974 IEEE TRANSACTIONS ON SMART GRID, VOL. 5, NO.

2, MARCH 2014

A Decentralized Storage Strategy for Residential

Feeders With Photovoltaics
Francesco Marra, Member, IEEE, Guangya Yang, Member, IEEE, Chresten Træholt,
Jacob Østergaard, Senior Member, IEEE, and Esben Larsen

Abstract—This paper proposes a decentralized storage strategy

to support voltage control in low-voltage (LV) residential feeders
with high photovoltaic (PV) capacity installed. The proposed
strategy is capable of preventing overvoltage situations during
high PV generation periods, by the use of locally controlled battery
energy storage systems (ESS) at the PV system grid interface. The
traditional way of operating a domestic ESS is based on charging
the battery as soon as the PV generation exceeds the consumption,
without taking into account overvoltage events during high PV
generation hours; the proposed storage concept improves the
traditional approach, thanks to the provision of voltage support.
A novel method, based on voltage sensitivity analysis, identifies a
common power threshold that triggers the ESSs activation in the
feeder. A Belgian residential LV feeder is used as a case study.
Time-series simulations based on 1-year load and generation Fig. 1. Conventional storage strategy and proposed strategy compared.
profiles verify the method findings and quantify the ESS size in
terms of storage power and energy level.
Index Terms—Decentralized control, energy storage, photo- 70% of the nominal peak-output power of the PV system; this
voltaic systems, power quality, voltage control. corresponds to an inherent curtailment of up to 30% when the
peak-output power is being generated. Some alternative solu-
tions to active power curtailment may involve domestic load
I. INTRODUCTION shifting, to increase local consumption, and energy storage [7],

W ITH the increasing penetration of photovoltaic (PV),

there is a great potential of relieving the loading of
low-voltage (LV) distribution grids. Meanwhile, the LV grid op-
eration encounters more and more uncertainties with regard to
voltage quality [1], [2]. In several European countries and re-
gions in Germany, Spain, Belgium and others, several LV grids
have reached high PV penetration levels with consequences on
the quality of the supply voltage [3].
High generation and low load conditions are the pre-condi-
tions of power flow inversion in a grid feeder; these situations
are likely on a daily basis and may result in overvoltage events
at the different buses [4]. Therefore, solutions to increase the
consumption at houses with roof-top PV, during periods of high
solar irradiation, are needed to preserve voltage quality. In Ger-
many, the vast majority of PV plants, about 13 GW, are con-
nected at the LV grid, causing overvoltage events in different
areas during the year [5]. To avoid such events, from January
2012, a fixed limitation of the active power feed-in by each
PV system is mandatory [5]. As of today, the limit is set to the
Manuscript received February 12, 2013; revised June 15, 2013 and August 10,
2013; accepted September 03, 2013. Date of publication September 26, 2013;
date of current version February 14, 2014. Paper no. TSG-00100-2013.
The authors are with the Electrical Engineering Department, Technical Uni-
versity of Denmark (DTU), Kongens Lyngby DK-2800, Denmark (e-mail: {fm,
gyy, ctr, joe, ela}@elektro.dtu.dk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2013.2281175
[8].
The state of the art on energy management for houses with
PV has mainly focused on demand side management (DSM), to
shape the domestic electric load during periods of high PV gen-
eration and to minimize the energy drawn from the mains [9].
Since the number of flexible domestic appliances is very lim-
ited and because these are not necessarily used on a daily basis
(e.g. the washing machine), the local consumption of PV gen-
eration can be significantly increased with the deployment of
a battery energy storage system (ESS). The conventional size
of domestic ESSs for houses with PV is in the range of 5 kWh
[9], [10], which allows obtaining a local consumption rate of
about 55% yearly [10]. The conventional way of performing
local consumption consists on activating the charging of the
ESS battery as soon as the PV output power is greater than the
house electric load, Fig. 1. However, this strategy does not pre-
vent overvoltage events during maximum PV generation hours
(12:00–14:00 PM), as the ESS battery gets fully charged during
morning hours of sunny days, well before the maximum PV gen-
eration period [9].
In this paper, a novel decentralized storage strategy, for res-
idential feeders with PV is proposed, prioritizing the charge of
ESS batteries around peak generation hours, as shown in Fig. 1.
Each ESS is activated at a certain power threshold, Pth, in Fig. 1,
which is identified by the method proposed. The selection of the
strategy to use, whether the “proposed” strategy or the “conven-
tional” one, Fig. 1, can be optimized using day-ahead solar irra-
diation forecasts [11]; however, these are not used in this paper.

1949-3053 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
MARRA et al.: DECENTRALIZED STORAGE STRATEGY FOR RESIDENTIAL FEEDERS
Fig. 2. Simplified single-line diagram of LV feeder with PV.
The decentralized storage strategy is implemented and sim-
ulated for a residential LV feeder with 23% PV penetration,
using measurement-based 1-year load and generation profiles;
this time interval is a quite representative for the goal of the
paper, thus the stochastic nature of PV generation is not mod-
eled. The advantages of the storage strategy, as opposed to de-
centralized reactive power options for voltage support are also
addressed.
II. LV FEEDERS WITH PV
LV distribution grids are designed and operated in a radial
fashion [4]. The purpose of the LV grid is to provide connection
from the medium-voltage (MV) grid to the supply points for the
individual customers. Typically, a European LV grid consists
of a secondary step-down transformer to step down the voltage
from the MV level to a 0.4 kV line-to-line voltage and of dis-
tribution lines or cables. Voltage regulation is passive in most
cases. When PV systems are connected at a feeder bus, Fig. 2,
the voltage magnitude at the point of common coupling (PCC)
is likely to increase and its expression can be approximated as
follows:
(1)
(2)
(3)
(4)
where
voltage variation vector at the bus;
current vector;
grid voltage vector and its conjugate;
nodal active and reactive power;
active and reactive power feed-in of the
PV system;
load active and reactive power;
equivalent resistance and reactance of the
feeder cable.
The voltage equation is obtained assuming negligible cable
losses, thus the R/X ration can be also neglected. To ensure a
975
correct operation of domestic load appliances, the power quality
Standard EN 50160 is considered, with regard to voltage magni-
tude variations [12]. It shall be ensured that during each period
of one week, the 95% of 10-minute average values of the supply
voltage shall be within the range 10% of , where is
the nominal mean value [12].
III. HOME ENERGY MANAGEMENT
The management of energy storage systems (ESSs) at houses
with PV requires an electrical setup which comprises smart me-
ters, smart sockets, for realizing the load-shift of different appli-
ances, and a main controller to realize load management [13],
[14]. The study in [10] analyzes the local consumption rate for
a cluster of households with annual consumption of 4000 kWh
and a 5 kW PV system. The local consumption rate amounts
to about 25%. To reach higher levels of local consumption, the
house shall be equipped with an ESS, whose operation is co-
ordinated with the PV system. With a 5 kWh battery-ESS, the
local consumption rate rises of about 30% showing a final local
consumption rate of about 55% [10].
The typical AC system in a house with PV and ESS is de-
picted in Fig. 3(a). The PV system is composed of a PV array
and a PV inverter whose AC output power is monitored by the
main controller via a real-time meter. The ESS is connected at
the same point of connection of the PV inverter and its input/
output power is also real-time monitored [15]. A wide range
of battery-ESS technologies can be utilized, such as lead-acid
and Li-ion batteries. However, due to the better performances of
Li-ion battery chemistries in terms of lifetime, energy-to-weight
ratio, self-discharge rate and charge- discharge efficiency, this
technology is adopted for the ESS in this paper [16].
The ESS is modeled with the charge-discharge equations as
expressed in (5) and (6), where is the discharging power and
is the charging power of the ESS battery, respectively;
is the energy stored in the battery at time ; is the duration
time of each interval [17]. The two coefficients and are
the discharge and charge efficiencies respectively.
(5)
(6)
The operation of the battery system should also take into ac-
count power and energy constraints. The maximum power limits
during charging/discharging can be described by (7) and (8) re-
spectively:
(7)
(8)
For simplicity, we refer to to indicate the power flow in-out
the ESS, thus including both the phases of charging and dis-
charging. Positive values of indicate the provision of voltage
support. The state-of-charge (SOC) and energy limits of an ESS
can be described as follows:
(9)
(10)
where
976
Fig. 3. Decentralized storage concept.
and are the minimum and
maximum energy levels of
the storage, defining the
usable energy window;
and are the minimum and
maximum SOC limits that
shall be set in relation to the
application.
IEEE TRANSACTIONS ON SMART GRID, VOL. 5, NO. 2, MARCH 2014
In this paper, the size of the ESS is not pre-determined, in-
stead it is one of the decision variables, in order to comply with
the voltage constraints in the feeder.
IV. PROPOSED METHOD
To lower the voltage profiles during overvoltage periods, the
ESS at each house with PV shall be operated during the max-
imum generation interval, as illustrated in Fig. 1. The system
architecture to achieve voltage support by decentralized storage
is depicted in Fig. 3. Every roof-mounted PV system is cou-
pled with an ESS at the same point of connection. The house
main controller controls the activation of the ESS. If the power
reaches a pre-defined threshold (e.g. the 70% of the
PV system peak-power), the main controller sends a charging
activation signal to the ESS control, corresponding to flag set
to 1, in Fig. 3(b). The flag is set to 0, otherwise. At the end
of the PV generation period, i.e. after sun set, the main con-
troller sets the flag to 2, so the ESS can be discharged back to
the initial state. Depending on the flag setting, the ESS control
generates an ESS activation signal corresponding to “Charge
on”, “Discharge on”, or Idle, as depicted in Fig. 3(c). The ESS
stores the incoming PV energy, by charging with a power level
. The charging process continues as long
as is greater than the threshold. The ESS control can
be also improved introducing a power dead-band to avoid fast
switching events.
A safe usage of the ESS battery should be ensured at all times,
during charging or discharging operation. This task is achieved
by the SOC control, which is the inner control loop of the storage
system. The charging of the ESS battery should be limited to
a SOC window of 20–90%, as determined in [18], for Li-ion
battery types. Other types of batteries may be considered for the
same concept, involving potentially a different usage window.
To provide voltage support in the feeder, the main problem to
solve is the identification of the ESS maximum charging power
under worst case conditions of maximum generation and no
load. This problem is solved using linear programming (LP)
method, based on voltage sensitivity analysis.
V. PROBLEM FORMULATION
The main problem is to identify the power threshold for acti-
vating the ESSs in the feeder, considering that voltage violations
should be avoided. The objective function in (11) shall mini-
mize the ESS size while securing all bus voltages within 1.1
p.u. The method identifies a value for , between 0 and 1, that
represents the share of PV peak-output power used by the ESS
for charging. For instance, the feed-in power limitation of 70%
[5] is equivalent to adopting . The problem of interest
can be solved with linear programming (LP), with the following
formulation:
Objective function:
(11)
MARRA et al.: DECENTRALIZED STORAGE STRATEGY FOR RESIDENTIAL FEEDERS 977

Constraints: TABLE I
PV IN THE FEEDER—23% PENETRATION, ORIGINAL SCENARIO

(12)

where is the number of PV plants in the feeder; is

the charging power of the -ESS; and are the min-
imum and maximum ESS discharging and charging power, re-
spectively; is the column vector containing all bus voltages in
the feeder; is the maximum voltage magnitude allowed.
Constraints (12) are: a) the operative constraints that bound the
voltage magnitude at all buses within the limits; b) the operative
power range for each ESS; c) a common value for respec-
tively. It shall be noted that the voltage constraint a) includes
the network equations that incorporate the characteristics of ca-
bles, load and generation. The expression in (13) describes the
voltage constraint at the generic bus . Reactive power contri-
butions are neglected.

(13)
where

busbar grid voltage magnitude (p.u.); Fig. 4. Single-line diagram of the LV grid feeder.

voltage magnitude at bus (p.u.);

number of buses in the feeder;
active power feed-in by PV on bus ;
aggregated ESS charging power at bus ;
voltage sensitivity coefficient of bus , to the
active power exchanged (p.u./MW).
Due to the use of load flow equations, the aim of the method
is not to precisely quantify the ESS size for voltage support,
but rather to estimate the value of required for a certain PV
penetration level; this value is also dependent on the number of
PV systems in the feeder and on their rated power. With 1-year
time-series simulation, the value of can be confirmed or ad-
justed according to the voltage profiles obtained.
VI. CASE STUDY
The Belgian residential LV feeder [20] is used as case study.
The feeder comprises 7 buses, of which 4 out of 7 hosting
PV systems: bus 2, 4, 5 and 7, respectively. It is composed
of NA2XRY type LV cables and it is part of a larger LV grid
which includes 9 feeders in total [20]. The feeder supplies 33
houses and it embeds 9 single-phase roof-mounted PV systems,
with total PV capacity installed of 42.6 kW. The PV capacity
per bus is indicated in Table I. On bus 2, 4, and 5, each PV
system size corresponds to the value indicated in Table I, while
for bus 7, 6 PV systems contribute as follows: 2 plants of 4.8
kW each; 2 plants of 4.25 kW each and 2 plants of 4.4 kW
each, respectively. The definition of PV penetration used in this
paper refers to the one given in [21]:
(14)
The PV penetration in the feeder is the ratio between the total
PV installed capacity to the nominal feeder capacity. The feeder
capacity is intended as the capacity of the first line section of the
feeder, corresponding to 185 kVA in this case. It follows that a
PV penetration of 23% is present in the feeder.
A. Method Implementation
The method proposed is demonstrated on the grid feeder of
Fig. 4. To reproduce a worst case scenario for voltage magni-
tude variations, the assumption of maximum generation and no
load in the feeder is made; this should ensure a margin from the
voltage limit of 1.1 p.u. with real load and generation profiles.
The method identifies a value of , leading to a maximum
ESS charging power of about 1 kW per house with PV system.
A second scenario assuming 50% PV penetration is investi-
gated with the method. The PV installed capacity for this sce-
nario is indicated in Table II. With this PV penetration level, the
method identifies a value of , leading to a maximum
charging power per ESS of about 6 kW. It is observed that, by
doubling the PV penetration in the feeder, the ESS power re-
quirement increases by 6 times. As a consequence, it is also ex-
pected that the ESS energy level is significantly increased in the
978 IEEE TRANSACTIONS ON SMART GRID, VOL. 5, NO. 2, MARCH 2014

TABLE II
PV INSTALLED CAPACITY IN THE FEEDER—50% PENETRATION

Fig. 8. Power losses in the LV feeder.

TABLE III
ENERGY QUANTITIES IN THE LV FEEDER

Fig. 5. Load in the LV feeder.

power range of 15 to 35 kW; 2) the highest concentration of days

with voltage above 1.1 p.u. is observed in spring and summer.
With regard to grid losses, the total energy losses in the feeder
are calculated considering the losses on all feeder cables; the
total power losses profile, during the 12-month period, is de-
picted in Fig. 8. It is observable that losses are lower in spring
and summer and higher in winter and autumn.
Fig. 6. PV generation in the LV feeder. With regard to energy levels, the total energy consumption,
energy production and energy losses are calculated and summa-
rized in Table III.

VII. DECENTRALIZED STORAGE

The method identified a value of for the scenario
with 23% PV penetration and a value of for 50% PV
penetration. With time-series simulations, the following cases
are investigated:
Fig. 7. “Generation-load” in the LV feeder. Case 1) which considers , as obtained by the method
application with 23% PV penetration; all private
ESSs are activated when .
second scenario; both scenarios are investigated via time-series Case 2) which considers ; all private ESSs are acti-
simulations using 1-year load and generation profiles. vated with , as required by
the Renewable Energy Sources Act, EEG 2012, in
B. Load and Generation in the Feeder Germany [6].
Case 3) which considers , as obtained by the
The ESS operation is analyzed with time-series simulations method application with 50% PV penetration; all
for a 12-month period using 15-minute average load and gen- residential ESSs are activated with
eration profiles. A previous study performed for the feeder has .
shown overvoltage events along the feeder buses [22]. While for case 1 and case 3 are applied as a result of the
The aggregated load and generation in the feeder, from the proposed method, the value of for case 2 is predetermined, as
1st of January till the 31st of December, are depicted in Fig. 5 it is based on an existing grid connection requirement [6].
and Fig. 6, respectively. The 12-month simulation shows that
voltage rise events at bus 7 are the most critical, due to voltage
A. Case 1
magnitude violations in 98 days.
In Fig. 7, the difference “PV generation (G)-load (L)” in those With , the feed-in power by each PV system is at most
days with voltages magnitude above 1.1 p.u. is depicted. For the 80% of . The 1-year charging activity of the 9 ESSs
graphical reasons, for each of these days, only one value of G-L is depicted in Fig. 9. The maximum charging power is about 1.1
is depicted, though this value may change during the entire over- kW, which is in line with the method results. The phase voltage
voltage period. Some major findings are: 1) overvoltage events profiles , and of the most critical feeder location, bus
occur with a positive “G-L” difference that corresponds to the 7, are depicted in Fig. 10. The profiles are clearly below the
MARRA et al.: DECENTRALIZED STORAGE STRATEGY FOR RESIDENTIAL FEEDERS 979

Fig. 9. ESS charging power for voltage support, .

Fig. 13. ESS energy required for voltage support, with .

Fig. 10. Phase voltage profiles on bus 7, with .

Fig. 14. ESS energy required for voltage support, with .

of is required to comply with the voltage limit of 1.1

Fig. 11. ESS charging power for voltage support,
p.u.
The assumptions made to increase the PV penetration level
from 23% to 50% have been the following:
— same number of PV systems as in the original case, with
.
augmented installed capacity per system;
— twice the number of PV systems than the original case, i.e.
18 PV systems instead of 9.
With 9 PV systems in the feeder, the maximum charging
power of each ESS is on average 6 kW; while, with 18 PV sys-
tems, the maximum charging power results of about 3 kW.

VIII. ENERGY LEVELS QUANTIFICATION

Fig. 12. Phase voltage profiles on bus 7, with . , the required ESS energy is 1.1 kWh, Fig. 13. This
voltage magnitude of 1.1 p.u., showing a margin of about 0.01
p.u. from the limit, which is acceptable.
B. Case 2
With , the feed-in power of the PV system is at most
the 70% of . The charging activity of the residential
ESSs is depicted in Fig. 11. In this case, the maximum charging
power is about 1.7 kW and the ESSs operate for 147 days. The
phase voltage profiles , and of bus 7 are depicted in
Fig. 12. Compared to the voltage profiles obtained for
a wider margin from the limit of 1.1 p.u. is observed.
C. Case 3
With 50% PV penetration, the total feed-in power by PV sys-
tems is at most 92 kW. Such power amount leads to 231 days
with voltage magnitude violations. While the method applica-
tion found a value for of 0.55, simulations show that a value
The ESS energy levels are calculated based on the power pro-
files obtained for case 1, 2 and 3. With 23% PV penetration and
value is obtained by selecting the day with the highest storage
power and energy content. Whereas, with , the required
ESS energy is 3.8 kWh, Fig. 14.
With 50% PV penetration and , the required ESS
energy level with 9 ESSs in the feeder is 28 kWh per system,
whereas 14 kWh, if we consider twice the number of ESSs.
It shall be considered that the identified energy levels are re-
ferred to the usable energy window (in kWh) of the ESSs. Ac-
cording to [18], to operate the ESS battery within a linear SOC
region and to avoid deep discharge cycles, the battery should
be oversized in relation to the usable energy window identified
, with simulations. For each ESS battery, a usable SOC window
of 20% to 90% is considered in this paper, complying with the
recommendations made on the use of Li-ion batteries in [18].
The nominal battery capacity , the minimum and maximum
energy constraints and can be obtained and these
are indicated in Table IV.
With 23% PV penetration, voltage support with re-
quires a nominal battery capacity of about 1.6 kWh; instead,
980
TABLE IV
ESS OPERATIVE CONSTRAINTS
Fig. 15. Estimated ESS battery lifetime for the different scenarios.
with , a battery capacity of 5.4 kWh should be used.
It is also evident that the maximum charging and discharging
power of the ESS is not a critical requirement for the provision
of voltage support; in fact, the power level of 2 kW is enough
for both -cases.
Significant differences are observed in power and energy
levels with the 50% PV penetration scenario. With either 9
ESSs or 18 ESSs in the feeder, the storage option is not eco-
nomically feasible by private PV owners.
A. ESS Battery Lifetime
The estimation of the battery lifetime is limited to the sole
voltage support operation of the ESS, regardless of the ESS op-
eration in days without voltage problems. The number of cycles
performed during the 12-month period to support voltage con-
trol, and the declared battery lifetime (number of cycles) at 80%
depth-of-discharge (DOD), are considered for the estimation of
. Considering a Li-ion battery with 1500 declared cycles [23],
is approximated as:
(15)
For the three scenarios of , the diagram of Fig. 15 is obtained.
Under the 23% PV penetration scenario, it appears that
voltage support using decentralized storage does not impact
significantly the battery lifetime, if we consider that today’s
Li-ion batteries are warranted for about 8–10 years [23]. A
considerably shorter lifetime is obtained with the 50% PV
penetration scenario; this is due to the higher number of cycles
required for voltage support during one year.
IEEE TRANSACTIONS ON SMART GRID, VOL. 5, NO. 2, MARCH 2014
TABLE V
ENERGY LOSSES IN THE LV FEEDER FOR DIFFERENT SCENARIOS
B. Energy Losses in the Feeder
A comparison of energy losses in the feeder during the
12-month period with 23% PV penetration is performed ac-
cording to the cases of:
1) No storage;
2) Decentralized storage (DS) with ;
3) Decentralized storage (DS) with ;
The results obtained are indicated in Table V. With ,
the energy losses are reduced by 6.6% compared to the case
without storage. If voltage support is operated with ,
the energy losses are reduced of about 7.3%. Losses can be fur-
ther reduced if the ESSs operate also in those days without over-
voltage problems. In [22], reactive power solutions are used for
the same feeder to provide voltage support.
In all cases, a significant reactive power import is observed,
with increased energy losses. Furthermore, with reactive power
options by PV units, the PV active power output shall be reduced
according to the rated capacity of the unit; this aspect represents
an indirect active power curtailment.
IX. CONCLUSION
In this paper, a decentralized storage concept for feeders with
high PV penetration has been set forth that can be potentially
used as a storage planning tool.
The main contribution of the work is based on the qualita-
tive results obtained with the method, particularly applicable to
situations of LV feeders with high PV penetration. The deploy-
ment of private ESSs at houses with PV appears a promising
option for accommodating increased PV penetration levels. The
storage strategy proposed improves the traditional way of con-
trolling ESSs at houses with PV: battery charging is activated
according to an optimized power threshold, instead of “as soon
as ”. The strategy allows also a reduction of energy
losses in the feeder, compared to the scenario without storage
and to scenarios using reactive power options.
For sizing the domestic ESSs, a method based on voltage sen-
sitivity analysis has been proposed, which identifies the power
threshold for the charging activation of the ESS battery. By
testing the method on a simulated feeder with 23% PV pene-
tration, it is found that all ESSs shall be activated starting from
the 80% of the peak-output power of each PV system. Instead,
the application of the 70% power threshold, as required by EEG
2012, would lead to over sizing the ESS batteries. A battery life
of 15 years is obtained with the 80% power threshold, while 10
years is obtained with 70%.
By scaling up the PV penetration level to 50%, the method
highlights that the power limitation of 70% does not guarantee
the required voltage support in the feeder. Instead, a power limi-
tation of 50% from each PV system is required. In the latter case,
MARRA et al.: DECENTRALIZED STORAGE STRATEGY FOR RESIDENTIAL FEEDERS
the required ESS size results amplified of about 6 times com-
pared to the original penetration level; this would make voltage
support by private PV owners uneconomical.
Though the decentralized storage strategy proposed is applied
to a specific case study, other distribution networks with dif-
ferent topologies and load patterns have been analyzed by the
authors, obtaining similar qualitative results. Distribution net-
work operators can potentially benefit from the findings of the
paper, as grid reinforcement and active power curtailment can
be postponed.
REFERENCES
[1] A. Bracale, P. Caramia, G. Carpinelli, A. Russo, and P. Verde, “Site
and system indices for power-quality characterization of distribution
networks with distributed generation,” IEEE Trans. Power Syst., vol.
26, no. 3, pp. 1304–1316, Jul. 2011.
[2] M. H. J. Bollen, Understanding Power Quality Problems: Voltage Sags
and Interruptions. Piscatway, NJ, USA: IEEE Press, 2000.
[3] B. Renders, L. Vandevelde, L. Degroote, K. Stockman, and M. H. J.
Bollen, “Distributed generation and the voltage profile on distribution
feeders during voltage dips,” J. Electr. Power Syst. Res., vol. 80, no.
12, pp. 1452–1458, 2010.
[4] M. Delfanti, M. Merlo, M. Pozzi, V. Olivieri, and M. Gallanti, “Power
flows in the Italian distribution electric system with dispersed genera-
tion,” in Proc. 20th Int. Conf. Exhib. Electricity Distrib., 2009, Part 1,
CIRED.
[5] T. Stetz, F. Marten, and M. Braun, “Improved low voltage grid-integra-
tion of photovoltaic systems in Germany,” IEEE Trans. Sustain. Eng,
vol. PP, no. 99, 2012.
[6] Renewable Energy Sources Act (EEG) 2012 [Online]. Available:
www.bmu.de
[7] B. H. Bailey, J. R. Doty, R. Perez, R. Stewart, and J. E. Donegan,
“Evaluation of a demand side management photovoltaic system,” IEEE
Trans. Energy Conv., vol. 8, no. 4, pp. 621–627, Dec. 1993.
[8] S. J. Chiang, K. T. Chang, and C. Y. Yen, “Residential photovoltaic
energy storage system,” IEEE Trans. Ind. Electr., vol. 45, no. 3, pp.
385–394, Jun. 1998.
[9] M. Castillo-Cagigala, E. Caamaño-Martínb, E. Matallanasa, D. Masa-
Boteb, A. Gutiérreza, F. Monasterio-Huelina, and J. Jiménez-Leubea,
“PV self-consumption optimization with storage and active DSM for
the residential sector,” J. Solar Energy, vol. 85, no. 9, pp. 2338–2348,
Sep. 2011.
[10] J. Binder, C. O. Williams, and T. Kelm, “Increasing PV self-consump-
tion, domestic energy autonomy and grid compatibility of PV systems
using heat-pumps, thermal storage and battery storage,” in Proc. 27th
Eur. Photov. Solar Energy Conf., Frankfurt, Germany, 2012.
[11] E. Lorenz, J. Hurka, D. Heinemann, and H. G. Beyer, “Irradiance fore-
casting for the power prediction of grid-connected photovoltaic sys-
tems,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sensing, vol.
2, no. 1, pp. 2–10, Mar. 2009.
[12] EN 50160, “Voltage characteristics of electricity supplied by public
distribution networks,” CENELEC. Brussels, Belgium, 1999.
[13] T. Ganu, D. P. Seetharam, V. Arya, R. Kunnath, J. Hazra, S. A. Husain,
L. C. Da Silva, and S. Kalyanaraman, “nPlug: A smart plug for allevi-
ating peak loads,” in Proc. 3rd Int. Conf. Future Energy Syst.: Where
Energy, Computing Comm. Meet (e-Energy), May 2012.
[14] F. Shahnia, M. T. Wishart, A. Ghosh, G. Ledwich, and F. Zare, “Smart
demand side management of low-voltage distribution networks using
multi-objective decision making,” IET Generat., Transm. Distrib., vol.
6, no. 10, pp. 968–1000, Oct. 2012.
[15]
X. Liu, A. Aichhorn, L. Liu, and H. Li, “Coordinated control of
distributed energy storage system with tap changer transformers for
voltage rise mitigation under high photovoltaic penetration,” IEEE
Trans. Smart Grid, vol. 3, no. 2, Jun. 2012.
[16] K. Chen, A. Bouscayrol, A. Berthon, P. Delarue, D. Hissel, and R.
Trigui, “Global modeling of different vehicles,” IEEE Veh. Tech. Mag.,
vol. 4, no. 2, pp. 80–89, Jun. 2009.
981
[17] S. X. Chen, H. B. Gooi, and M. Q. Wang, “Sizing of energy storage for
microgrids,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 142–151, Mar.
2012.
[18] F. Marra, G. Y. Yang, C. Træholt, E. Larsen, C. N. Rasmussen, and S.
You, “Demand profile study of battery electric vehicle under different
charging options,” in Proc. IEEE Pow. Eng. Soc. Gen. Meet., Jul. 2012.
[19] “DIg Silent PowerFactory, User Manual,” .
[20] F. Marra, Y. T. Fawzy, T. Bülo, and B. Blažič, “Energy storage options
for voltage support in low-voltage grids with high penetration of pho-
tovoltaic,” in Proc. IEEE Innov. Smart Grid Techn., ISGT 2012, Oct.
2012.
[21] V. Mendez Quezada, J. Rivier Abbad, and T. Gómez San Román, “As-
sessment of energy distribution losses for increasing penetration of
distributed generation,” IEEE Trans. Power Syst., vol. 21, no. 2, pp.
533–540, May 2006.
[22] B. Blazic, I. Papic, B. Uljanic, B. Blatterie, C. Dierckxsens, K. De Bra-
bandere, W. Deprez, and T. Fawzy, “Integration of photovoltaic sys-
tems with voltage control capabilities into LV networks,” in Proc. of
1st Int. Workshop on Integr. of Solar Pow. Into Pow. Syst., 2011.
[23] LTC, GAIA Advanced Lithium Battery Systems, Handling Instructions
for the Lithium Ion Cell Type HE 602030 NCA-55 Ah/198 Wh [On-
line]. Available: www.gaia-akku.com
Francesco Marra was born in Copertino, Italy, in 1984. He received the B.Sc.
and M.Sc. degrees in electronic and mechatronic engineering from Polytechnic
of Turin, Turin, Italy, in 2006 and 2008, respectively. Since June 2013, he has
been pursing the Ph.D. degree in electrical engineering at the Technical Univer-
sity of Denmark, Lyngby, Denmark.
His fields of interest include renewable energy integration and control engi-
neering.
Guangya Yang received the B.Sc. and M.Sc. degrees, all in electrical engi-
neering, from the Shandong University, China, in 2002 and 2005, respectively.
In 2008, he received the Ph.D. degree in electrical engineering from the Univer-
sity of Queensland, Australia.
Currently, he is a Research Scientist with the Department of Electrical Engi-
neering of the Technical University of Denmark, Lyngby, Denmark. His fields
of interest include power system operation and control, renewable energy inte-
gration and wide-area system monitoring and protection.
Chresten Træholt received the M.Sc. degree in electrical engineering in 1987
and the Ph.D. degree in materials science in 1994, both from the Technical Uni-
versity of Denmark, Lyngby, Denmark.
Since then, he has spent several years on electron microscopy and materials
research at the Technical University Delft, the Netherlands as well as several
years of experience with the superconductor cable industry. His current fields of
interest include smart grids, renewable energy, PV, wind power, electric vehicles
and superconductivity.
Jacob Østergaard is head of the Centre for Electric Technology and head of
section for Electrical Energy Systems at DTU Electrical Engineering.
He is also head of the experimental platform for electricity and energy, Pow-
erLabDK. He is member of the Advisory Council of the EU Technology Plat-
form SmartGrids. He is coordinator of the M.Sc. program in Wind Energy (elec-
tric). His research focuses on the development of future intelligent power system
with increased share of decentralized and environmentally friendly electricity,
including wind energy.
Esben Larsen received the M.Sc. degree in electrical engineering from the
Technical University of Denmark (DTU), Lyngby, Denmark, in 1977.
He is currently Associate Professor at DTU. His main areas of interest in-
clude: high voltage engineering, wind power, photovoltaic, geothermal, hydro
power, micro combined heat and power. He has been manager of “Information
and Knowledge Center of Electric Vehicles” at DTU, in 2000–2003.