness and capability, in dealing
with a large number of constraints, with respect to other computational intelligence techniques [3638]. The framework includes,
also, a Neural Network based software for solar power forecasting.
The simulations carried out in this work have provided the means
to evaluate the yearly energy overall cost for each of the addressed
configurations. Therefore, the importance of the electrical and
thermal energy storages within the system has been evaluated,
but also the fact that the Leaf House energy management system
can be improved with the adoption of adequate hardware
modifications and storage management strategies.
The paper is organized as follows: Section 2 presents the
microgrid used as reference, along with operational results and
related comments. In Section 3 the energy management simulation
framework and the solar power forecasting algorithm are presented, whereas Section 4 presents the case study and the scenarios being examined. The simulation results are reported and
discussed in Section 5. Section 6 draws the conclusions of the work.
2. Microgrid under study: the Leaf House building
The Leaf House (see Fig. 1) is one of the six international case
studies selected by the IEA Task 40/ECBCS Annex 52: Towards
Net Zero Energy Solar Buildings [39,40]. Built in 2008, the Leaf
House is located in Angeli di Rosora, Ancona, Italy (latitude
43280 43.16 N, longitude 13040 03.65 E, altitude 130 m above sea
level). The site is characterized by a moderate climate: annual
temperature between 5 and 37 C; 1688 degree day, mean annual
horizontal solar radiation 302 W/m2. The Leaf House is a three
stories building, hosting three couples of twin flats. Two apartments are occasionally occupied, while the remaining four flats
are occupied by two lodgers each. The building is south oriented
and the ratio between the lengths of the south and east facades
was set to 1.34 to maximize solar gains during the colder season.
The roof, the solar thermal panels and the balcony have been
designed to behave as solar shadings during the hottest months.

Fig. 1. The Leaf House.

G. Comodi et al. / Applied Energy 137 (2015) 854866

856
Table 1
List of sensors.
Thermal energy
meter

Brunata HG  Q =S
Precision class: EN1434 class 2
Pt500 sensor temperature

Water flow meter

Turbine flow-meter
Nominal flow: 2.5 6 10 m3/h
Signal: 100 l/pulse

Temperature sensors Series: TR 10 Omnigrad M

Accuracy: 3% (range 200 to +600 C)
Precision: 0.2 K (range 200 to +650 C)
Electric meters

Beckhoff module KL3403

Mesured values: current (I); voltage (V); effective
power (P);
power factor cos /; I, V and P peak values; frequency
Accuracy: 0.5% full scale; (V/I), 1 % of calculated value

Data Acquisition
System

PLC Beckhoff (CX1100)

Operating/storage temperature: 0 to +55 C/25 to
+85 C
Relative humidity: 95%, no condensation

According to [28], a microgrid is essentially an active power

grid, which includes distributed generation, storage systems and
multiple loads at main grid voltage level. From the functional point
of view, the sources within the microgrid are provided with power
electronic interfaces (PEIs) and controls, to coordinate the activity
of each element of the system, in order to guarantee the required
power quality and energy output. The microgrid presents itself to
the main utility power system as a single controlled unit, and
meets local energy needs for reliability and security.
Table 1 lists the main sensors used in the monitoring activity
and their characteristics.
2.1. Electrical energy subsystem
The outline of the residential microgrid is presented in Fig. 2. A
20 kW peak power photovoltaic plant is integrated into the
building and produces the electrical energy needed to supply the
building. The plant is made of 150 m2 of mono crystalline modules,

it is south oriented and it is installed on the 21 sloped roof. It

provides energy not only to the HVAC system but also to the apartments lighting and loads.
Two energy storage systems have been installed in July 2012.
The Leaf House PV power plant is made of three single-phase
circuits, connected in order to obtain a balanced three phase circuit. Also the six apartments are grouped in pairs, each pair connected to form a single phase circuit. The resulting three circuits
are then connected to form a three phase circuit. The two energy
storage systems serve respectively two couples of flats and they
store energy from two out of the three PV arrays. This choice
was made because just four of the six apartments are regularly
occupied.
The main elements of the system are:
 a 5.8 kW h Li-ion battery and its inverter based interface that
manages the charge and discharge policies;
 a 6 kWp PV array and the corresponding inverter;
 the energy manager that coordinates the storage, the production and the demand.
The microgrid is connected to the main grid through a point of
common coupling (PCC) (represented by the circuit breaker CB0 in
Fig. 2). Self-consumption is maximized by storing the energy when
the PV production exceeds the consumption while feeding it
directly to the building otherwise. To maximize the reliability of
the system, the control strategies have been defined following
the policies listed below:
 the batteries serve only the apartments; the central HVAC system is not supplied by the electrical energy storage system;
 during the discharge phase, the battery supplies about the 90%
of the demand, the remaining 10% is always supplied by the
grid, thus avoiding the event of energy being feed back to the
main grid, since it is not allowed by the national law on PV feeding tariffs;
 the discharge process is not allowed below the 30% of the nominal capacity of the device, the reason being that, below said
value, the discharge process becomes very fast and the readings

Fig. 2. Electrical scheme.

G. Comodi et al. / Applied Energy 137 (2015) 854866

857

Fig. 3. Thermal scheme.

Table 2
hot water energy demand and supply.

2010
2011

Hot water
demand
(kW h) (A)

Thermal solar
production
(kW h) (B)

Thermal energy
auxiliary source
(kW h) (C)

B=A
(%)

C=A
(%)

2986
3149

3039
2304

1950
1789

102
73

65
57

of the residual energy level become unreliable, and may lead to

a loss of control over the discharge process; as such, in the event
of the battery being completely discharged, the system may
shut down and may not restart automatically because of the
voltage mismatch between the inverter and the battery; in
addition, it should be reminded that deep discharges tend to
reduce the battery lifetime;
 if the battery residual energy level fall at 30%, and the PV production is not sufficient to recharge the storage, the battery
manager draws about 100 Watt from the main grid, to supply
the battery and prevent its residual energy level to fall below
30%; the amount of power drawn from the grid has been identified during the preliminary test phase and covers both the
power needed by the battery control system and the battery self
discharge;
 the battery supplies the load when the demand is greater than
the PV production. However, after the charge phase begins, the
battery does not supply the load until the residual energy level
has risen up to, at least, the 65% of its residual nominal capacity;
this policy is required to limit the frequency of charge and discharge cycles.

Radiant floors provide the apartments with the requested

amount of thermal energy during cold seasons, while removing
the heat in excess during hot ones. The association of the ground
source and radiating floors reduces the gap between the heat
pump condensing and evaporating temperatures, increasing the
heat pump efficiency. Seven flat plane solar thermal collectors

(a) Summer days.

Provided that the system under study was part of a scientific

project, the turn-key cost of the electrical storage was (in 2012)
2650 /kW h, with a yearly O&M cost of 500 for each battery.
2.2. Thermal energy subsystem
The building hosts a central geothermal heat pump (GHP)
supplying cooling, heating and domestic hot water. The electric
peak-power of the heat pump is 16.6 kW.

(b) Winter days.

Fig. 4. Hot water energy demand, thermal energy produced by the solar plant and
thermal energy provided by the auxiliary boiler.

858

G. Comodi et al. / Applied Energy 137 (2015) 854866

(2.6 m2 each), according to the season, support the production of

domestic hot water; the solar based thermal energy plant is coupled with a 1300 l water tank acting as a thermal energy storage. An auxiliary electric boiler, of about 15 kWt (thermal
kilowatt), also supports the thermal energy production, in the
occurrence of low solar based thermal energy production in winter season. The set-point of the domestic hot water temperature
is 45 C. The temperature of the water supplied to the apartments by the heating system is 30 C during winter season and
18 C during summer season. The amount of heating and cooling
energy, provided to the apartments, is regulated through digital
valves that control the water flow in the radiant floor coils.
The valves are driven by a PWM (Pulse Width Modulation) controller, that acts based on the internal air temperature in each
thermal zone. The outline of the Leaf House thermal circuit is
presented in Fig. 3.
The cost of the 1300 l storage tank was 2000 , installation
included; the yearly maintenance cost is 100 .
2.3. Present day energy management: operational results and
preliminary comments
The monitoring activity of the energy parameters of the Leaf
House started in 2010. In the following, we present a summary
of the main operational results, related to both thermal and electrical storage management. Table 2 reports the yearly aggregated
results related to hot water energy demand, and supply, relating
to 2010 and 2011.
The high performance of the solar based thermal energy plant
are due to the system being oversized in order to meet the
winter energy demand. In two years the auxiliary boiler has provided 3739 kW h of energy. During summer the solar thermal
system has collected more energy than required, therefore part
of said energy has remained unused. That, and the additional
thermal losses, explains the sum of the energy supplied, by both
the solar thermal plant and the auxiliary boiler, exceeding the
100% of yearly hot water demand. Fig. 4 shows the hourly trend
of the hot water energy demand, the hourly trend of the thermal
energy produced by the solar plant and the hourly trend of the
thermal energy provided by the auxiliary boiler. In particular,
Fig. 4(a) refers to a ten-days period of summer and Fig. 4(b)
refers to a ten-days period of winter. Fig. 4 highlights the benefits provided by the thermal energy storage:
 It allows the solar plant to satisfy almost the 100% of the hot
water summer demand: no energy has been provided by the
auxiliary boiler in summer; indeed, depending on the evaluated
year, the maximum number of consecutive days, during which
the auxiliary boiler operation has not been necessary, ranges
between 77 and 94.
 It smoothes the thermal peak-power of the energy supply
systems: the yearly maximum of the hourly average value of
the hot water peak demand of the cluster of apartments is
slightly less than 12 kWt; even so, Fig. 4(b) shows that the
hourly average power provided by the auxiliary boiler is about
1 kWt; the yearly maximum of the hourly average value of the
peak power provided by the auxiliary boiler has been 3.5 kWt.
This smoothing effect will most likely lead, during 2014, to
the replacement of the currently installed electric boiler, of
about 15 kWt, with a 2.73 kWe (electrical kilowatt) waterair
heat pump; hypothesizing a HP COP of 3, the thermal energy
production peak-power will drop down to 8.2 kWt (45%).
Assuming a 96% efficiency of the electrical boiler, its replacement with a HP, will furtherly reduce the electrical peak consumption, required by hot water production, from 15.6 kWe
down to 2.73 kWe (82.5%).

 It is a fundamental component for residential demand side

management; indeed, the load profile describes the typical
energy demand of an employee, who needs hot water mainly
early in the morning and late in the afternoon, with a minimum
contemporary factor with solar production. Thermal energy
storage thus allows to collect renewable energy during daytime and to use it during night-time.
Data pertaining the electrical energy management have been
collected every 15 min, starting from the installation of the electric
storage system. Table 3 summarizes one-year operational results,
from 1 November 2012 to 31 October 2013, achieved by the two
couples of apartments, each couple taking advantage of a
5.8 kW h lithium ion battery.
Data reported in Table 3 show the role played by the electrical
energy storage in the maximization of renewable energy selfconsumption, in the reduction of the amount of energy feed back
to the main grid and, therefore, in the increase of self-sufficiency
of the microgrid. Since the management policies of the two electrical energy storages are the same, the slight differences between
the results achieved by the two couples of apartments are mainly
due to the different load profiles. Thanks to the electrical energy
storage, the percentage of energy being self-consumed (column g
in Table 3) is 58.1% for the first couple of apartments and 63.5%
for the second one. The actual number of hours, during which full
self-sufficiency has been achieved, are 3163 and 3356, respectively. Without electrical energy storage, the figures of self-consumption would had been 30.4% and 34.6% respectively, that is
the direct consumption of the electricity produced by the PV plant
(column f). The availability of batteries allows to store about the
17.5% of the electricity produced by the PV plant (column h). It is
worth noting that, since the batteries have been managed to store
the PV electricity exceeding the apartments electricity demand, the
energy being stored (column b) would have been fed back into the
main grid. Actually, not all the energy transferred to the batteries
has been produced by the PV system. Given the previously
described electric storage management strategy, if the battery
residual energy level drop to 30% of its nominal capacity, and no
renewable energy is available, the battery manager constantly
draws about 100 W from the main grid, to prevent the residual
energy level of the battery from dropping lower. Because of this,
about 14% of electricity transferred to the batteries comes from
the main grid, thus reducing down to about 15% the increase in
exploitable PV electricity that can be achieved thanks to the electrical energy storage. In column i, of Table 3, it is reported the
yearly average efficiency of the two batteries: 63.7% for the one
coupled with the 12 apartments and 65.3% for the other one.
These efficiency figures describe a much less performing device
with respect to the manufacturer certification, which guarantees
a nominal level of efficiency up to 95%. The analysis of recorded
data has pointed out also, a significant efficiency drop, when the
residual energy level of the batteries rise from 97% to 100% of its
nominal capacity, during recharge. Tests are currently being
carried out to improve the overall performance of the electrical
storage system, by modifying the charging conditions: an increase
of the battery efficiency up to 8590% is expected.
Fig. 5 shows the trend of energy consumption, PV production,
energy being charged and discharged and the state of charge of
the storage for the battery coupled with apartments 12. As stated
above, data have been acquired every 15 min. In particular,
Fig. 5(a) shows the trend for a 8-days period, while Fig. 5(b) focuses
on a two-days period, characterized by both a sunny and a variable-weather day. The trend in Figs. 5(a) and (b) reflects the batteries management strategies previously described. The electrical
energy storage, as well as the thermal energy one, has proven its
effectiveness as DSM tool. Even more so in the particular case of

G. Comodi et al. / Applied Energy 137 (2015) 854866

859

Table 3
Operational results from 1 November 2012 to 31 October 2013.

Ap.12
Ap.34

a
Cons.
(kW h)

b
Energy charged
(kW h)

c
Energy dischar.
(kW h)

d
PV Prod.
(kW h)

e
self-cons. from
PV (kW h)

f
e/a (%)

g
(c + e)/a (%)

h
b/d (%)

i
c/b (%)

l
Autonomy
from grid (h)

m
Nighttime
charge (%)

2720
2844

1183
1257

753
821

6635
7179

827
984

30.4
34.6

58.1
63.5

17.8
17.5

63.7
65.3

3163
3356

14.4
13.9

the Leaf House, whose electric load profiles describe the typical
demand of an employee who needs the electricity the most when
PV production is low or missing. During daytime, PV production
supplies the base-load of the apartments, whereas the surplus of
renewable electricity is stored in the battery. When PV production
is low or missing, the battery supplies the appliances with electricity until the residual energy level drop to 30% of its nominal capacity. When this threshold is reached, the battery manager starts to
draw power from the main grid (100 W) until the PV production
begins anew. From this point onwards, the trend of the energy
being stored results from the trend of PV production minus the
amount of energy consumed by the two apartments.
3. Energy management simulation framework
The suggested framework is meant to simulate the behavior of
the energy management subsystem at a high level of abstraction.
The aim is to focus on the energy fluxes and their balance, so that
the energy management subsystem can be tailored and evaluated
early in time, during the design process. Moreover, as a design tool
that focuses on the energy flows, the devised framework does not
require to simulate the system from a physical point of view. In
other words there is no need to account the evolution, over time,
of low level parameters such as the voltages of electrical storage
devices, or the temperatures within thermal storage devices. Once

(a) 8-days overview.

the most appropriate configuration of the structure has been

selected, the proposed approach can be complemented with a
more accurate modeling of the system, if a deeper insight of its
performance is required.
3.1. The structure of the model and the MILP problem
The proposed model has been devised as a set of blocks and elements. The blocks model the substructures within the main energy
management subsystem. The elements model the single devices
that compose each block. Elements can be arranged together to
model a block, by defining the constraints that bind together the
variables of chosen elements. Depending on the constraints, different structures, or even different policies, can be evaluated through
simulation.
In the current work, a real life microgrid is used as a reference,
namely the Leaf House. In this environment it is possible to identify an electrical management subsystem, that can be partitioned
in three independent blocks. Each block involves two apartments,
out of the six within the building, and a photovoltaic plant. Additionally, two blocks also include an electrical storage device each.
Within the management of thermal resources, it is possible to
identify hot water production, heating and cooling. Although
heating and cooling are provided by the same heat pump, which
operates alternatively as a cooler or heater, in the current work,
to improve the robustness and flexibility of the framework, heating
and cooling are modeled as independent blocks. Hot water production is modeled as an additional independent block.
In total, therefore, six blocks have been modeled. By arranging
the elements within each block, and by organizing the blocks
within the management subsystem, the performance of 8 alternative configurations have been evaluated. For each configuration a
dedicated routine has been created, while maintaining the same
parametrization for the model components and the same energy
production and load demand profiles.
Although the resulting model can be easily extended to account
non linear constraints, in the current scenario only linear constraints have been adopted. Therefore the simulation problem
has been addressed as a MILP problem. On purpose, since the constraints take the form of a block matrix, Mathworks MatLab is
adopted as simulation environment. The solver, namely SCIP, has
been adopted to complement said environment. On purpose, the
Opti Toolbox 2.05 [41] has been used as interface between the solver and the MatLab environment. The framework is hosted on a
Laptop PC, based on the Intel Core i7 CPU series, with 8 GB of
RAM, and running on the Microsoft Windows 8 64-bit OS.
3.2. Energy management model

(b) 2-days focus.

Fig. 5. Trend of electric parameters monitored.

The model operates on a single time slot at a time. The objective

function to be minimized is the sum of the costs of energy purchased and energy surplus. The aim is to avoid either unnecessary
purchase and, at times, unnecessary sale of electricity, thus promoting storage, since the sale price is usually much lower than
the purchase price. The variable modeling the energy fluxes within
the environment can be addressed as following:

G. Comodi et al. / Applied Energy 137 (2015) 854866

860

 EPa : amount of energy purchased to supply the entire structure;

 ESa : amount of energy surplus resulting from the entire
structure;
 Ep : the energy purchase price;
 EPi : energy required by the i-th electrical block with i 1; 2; 3;
 Eg i : energy required by the i-th electrical storage (to guard
against over discharge) with i 1; 2;
 EPhw : electricity consumption of the water boiler;
 EPhe : electricity consumption of the heat pump during heating
phases;
 EPre : electricity consumption of the heat pump during cooling
phases;
 ESi : energy production surplus provided by the i-th electrical
block with i 1; 2; 3.
 EPth : electricity demand of the thermal blocks not covered by
the available electricity surplus;
 Capi t 1: residual capacity level (of the i-th electrical storage)
at the end of the t-th time slot;
 Capi t: residual capacity level (of the i-th electrical storage) at
the beginning of the t-th time slot;
 Chi t: charged energy amount (of the i-th electrical storage)
during the t-th time slot;
 Dii t: discharged energy amount (of the i-th electrical storage)
during the t-th time slot;
 Dei : demanded energy amount (of the i-th electrical block);
 PV i : energy production of the photovoltaic panel (of the i-th
electrical block);
 Cheff : charge efficiency of each storage device;
 Dieff : discharge efficiency of each storage device;
 Sdi : self discharge of each storage device;
 Chg : hourly energy amount, provided by the main grid, to
compensate the self discharge;
 bg i : binary guard that equals zero if Capi t is higher than 30%;
 Ephw : heat production of the water boiler;
 Chhw : heat collected by the water tank;
 Dihw : heat lost as the water leave the tank;
 Shw : heat provided by the thermal solar panel;
 Dehw : heat demand corresponding to the hot water being used;
 Caphw t 1: energy level of the water tank at the end of the tth time slot;
 Caphw t: energy level of the water tank at the beginning of the
t-th time slot;
 Ephe : heat production of the heat pump;
 Chhe : heat collected by the heater storage;
 Dihe : heat output of the heater storage;
 Dehe : heat demand of the environment;
 Caphe t 1: energy level of the heat storage at the end of the
t-th time slot;
 Caphe t: energy level of the heat storage at the beginning of the
t-th time slot;
 Epre : chill production of the heat pump;
 Chre : chill collected by the chill storage;
 Dire : chill output of the chill storage;
 Dere : chill demand of the environment;
 Capre t 1: energy level of the chill storage at the end of the
t-th time slot;
 Capre t: energy level of the chill storage at the beginning of the
t-th time slot.
Therefore the objective function can be written as:

Q EP a ESa  Ep

meaning that either energy purchase and the sale of energy surplus
are not desirable, and thus they are accounted as a cost. In other
words the objective function is used to model a policy rather than
the actual energy cost. Of course, depending on the policy of choice,

the objective function can be restricted not to account the energy

surplus as a cost.
Almost every configuration adopted in the current work is
composed of three electrical blocks and three thermal blocks. Since
the thermal blocks require electrical energy to provide the thermal
management two alternatives are possible.
On the one hand, if the model is configured not to supply the
electricity surplus to the thermal blocks, the energy amount to
be purchased can be computed as:

EPa EP 1 EP 2 EP 3 Eg 1 Eg 2 EPhw EP he EP re

which results in the energy surplus, computed as:

ESa ES1 ES2 ES3

being either stored or discarded.

On the other hand, if the model is configured to route the electricity surplus to the thermal blocks, Eqs. (2) and (3) can be revised
in the following manner:

EPa EP 1 EP 2 EP 3 Eg 1 Eg 2 EPth

EPth  ESa EPhw EP he EP re  ES1 ES2 ES3

In this case, the variable EP th has been introduced to account the

electricity demand of the thermal blocks that is not covered by
the available electricity surplus. Conversely, ESa represents the electricity surplus that still exceeds the needs of the whole structure.
Concerning the electrical blocks, the amount of electricity to be
purchased, and the electricity surplus, are bound to production and
demand according to the following:

EPi  ESi  Chi Dii Dei  PV i

meaning that purchase or sale, while accounting the energy managed by the storage, must match the difference among production
and demand.
The electrical storage devices are modeled, each, by computing
its residual capacity level as:

Capi t 1 Capi t Chi t  Cheff  Dii t  Dieff Sdi Chg

 Cheff  bg i t

thus accounting charge and discharge efficiency as well as self

discharge. Also, the model includes a charge level guard. If the battery residual energy level drops to 30%, a charger, that compensates
for self discharge, is enabled so that the battery depletion due to self
discharge is prevented.
A few constraints have also been devised to model the original
policy used to manage the storage devices:

Chi t

0;

if PV i t 6 Dei t

PV t  Dei t; if PV i t > Dei t

8 i
if Capi t < 65
>
< 0;

PV i t < Dei t
Dii t
>
t

PV
t;
if
0:9

De
i
i
:
Capi t P 65%

For instance, Eq. (8) prevents purchased electricity from being

stored, whereas Eq. (9) prevents storage device from being deeply
discharged. These constraints are intended to achieve a conservative management of the storage, and they are paired with an explicit avoidance of energy sale. Thus they are complemented with the
objective function presented in Eq. (1). Conversely, to adopt a more
aggressive or, which is the same, a less conservative policy, said
constrains are not applied. In this case, also, Eq. (1) does not
account the energy surplus amount.
Similarly to the electrical blocks, with regard to the hot water
management, the balance between consumption and production
has been modeled:

G. Comodi et al. / Applied Energy 137 (2015) 854866

Ephw  Chhw Dihw Dehw  Shw

10

thus accounting hot water production and demand, and the management of a hot water tank.
The energy capacity of the water tank can be computed as:

Caphw t 1 Caphw t Chhw t  Dihw t

11

assuming that heat loss over time from the storage is negligible.
In the case of the heating block and the refrigeration block, the
model is almost identical, the only exception being the lack of
energy production from renewable sources, and the use of the heat
pump in place of the water boiler. For instance, being the heat
balance in the form Ephe  Chhe Dihe Dehe , the integration of hot
water and heat production has been modeled as a linear combination of said balance with Eq. (10), in place of two separate equations.
3.3. Solar power forecasting
In this paper an on-line learning procedure is proposed, to
predict the PV output and the Solar Thermal power output. On purpose, Radial Basis Function Networks (RBFNs) have been used.
These networks have been widely used for nonlinear system identification [42] because they have the ability both to approximate
complex nonlinear mappings, directly from inputoutput data,
with a simple topological structure that avoids lengthy calculations, and to reveal how learning proceeds in an explicit manner
[43]. The proposed on-line learning algorithm is based on the
Extended Minimal Resource Allocating Network (EMRAN)
technique, that adds hidden neurons to the network, based on
the innovation of each new RBFN input pattern which arrives
sequentially. As stated in [42], to obtain a more parsimonious network topology, a pruning strategy is introduced. This strategy
detects and removes, as learning progresses, those hidden neurons
which provide little contribution to the network output. If an
observation has no novelty then, the existing parameters of the
network are adjusted by an Extended Kalman Filter (EKF). In this
paper the performance of the filter is improved by an on-line
adjustment of the noise statistics, obtained by a suitably defined
estimation algorithm; the proposed Adaptive Extended Kalman Filter (AEKF) is able to adaptively estimate the unknown statistical
parameters [44,45]. To minimize the computational effort, in
real-time implementation, a winner neuron strategy is incorporate in the learning algorithm [42,45]. In this work the proposed
RBFN-based prediction algorithm is used, because it has the
capability to adapt on-line as operating conditions vary (i.e. night
and day and season succeed). Also these Networks do not need a
training dataset and they can be used in different case studies
without a training stage. The EMRAN estimation algorithm
enhanced by the AEKF is called EMRANAEKF algorithm [44], and
is shown in Fig. 6. In this work the input data consists of a tapped
delay line of 10 samples of the past data, each sample corresponding to a one-hour time interval. The forecast is aimed to predict
external temperature, solar irradiation and photovoltaic power
data, so the input dimension is 10  3. The output data is a onehour ahead forecast of solar irradiation and photovoltaic power
production.

framework as a design tool, a real life energy management system

has been analyzed in order to identify its building blocks. Thereafter the identified blocks have been modeled and thus 8 different
configurations, with growing complexity, have been derived from
the original structure.
In the first configuration, only production and demand of
electricity, hot water, heating and cooling has been evaluated. In
this case, therefore, no storage device has been included. The setup is composed of three electrical blocks, a hot water block, a heating block and a cooling one, but none of them is able to store energy.
The second configuration differs from the first one, by the addition
of three thermal energy storage devices, which pertain hot water,
heating and cooling respectively. The third configuration, on the
other hand, differs from the first one by the addition of two electrical energy storage devices. Thus, two electrical blocks are able to
store energy, whereas the remaining electrical block, and the three
thermal ones do not provide any form of storage. The fourth configuration includes two electrical energy storage devices and three
thermal energy storage devices. The four configurations already
described have, as common traits, that the building blocks are independent from each other, meaning that the energy surplus from
each block is not routed to the others. Also, when the electrical storage is used, a management policy, based on the constraints presented in Eqs. (8) and (9), is used. As said, this policy is meant to
prevent the storage from being recharged from the main grid, and
from being discharged if the residual capacity is below 65%. Also,
the fourth configuration presented matches the actual set up of
the real environment used as a reference.

4. Case study and scenarios

As mentioned above, the suggested framework is intended as a
design aid, meant to evaluate the performance of different energy
management alternative systems, which are to be integrated in the
microgrid context of choice. It follows that, given the needs and
habits of the user, the designer can identify the best performing
solution early in time, without the need of an in depth analysis
of each candidate solution. To evaluate the performance of the

861

Fig. 6. Flow chart of the EMRANAEKF algorithm.

862

G. Comodi et al. / Applied Energy 137 (2015) 854866

To evaluate additional improvements, few incremental changes

have been applied. From configuration 5 onwards, for instance, the
constraints presented in Eqs. (8) and (9) have been removed. Thus,
configuration 5 differs from configuration 4 in the electrical storage management policy. In configuration 6, in addition, the management of hot water and heating has been integrated. Then, in
configuration 7, the integration of electrical and thermal management is included, as presented in Eq. (5). Notably, the three electrical blocks remain independent in each configuration. As such, the
surplus from an electrical block is never routed to another electrical block, therefore, even in configuration 7 only the heat pump
and the water boiler can take advantage of the electricity surplus.
Since in the configuration 7 each electrical block remain independent from the others, as a mean to evaluate a further integrated
design, the configuration 8 is also proposed. In the case of configuration 8 the electrical subsystem has been modeled as a single
phase circuit, meaning that the solar energy production is the
sum of the production of the three panels, the energy demand is
the demand of the 6 apartments, and the electrical storages, combined, support the whole demand and production of the building.
As such, in configuration 8, electricity demand and production
are managed at building level rather than at electrical block level.
The proposed configurations are briefly listed as follows.
Configurations under test
1. storage devices not used
2. thermal storage
3. electrical storage with a conservative electrical storage
management policy
4. thermal and electrical storage with a conservative electrical storage management policy
5. thermal and electrical storage
6. thermal and electrical storage with the integrated production of hot water and heat
7. thermal and electrical storage with the integration of thermal and electrical management
8. thermal and electrical storage with the integration of thermal and electrical management at building level
The first test is based on historical data. The input is composed
of the electrical and thermal demand, the latter involving hot
water, heating and cooling needs. Additionally, photovoltaic
energy production, thermal solar energy production and residual
stored energy are assigned to the system. For instance the simulation covers an entire year spanning from 1 November 2012 to 31
October 2013. For each hour a simulation step is carried out.
Test 1 algorithm
1. the simulation routine receives the input data for the t-th
slot and simulates the energy management;
2. the simulation routine computes the energy fluxes at the tth time slot, and the residual stored energy at the end of
the same time slot;
3. the energy fluxes are recorded along with their costs and
the residual stored energy;
4. the input set of the t 1-th time slot is composed and a
new simulation step is carried out.
Also, at the beginning of the first step, the thermal storage
devices are fully charged, the electrical storage devices are at 50%
of their nominal capacitance level.
After the simulation described, to evaluate if the suggested
framework is also able to plan the activity of the storage early in
time, an additional test has been carried out. In this case the histor-

ical data, as regards photovoltaic energy production and thermal

solar energy production, has been replaced with its forecast counterpart. For instance, since the forecast is hour ahead, the test
requires two simulations per hour.
According to what stated above, the same framework can be of
use, not only to evaluate the system performance ahead of time,
but also to operate an actual resource management. In the forecasting case study then, since an offline planning is carried out, it
is also possible to search for an optimal scheduling scheme.
Test 2 algorithm
1. at the t-th time slot, the energy management, based on
forecast data, is carried out;
2. the planned storage activity (charge and discharge), at the
t-th time slot, is then recorded, along with the expected
costs;
3. the recorded storage activity is assigned, as an input, along
with historical data, to execute the planned management;
4. the resulting energy cost, at the t-th time slot, is computed,
and recorded along with the residual stored energy;
5. the process is repeated for the next time slot.

Notably, the distinction of the planning phase and the execution

phase it is not simply limited to the way the framework is used. In
order to separate properly the two phases, even the simulator configuration shall differ from the manager set-up, so that planning
and execution may not overlap. In regard of this, it is possible to
conclude that a policy is an essential element of a decisional
process, and thus its role, within the management process, is clear.
Conversely, however, its role during the execution process is
dubious, since the execution of a plan shall be carried out without
question. If that is not the case, in fact, the execution of the plan,
and its management, overlap, leading to management conflicts.
For these reasons, in order to distinguish clearly the management
phase and the execution phase, the constraints presented in Eqs.
(8) and (9), are applied to the configurations 3 and 4 during the
management process, but they are removed during the execution
phase.
In both tests, concerning the details of electrical configuration,
three electrical blocks have been modeled, each of them featuring
a photovoltaic panel of about 6 kW of peak power, and two of them
including an electrical storage device of 5.8 kW h each. Concerning
the thermal configuration, a hot water block has been modeled
including a solar thermal panel, an electrical water boiler of about
15 kW and a water tank of 1300 l. The water tank has been modeled assuming a DT of about 34 C between the temperature of
the output water and the temperature of the input water, thus
50 kW h of maximum stored energy has been assumed for the
water tank. Relating to the heating and cooling blocks, a heat pump
of about 16.6 kW has been used. The water storage, of about 400 l,
has been modeled assuming a DT of about 10 C, leading to roughly
7 kW h of maximum stored energy. For simplicity sake, two separated storage devices and heat pumps have been assumed, one set
for heating and the other for cooling, thus separating the heating
management from the cooling management. As regards the electricity price, a two tiered tariff has been assumed, distinguishing
peak hours from 8 a.m. to 7 p.m., from off peak hours. In particular
the peak hours tariff amounts to 0.138 , whereas the off peak
hours tariff amounts to 0.129 .
5. Results and comments
In the first test, the energy management is simulated, based
on historical data, for each of the proposed configurations. The

G. Comodi et al. / Applied Energy 137 (2015) 854866

performance is summarized in Table 4, which reports, in each

column, the results of the corresponding configurations.
The reported amounts present, in order, the cost of the energy
purchased to supply the electrical blocks and the cost of the energy
purchased to supply the thermal blocks. Next, the overall energy
purchased amount, and surplus are reported. At last, the residual
energy stored within each storage device is presented.
The comparison of the results, within the synoptic table is quite
straightforward. In the 1st configuration, the storage devices are
not included, therefore the energy surplus is discarded immediately. As a consequence, the system is forced to purchase
additional energy in order to match the demand, either electrical
or thermal, each time the local production is lacking.
In the 2nd configuration, on the other hand, the availability of
the thermal storage grants the system the ability to preserve the
thermal energy surplus thus lowering the electrical needs of the
thermal system. In this case, however, since the solar collector only
support the hot water production, the savings are quite limited and
amount to 38 per year.
In the 3rd configuration, the storage devices support the activity
of two, out of the three electrical blocks. Due to the higher energy
demand and production, therefore, an enhanced performance is
possible, with respect to the 2nd configuration. In particular
130 can be saved with respect to the 1st configuration.
Also, since each block operates independently from the others,
when both thermal and electrical storage devices are used, the saving from both the hot water block and the electrical blocks concurs
to reduce the overall energy bill, as shown from the results
obtained by the 4th configuration.
The 5th configuration proves to be a little less performing than
the 4th one. In this case, in fact, the savings provided by the addition of the electrical storage amount to 125 , with respect to the
reference case, represented by the 1st configuration. This is likely
due to the fact that, without accounting the energy surplus in
the objective function (Eq. (1)), the manager is not forced to store
the energy surplus, thus achieving a sub optimal management of
the energy surplus. On the other hand however, the absence of
the constraints, described in Eqs. (8) and (9), also grants the energy
manager with two additional degree of freedom, thus improving
the responsiveness of the storage management. Additional tests,
not shown here for the sake of conciseness, reveal in fact that, by
accounting the energy surplus within the objective function, an
additional saving of about 20 can be achieved.
The 6th configuration shows that the integration of hot water
production and heating can greatly enhance the efficiency of the
system. In fact, while the solar collector may support the heat production, thus reducing the energy required to heat the building, the
heat pump can provide the extra heat required to support hot
water production in place of the water boiler. In results, roughly
488 can be saved with respect to the 5th configuration due to a
more efficient management of thermal resources. The overall
savings, therefore, rise up to about 650 if compared against the
performance of the 1st configuration.
In the same perspective, the integration of electrical and thermal management, with the ability to route the electricity surplus
to the heat pump and the water boiler, further improve the ability
of the system to exploit the renewable resources. For instance, the
7th configuration, achieves an additional reduction of the overall
bill, of about 245 . If compared against the results of the 1st
configuration, the purchased energy amount is practically halved,
whereas the energy surplus is lowered to roughly the 60% of its original amount.Similar results are achieved if the 8th configuration is
used. When the electrical energy is managed from the building
standpoint, a different allocation of the resources may be possible.
As such, with respect to the 7th configuration, a further reduction
of the energy cost due to the electrical block is achieved, saving

863

about 52 . At the same time however, since the energy surplus

routed to the heat pump is also lowered, the energy cost due to
the thermal blocks increase slightly, requiring an extra of about
9.05 . The overall saving, therefore, amounts to 43 . In this case,
then, the algorithm is able to promote the energy purchase during
the off peak hours, thus achieving a lower energy bill in spite of an
increased amount of purchased energy. This management strategy
entails an enhanced self-consumption of the renewable energy
produced by the PV system, as confirmed by the decrease of the
surplus energy amount.
While the results proposed in Table 4 are collected at the end of
a simulation that cover an entire year time span, partial results
have been collected during the entire simulation process. In Figs. 7
and 8 respectively, the energy cost due to the electrical blocks and
the energy cost due to the thermal blocks are proposed.
At a glance, although the plots depict the same situation presented within Table 4, they also reveal that the actual results
are consistent through the entire simulation. In addition, it is
possible to assess in which way the costs increase through the
year, and thus the ability of each configuration to save energy.
As such, in Fig. 7 it is possible to observe that the energy
demand of the electrical blocks remains roughly the same
through the year, and therefore the cost increase, over time, is
almost linear.
With respect to the energy demand of the thermal blocks in
Fig. 8, in the other hand, it is possible to observe four seasonal
changes, since the simulated time interval begins in November.
In the first quarter of the plot, which roughly correspond to the
winter season, the energy cost and demand increase with a sustained rate, due to the heat demand of the building. In the second
quarter of the plot, that is the spring season, nor heating nor
cooling are required, thus the cost remains almost constant. In
summer, again, the cost over time shows a fast rate of growth that
decreases in the last quarter. Also, the less efficient the configuration, the higher the cost rise over time.
In relation to the case of the scheduling approach, the robustness to data uncertainty has been also evaluated. In this case, in
fact, the storage activity is planned by means of the forecast data.
This means that the prediction error may affect the performance of
the system, in the different addressed configurations.
The performed tests, in particular, have shown two distinct
trends in connection with the management of electrical and thermal storage devices. In relation to the energy demand due to the
thermal blocks, in particular, only minor changes, with respect to
the previous test, can be highlighted. The electrical management
however has shown a significant difference.
The 1st configuration, since it lacks any form of storage, has
achieved the same results shown in the previous test. The 2nd
configuration, although to a smaller degree with respect to the
previous test, is still able to improve its performance with respect
to the 1st one. The 3rd configuration, conversely not only has not
provided the expected benefits, but also worsened the situation
with respect to the reference case, although to a small degree.
As a result, in the 4th configuration, by integrating both
electrical and thermal storage, the overall performance remain
almost unchanged with respect to the reference case. In the 5th
configuration, the adoption of a less conservative management policy, partially ease the problems originating from the prediction
error.
In the 6th configuration, the cost, due to the energy demand of
the thermal block, remains almost unchanged with respect to the
previous test, thus confirming that the thermal storage subsystem
is fairly robust against prediction errors. This result is likely connected to the fact that a negligible thermal loss has been assumed
to model the thermal storage, whereas a quite significant self
discharge has been applied to the electrical storage. As such, the

G. Comodi et al. / Applied Energy 137 (2015) 854866

864

Fig. 7. Energ y cost due to the electrical blocks. The identification number is based
on the entries listed in Section 4.

Fig. 8. Energy cost due to the electrical blocks. The identification number is based
on the entries listed in Section 4.

prediction uncertainty affects the activity time of the storage and

therefore influences the storage management strategy.
From this perspective, the results achieved by the 7th configuration appear quite clear. The thermal blocks, in fact, remain
almost unaffected by the prediction error, and no significant difference, with respect to the previous test, is present in relation to the
energy cost due to the thermal blocks. The energy cost due to the
electrical blocks, on the other hand, shows a further increase with
respect to the 6th configuration. Since the energy surplus is to be
routed towards the thermal blocks, the management of the electrical storage requires an increased accuracy, and thus the process
becomes less robust against prediction errors. On the other hand,
the 8th configuration seems to be the most robust among the
others under test. Indeed, differently from the 7th configuration,

the energy cost saving due to the thermal blocks is not paired with
an increased energy cost due to the electrical blocks. Rather,
although limited, even the energy cost due to the electrical blocks
is reduced. In the light of this, 8th configuration allows achieving
the best performance within the proposed set of tests.
In conclusion, from the first test set it is possible to observe that
the ability to store both the thermal and the electrical energy usually improves the performance of the energy management, even
more so if the inclusion of storage devices is complemented with
the integration of the different blocks within the system. In particular, the comparison of the 7th and 8th configuration against the
4th configuration, which is characterized by the same constraints
for storage activity actually adopted in the Leaf House, suggests
that there is room for new design choices within the energy management subsystem, and that these choices are likely to improve
the related efficiency. The second test set results highlight that
while the thermal energy management shows a fair amount of
robustness against the error prediction, the management of the
electrical storage can be improved based on this perspective.
Therefore, future efforts will be targeted on this direction. For
instance, the comparison of the 5th configuration against the 4th
one, suggests that an appropriate choice of the management policy,
in dependence on the expected liability of the forecaster on hourly
basis, may solve the issue. Nevertheless, the last two configurations, proposing an integrated management of the electrical and
thermal storage capabilities, still allow to achieve a significant
reduction of the annual energy cost with respect to other
approaches.

5.1. Economic considerations

Energy management strategies, investigated in previous sections, shall not be evaluated in terms of operational savings only.
Indeed, the investment cost often represents a significant barrier
against energy storage installation. In the previous section, simulations have demonstrated that the higher the electrical and thermal
storage integration level, the bigger the amount of microgrid selfconsumption and the annual energy saving. Nevertheless, this
result would be obtained at the expense of a high investment cost.
In Tables 4,5 we present both the capital and maintenance costs of
energy storage systems installed in the Leaf House, in order to
assess the energy savings obtained while accounting the increase
of investment cost required by each configuration proposed. The
comparison between investment costs and achievable savings
shows that the energy storages being proposed are not profitable
at present time. In the case of thermal storage, the high investment

Table 4
Energy management, based on historical data, for all addressed system configurations. The identification number is based on the entries listed in Section 4.
Configurations

Energy cost due to electrical blocks

Energy cost due to thermal blocks
Energy overall cost
Purchased energy amount
Surplus energy amount
Residual electrical energy
Residual thermal energy
Residual refrigerating energy
Residual hot water energy
Electrical storage invest cost
Thermal storage invest cost
Total invest cost
Electrical storage maintenance yearly fee
Thermal storage maintenance yearly fee
Storage maintenance total yearly fee

()
()
()
(kW h)
(kW h)
(W h)
(W h)
(W h)
(W h)
()
()
()
()
()
()

618.85
1356.69
1975.54
14,663
20,630
0
0
0
0
0
0
0
0
0
0

618.85
1318.76
1937.61
14,682
20,630
0
1750
1750
12,500
0
2000
2000
0
100
100

488.50
1356.69
1845,15
13,964
17,778
3334
0
0
0
30,740
0
30,740
1000
0
1000

488.50
1318.76
1807,26
13,681
17,778
3334
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

493.42
1318.76
1812.18
13,715
17,396
3396
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

493.42
830.59
1324.01
9874
14,936
3396
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

493.03
585.25
1078.28
7426
13,666
3396
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

441.08
594.30
1035.38
7926
11,181
6267
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

G. Comodi et al. / Applied Energy 137 (2015) 854866

865

Table 5
Energy management, based on forecast data, for all addressed system configurations. The identification number is based on the entries listed in Section 4.
Configurations

Energy cost due to electrical blocks

Energy cost due to thermal blocks
Energy overall cost
Purchased energy amount
Surplus energy amount
Residual electrical energy
Residual thermal energy
Residual refrigerating energy
Residual hot water energy
Electrical storage invest cost
Thermal storage invest cost
Total invest cost
Electrical storage maintenance yearly fee
Thermal storage maintenance yearly fee
Storage maintenance total yearly fee

()
()
()
(kW h)
(kW h)
(W h)
(W h)
(W h)
(W h)
()
()
()
()
()
()

618.85
1356.69
1975.54
14,946
20,630
0
0
0
0
0
0
0
0
0
0

618.85
1349.76
1968.61
14,895
20,630
0
1750
1750
12,500
0
2000
2000
0
100
100

623.99
1356.69
1980.68
14,979
18,791
10,455
0
0
0
30,740
0
30,740
1000
0
1000

623.99
1349.76
1973,75
14,928
18,791
10,455
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

620.38
1349.76
1970.14
14,891
18,382
10,455
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

620.38
830.59
1450.97
10,976
18,382
10,455
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

655.20
588.25
1243.45
9470
13,453
10,441
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

614.93
616.87
1231.8
9384
11,164
13,340
1750
1750
12,500
30,740
2000
32,740
1000
100
1100

cost results from the storage being over-sized. Indeed, since the
Leaf House has been designed accounting the Annex 52 of the
EBC Research Programme, annex by the title Towards Net Zero
Energy solar buildings, the thermal storage has been designed in
order to supply most of hot water demand with solar thermal
panels also during winter season. Nevertheless, configurations 6,
7 and 8 show that the use of the thermal storage, integrated in
the heating system management, would be a profitable solution,
with no additional cost with respect to the present investment cost
conditions. On the contrary, the involvement of electrical storage
does not seem to guarantee high return of investment rates in
the addressed simulated case studies, due to the high purchase
and maintenance costs for each device. On purpose, a deep analysis
of the economic sustainability of the storage systems installed in
the Leaf House, involving the computational framework as design
tool, is left to future investigations.

6. Conclusions
In the first part of the paper we presented the operational
results of a residential microgrid, the Leaf House, composed by
six apartments, equipped with a 20 kWp PV plant, a solar thermal
plant, a geothermal heat pump, a hot water thermal storage of
1300 l and two 5.8 kW h batteries, each serving a couple of
apartments. Thermal energy storage has been demonstrated to
be a fundamental component for residential demand side management. Indeed, it has allowed both to smooth thermal peak-power
of the energy supply systems and to collect renewable energy
during day-time to use it during night-time. The latter feature is
particularly useful since the hot water load profile of the occupants
had a minimum contemporary factor with solar production. Furthermore, the thermal storage has allowed to satisfy almost the
entire hot water summer demand through solar thermal energy
production. The two electrical storages have allowed to maximize
the self-consumption of renewable PV energy, thus reducing the
amount of electricity fed back to the main grid and, consequently,
increasing the self-sufficiency of the microgrid. Thanks to the electrical energy storage system the percentages of self-consumed
energy have been 58.1% (first battery) and 63.5% (second battery).
The resulting number of hours during which the microgrid has
been fully self-sufficiency have been 3163 and 3356, respectively.
In the second part of the paper a computational framework
aimed at performance evaluation of the energy system of interest,
has been presented, including suitable energy management strategies therein. The framework is indeed featured by the model of
diverse energy blocks composing the energy system of interest,

i.e. the Leaf House, the solar power forecasting algorithm and the
MILP based optimization technique for energy management. From
this perspective, the framework can play the role of a design tool to
assess the performance of alternative management solutions. On
purpose, 8 different system configurations, with growing complexity, have been derived from original structure and tested through
simulation.
We have carried out two kind of tests: the first one has been
based on historical data, whereas second one has been relying on
forecasts of photovoltaic energy production and solar based thermal energy production. Both kind of test sets have allowed to
gauge the gain provided by the storage devices within the Leaf
House energy system, and also to show to what extent an higher
electrical and thermal storage integration enables a deeper microgrid self-consumption, also allowing to achieve a significant annual
energy cost reduction. Despite the energy cost reduction achievable with the availability of storage systems in the Leaf House,
their high investment cost made them not really profitable at the
current price conditions for devices and energy purchase. In particular, while the thermal energy storage can be profitable when used
also for the heating system management, the batteries seem to be
still too high-priced to be competitive in the residential market.
Future works will be targeted to deeply analyze the economic
sustainability of the Leaf House in different operating scenarios.

References
[1] Bianchi M, Pascale AD, Melino F. Performance analysis of an integrated {CHP}
system with thermal and electric energy storage for residential application.
Appl Energy 2013;112(0):92838.
[2] Mongibello L, Capezzuto M, Graditi G. Technical and cost analyses of two
different heat storage systems for residential micro-CHP plants. Appl Therm
Eng 2013(0).
[3] Chen X, Wang Y, Yu H, Wu D, Li Y, Roskilly A. A domestic {CHP} system with
hybrid electrical energy storage. Energy Build 2012;55(0):3618. Cool roofs,
cool pavements, cool cities, and cool world.
[4] Nair N-KC, Garimella N. Battery energy storage systems: assessment for smallscale renewable energy integration. Energy Build 2010;42(11):212430.
[5] Hadjipaschalis I, Poullikkas A, Efthimiou V. Overview of current and future
energy storage technologies for electric power applications. Renew Sustain
Energy Rev 2009;13(6):151322.
[6] Blarke M, Lund H. The effectiveness of storage and relocation options in
renewable energy systems. Renew Energy 2008;33(7):1499507.
[7] Ma X, Wang Y, Qin J. Generic model of a community-based microgrid
integrating wind turbines, photovoltaics and {CHP} generations. Appl Energy
2013;112(0):147582.
[8] Nuytten T, Claessens B, Paredis K, Bael JV, Six D. Flexibility of a combined heat
and power system with thermal energy storage for district heating. Appl
Energy 2013;104(0):58391.
[9] Arteconi A, Hewitt N, Polonara F. Domestic demand-side management (DSM):
role of heat pumps and thermal energy storage (TES) systems. Appl Thermal
Eng 2013;51(12):15565.

866

G. Comodi et al. / Applied Energy 137 (2015) 854866

[10] Kempton W, Tomic J. Vehicle-to-grid power implementation: from stabilizing

the grid to supporting large-scale renewable energy. J Power Sources
2005;144(1):28094.
[11] Rezaie B, Reddy BV, Rosen MA. Energy analysis of thermal energy storages
with grid configurations. Appl Energy 2014;117(0):5461.
[12] Taylor PG, Bolton R, Stone D, Upham P. Developing pathways for energy
storage in the {UK} using a coevolutionary framework. Energy Policy 2013;
63(0):23043.
[13] Salah CB, Chaabene M, Ammar MB. Multi-criteria fuzzy algorithm for energy
management of a domestic photovoltaic panel. Renew Energy 2008;33(5):
9931001.
[14] Gelazanskas L, Gamage KA. Demand side management in smart grid: a review
and proposals for future direction. Sustain Cities Soc 2014;11(0):2230.
[15] Adika CO, Wang L. Smart charging and appliance scheduling approaches to
demand side management. Int J Electr Power Energy Syst 2014;57(0):23240.
[16] Pina A, Silva C, Ferro P. The impact of demand side management strategies in
the penetration of renewable electricity. Energy 2012;41(1):12837. 23rd
International conference on efficiency, cost, optimization, simulation and
environmental impact of energy systems, {ECOS} 2010.
[17] Giorgio AD, Pimpinella L. An event driven smart home controller enabling
consumer economic saving and automated demand side management. Appl
Energy 2012;96(0):92103. Smart Grids.
[18] Kyriakarakos G, Piromalis DD, Dounis AI, Arvanitis KG, Papadakis G. Intelligent
demand side energy management system for autonomous polygeneration
microgrids. Appl Energy 2013;103(0):3951.
[19] Xue X, Wang S, Sun Y, Xiao F. An interactive building power demand
management strategy for facilitating smart grid optimization. Appl Energy
2014;116(0):297310.
[20] Rodriguez-Hidalgo M, Rodriguez-Aumente P, Lecuona A, Legrand M, Ventas R.
Domestic hot water consumption vs. solar thermal energy storage: the
optimum size of the storage tank. Appl Energy 2012;97(0):897906. Energy
solutions for a sustainable world proceedings of the third international
conference on applied energy, May 1618, 2011 Perugia, Italy.
[21] Ammar MB, Chaabene M, Elhajjaji A. Daily energy planning of a household
photovoltaic panel. Appl Energy 2010;87(7):234051.
[22] Lilley W, Hayward J, Reedman L. Chapter 7 Realizing the potential of
renewable and distributed generation. In: Sioshansi FP, editor. Smart
Grid. Boston: Academic Press; 2012. p. 16183.
[23] Divya K, stergaard J. Battery energy storage technology for power systems
an overview. Electr Power Syst Res 2009;79(4):51120.
[24] Li J, Guo Y, Platt G, Ward JK. Renewable energy aggregation with intelligent
battery controller. Renew Energy 2013;59(0):2208.
[25] Chesi A, Ferrara G, Ferrari L, Magnani S, Tarani F. Influence of the heat storage
size on the plant performance in a smart user case study. Appl Energy
2013;112(0):145465.
[26] Arteconi A, Hewitt N, Polonara F. State of the art of thermal storage for
demand-side management. Appl Energy 2012;93(0):37189.
[27] Wang WM, Wang J, Ton D. Chapter 5 Prospects for renewable energy:
meeting the challenges of integration with storage. In: Sioshansi FP, editor.
Smart Grid. Boston: Academic Press; 2012. p. 10326.

[28] Chowdhury S, Crossley P. Microgrids and active distribution networks. 2009.

[29] Parisio A, Rikos E, Tzamalis G, Glielmo L. Use of model predictive control for
experimental microgrid optimization. Appl Energy 2014;115(0):3746.
[30] Zipperer A, Aloise-Young P, Suryanarayanan S, Roche R, Earle L, Christensen D,
et al. Electric energy management in the smart home: perspectives on
enabling technologies and consumer behavior. Proc IEEE 2013;101(11):
2397408.
[31] Liu C, Chau K, Wu D, Gao S. Opportunities and challenges of vehicle-to-home,
vehicle-to-vehicle, and vehicle-to-grid technologies. Proc IEEE 2013;101(11):
240927.
[32] Siano P. Demand response and smart grids: a survey. Renew Sustain Energy
Rev 2014;30(0):46178.
[33] Xue X, Wang S, Sun Y, Xiao F. An interactive building power demand
management strategy for facilitating smart grid optimization. Appl Energy
2014;116(0):297310.
[34] De Angelis F, Boaro M, Fuselli D, Squartini S, Piazza F, Wei Q. Optimal home
energy management under dynamic electrical and thermal constraints. IEEE
Trans Indus Inform 2013;9(3):151827.
[35] Severini M, Squartini S, Piazza F. Hybrid soft computing algorithmic
framework for smart home energy management. Soft Comput 2013;17(11):
19832005.
[36] Fuselli D, De Angelis F, Boaro M, Squartini S, Wei Q, Liu D, et al. Action
dependent heuristic dynamic programming for home energy resource
scheduling. Int J Electr Power Energy Syst 2013;48:14860.
[37] Arabali A, Ghofrani M, Etezadi-Amoli M, Fadali MS, Baghzouz Y. Geneticalgorithm-based optimization approach for energy management. IEEE Trans
Power Delivery 2013;28(1):16270.
[38] Soares J, Morais H, Sousa T, Vale Z, Faria P. Day-ahead resource scheduling
including demand response for electric vehicles. IEEE Trans Smart Grid
2013;4(1).
[39] Cellura M, Campanella L, Ciulla G, Guarino F, Nardi Cesarini D, Lo Brano V, et al.
A net zero energy building in Italy: design studies to reach the net zero energy
target. In: Proceedings of building simulation; 2011.
[40] Cellura M, Ciulla G, Lo Brano V, Orioli A, Campanella L, Guarino F, et al. The
redesign of an Italian building to reach net zero energy performances: a case
study of the SHC task 40-ECBCS Annex 52. ASHRAE Trans 2011;117(2).
[41] Currie J, Wilson DI. OPTI: lowering the barrier between open source optimizers
and the industrial MATLAB user. In: Sahinidis N, Pinto J, editors. Foundations of
computer-aided process operations, Savannah, Georgia, USA; 2012.
[42] Sundararajan N, Saratchandran P, Li Y. Fully tuned radial basis function neural
networks for flight control. London, Great Britain: Kluver Academic; 2002.
[43] Yingwei L, Sundararajan N, Saratchandran P. Performance evaluation of a
sequential minimal radial basis function (RBF) neural network learning
algorithm. IEEE Trans Neural Netw 1998;9(2):30818.
[44] Giantomassi A, Ippoliti G, Longhi S, Bertini I, Pizzuti S. On-line steam
production prediction for a municipal solid waste incinerator by fully tuned
minimal RBF neural networks. J Process Control 2011;21(1):16472.
[45] Corradini M, Fossi V, Giantomassi A, Ippoliti G, Longhi S, Orlando G, et al.
Minimal resource allocating networks for discrete time sliding mode control of
robotic manipulators. IEEE Trans Indus Inform 2012;8(4):73345.