Content uploaded by Francesco Tedesco
Author content
All content in this area was uploaded by Francesco Tedesco on Nov 27, 2017
Content may be subject to copyright.
A Command Governor Approach for the
Voltage Control in Smart Grids with
Distributed Generation and Storage
Devices
Alessandro Casavola, Francesco Tedesco and Maurizio Vizza
Dipartimento di Ingegneria Informatica, Modellistica, Elettronica e
Sistemistica, Universit`a degli Studi della Calabria, Via Pietro Bucci,
Cubo 42c, Rende (CS), 87036, Italy,
{casavola,ftedesco,mvizza}@dimes.unical.it
Abstract: High penetration of Distributed Generation (DG) plants in Medium Voltage
(MV) / Load Voltage (LV) power grids may lead to abrupt voltage raises. Typical critical
scenarios are represented by either low demand conditions or high power production from
renewable sources. Traditional approaches used to face such a situation involve both the
disconnection of the distributed generators and the curtailment of the generated power leading
to several disadvantages. However, new technologies allow an active orchestration between
some controllable devices of the grid, e.g. distributed generators, MV/HV transformers, storage
devices in order to maintain relevant system variables within prescribed operative constraints
in response to unexpected adverse conditions. This work addresses the online management of
distributed generators and storage devices. The approach is based on Command Governor (CG)
ideas and is based on the resolution at each time instant of an optimization problem containing
explicit constraints related to voltage bounds and operational limits of adopted devices.
Keywords: Voltage Regulation, Smart Grids, Distributed Predictive Control, Command
Governor, Storage devices.
1. INTRODUCTION
The increasing of smallscale distributed generation plants
based on renewable energetic sources paves the way to
new applications (e.g. electrical vehicle recharge stations)
whose developments would have otherwise required strong
investments for further power lines. Anyway the scatter
along power grids of local generators gives rise to a number
of technical challenges related to their safe exploitation.
An important issue induced by the higher and higher level
of penetration of DG systems is the voltage rising levels
along the distribution feeders. Several approaches have
been presented in the literature to cope with this issue.
Standard methods foresee an active action of an onload
tap changer (OLTC) in a feeder transformer (Ferry et al.
[2007]) in order to manage the voltage on the main bus of
the MV/LV network. Nevertheless, when a large number
of buses are connected to the distribution network, it may
become virtually impossible to maintain the buses voltage
proﬁles within their nominal ranges by only acting on the
OLTC tap positions, especially when the DG production
is high and reverse power ﬂow occurs. A conventional
expedient adopted in this case would be the immediate
disconnection of the DG sources that cause the voltage
limits violation. Unfortunately, many disconnections rep
resent an economic loss for the network and the DG plants
would result unusable.
Nowadays DG devices come equipped with inverters capa
ble to control their active and reactive power. This aspect
has inspired alternative and more eﬀective approaches for
the voltage control that are based either on the active
power curtailment (Reinaldo and Lopes [2011]) or on reac
tive power injection/absorption (Bletterie et al. [2012]) or
both reactive and active power regulation. All previously
mentioned methods carry out a decentralized action on
local generators and do not foresee any sort of coordina
tion among them. From one hand this aspect avoids the
need of communication infrastructures, on the other hand,
without systemwide coordination, sound algorithms are
diﬃcult to design. In fact, because control actions are com
puted on the basis of a partial knowledge of the network,
only suboptimal performance can be achieved.
The introduction of advanced power electronic and ICT
systems into modern smart grids opened to new commu
nication facilities involving DGs and OLTC and gave rise
to several more eﬀective control strategies mainly based on
the presence of a central controller (Casavola et al. [2011])
capable of monitoring and managing the entire MV/LV
grid.
More recently the use of storage devices in the grids,
usually exploited to increase selfconsumption, quite often
represents a new degree of freedom with respect to the
voltage control problem. In fact, without any storage
facility, the possibility to transfer reactive power exists
only when the generators are able to inject active power
too. On the contrary, the presence of storage devices allows
one to inject/absorb reactive power also when the local
Preprints of the 20th World Congress
The International Federation of Automatic Control
Toulouse, France, July 914, 2017
Copyright by the
International Federation of Automatic Control (IFAC)
10420
~ ~
~ ~
~ ~
HV
T
V1V2VN1
SDG1,1 SDG1,2 SDG
N,1
L11L12L1N1
Load 11 Load 12
VMV
~
Load N,1
V1V2VN2
SDG2,1 SDG2,2 SDGN,2
L21L22L2N2
Load 2.1 Load 2.2
~
Load N,2
V1V2VNl
SDGl,1 SDGl,2 SDGN,l
Ll1Ll2LlNl
Load l,1 Load l,2
~
Load N,l
Fig. 1. Distribution Network Scheme
generators are not able to transfer active power, moreover
they can be controlled to minimize the curtailment of DC
source during overvoltage conditions.
This paper proposes a Command Governor (CG) (Bem
porad et al. [1997]) supervision scheme for the online
coordination of OLTC, DGs and storage devices for the
optimal active power loss reduction and minimization of
voltage deviation from nominal conditions in the presence
of timevarying loads. In particular, here we extend the
work presented in Casavola et al. [2017] to the more chal
lenging case where storage units are present at generator
sides.
The CG is an eﬀective methodology for supervising a sys
tem within pointwiseintime setmembership constraints
on relevant variables. It can be used to modify the set
points of the controlled devices of the grids in such a way
that input/state constraints are satisﬁed while maintain
ing tracking performance (Gilbert et al. [1995]).
Following the same lines of Casavola et al. [2017], the
CG based strategies will be introduced and shown to be
eﬀective in accomplishing the voltage control problem.
The paper is organized as follows. The problem of the
voltage control for MV power grids is stated in Section
II. In Section III, the CG design scheme is presented for
the problem at hand. Computer simulations are ﬁnally
presented in Section IV and some conclusions end the
paper.
2. PROBLEM STATEMENT
2.1 Control Oriented Modeling
Power grids under analysis can be described through
the general scheme depicted in Figure 1 that repre
sents a generic radial distribution network operating in
medium/low voltage (MV/LV). There lfeeders are sup
plied through the main MV bus by an HV/MV transformer
equipped with onload tap changers (OLTC). The jth
feeder consists of Njnodes to which small size generators
and/or loads are connected. Both loads and generators,
the latter supposedly consisting of smallsize renewable en
ergy generators (wind, sun, hydro, etc.), are characterized
by a high degree of uncertainty because they follow the
customer behavior and the primary source availability. It
is assumed that distributed generators are endowed with
”smart” functionalities in the sense that it is possible to
specify, within certain limits, the desired amount of ac
tive/reactive power to be provided/absorbed with respect
to the grid. For this reason in the paper they will be de
noted as Smart Distributed Generators (SDG). Moreover,
it is assumed that each generator has a direct link with
a local storage device. Without loss of generalities the
voltage at the secondary sides of the HV/MV transformer
can be modeled as an ideal voltage source ( ¯
EMV ), whose
amplitude can be controlled by the OLTC tap positions.
In order to present a mathematical model of the above
described network, it will be assumed the presence of
balanced loads only, symmetrical generation and a linear
network behavior. Under such assumptions the network of
Figure 1 can be represented by the singlephase equivalent
circuit depicted in Figure 2 (see Casavola et al. [2011] for
details), where the set of N=N1+N2+... +Nlnodes is
highlighted.
The following notation will be adopted
˙
ZLi
,RLi+jXLi, i = 1, ..., N,
˙
ZCi=RCi+jXCi, i = 1, ..., N,
¯
Ei(t),EDi(t) + jEIi(t), i = 1, ..., N ,
¯
EMV (t),EM V D(t),
(1)
where the electrical parameters: ˙
ZLi, RLiand XLidenote
respectively the line impedances, resistances and reac
tances, ˙
ZCi, RCiand XCithe nominal load impedances,
resistance and reactances, ¯
Ei(t) and ¯
EMV (t) complex
numbers representing respectively the voltages on the ith
node and on the HV/MV transformer secondary side. For
1
Fig. 2. Equivalent circuit
the same reason, also each SDG unit will be assumed to
consist of linear subsystems and represented as a current
generator, namely
¯
Ji(t),JDi(t) + jJIi(t) (2)
Also in this case the currents ¯
Ji(t) drawing into ith node
are represented in complex form. In particular, notice
that the ith generator, being a smart device, allows
the manipulation, within certain limits, of its generated
complex power ¯
Si(t),¯
Ei(t)¯
J∗
i(t) = Pi(t) + jQi(t),
where Pi(t) and Qi(t) denote active and reactive power
respectively.
In the proposed setting, each SDG is provided with a
local storage device that is modeled via a quasikinetic
Preprints of the 20th IFAC World Congress
Toulouse, France, July 914, 2017
10421
battery model (Manwell J.F. and McGowan [1993]), which
consists of a tank storing a certain amount of energy xb
i(t)
at time step t, that evolves according to the following
discretetime diﬀerence equation
xb
i(t+ 1) = τxb
i(t) + ∆T P b
i(t) (3)
with τ≤1 denoting the selfdischarge decay (McEvoy et
al. [2003]), ∆Tthe sampling interval and Pb
i(t) a certain
fraction of the power Pr
i(t) generated by the ith local
SDG according to
Pb
i(t) = Pr
i(t)−Pi(t) (4)
where Pr
i(t) is the direct current power contribution pro
vided by the local renewable source at time t, i.e. Pr
i(t)≥0
if a source is present, otherwise Pr
i(t) = 0. For simplicity,
(3) presupposes that the round trip eﬃciency of the bat
teries is 1.
Hence, the following relations result for i= 1, ..., N
¯
ITi(t)−¯
Ii(t) + ¯
Ji(t) = 0 (5)
with ¯
Ii(t),¯
Ei(t)/˙
ZCi(t). Then, by resorting to the
Kirchhoﬀ’s circuit laws, the following impedance matrix
description can be achieved for the distribution network
¯
EMV (t)
¯
E1(t)
¯
E2(t)
.
.
.
¯
EN(t)
=
˙
Z1,1, ..., ˙
Z1,N +1
.
.
.....
.
.
˙
ZN+1,1, ..., ˙
ZN+1,N +1
¯
IMV (t)
¯
IT1(t)
¯
IT2(t)
.
.
.
¯
ITN(t)
(6)
where ¯
IMV (t) denotes the HV/MV transformer current
and ˙
Zi,i =ϕ(˙
ZLi,˙
ZCi) the entries of the impedance ma
trix. We shall now consider the presence of additional in
strumental disturbances acting on the system for modeling
load changes, injected active power and injected/absorbed
reactive power (e.g. Electric Vehicle charging stations). It
is wellknown that the injection of active/reactive power
alters the loads behavior in a nonlinear way. In order to
take these aspects into account, we allow the load distur
bances to vary to a large extent. Speciﬁcally, we assume
that the module of the current absorbed by the loads may
vary within the following range
[¯
Ii−δIi,¯
Ii+δIi] (7)
where δIi=γ¯
Iiaccounts for both measurement errors
over the voltage proﬁle of all system buses and for load
power variations. Allowable ranges corresponding to the
choice γ≈0.25 are typically considered in the literature,
see Barboza et al. [2004].
Therefore, from eqs. (5)(7) and deﬁnitions (1)(2), the
following discretetime state space description can be
derived through straightforward algebraic manipulations
x(t+ 1) = Φx(t) + Gg(t) + Gdd(t)
y(t) = Hx(t)(8)
where
x(t),[¯
E1(t),··· ,¯
EN(t), xb
1(t),··· , xb
N]T
g(t),[¯
EMV (t),¯
J1(t),··· ,¯
JN(t)]T
d(t),[δI1(t),...,δIN(t)]T
y(t),Hx(t)
(9)
and
y(t) = [ ¯
E1(t), ..., ¯
EN(t)]T(10)
Moreover, the allowable disturbances d(t) is supposed to
be conﬁned inside the compact set
Pi
+SnSn
Qi
PMAX c i
PMAX d i
Fig. 3. DG & Storage capability chart limits
D,{d:−[δI1,...,δIN]T≤d≤[δI1,...,δIN]T}(11)
The explicit structures of matrices Φ, H, G and Gdare not
included here for space reasons but details can be found
in Casavola et al. [2007].
2.2 Operational Constraints and Control Requirements
The system described by (8) is subject to the following set
of constraints:
•Load voltages ¯
Ei(t), i = 1,2, ..., N cannot be imposed to
equal the reference value Enom, because the voltage drops
are diﬀerent each others in the various distribution lines.
Therefore, the following constraints must be fulﬁlled at
each time instant:
(1 −α)E2
nom ≤  ¯
Ei(t)2≤(1 + α)E2
nom, i = 1, ..., 12,∀t
(12)
Deviations of ¯
Ei(t), i = 1,2, ..., N within ±10% (α∈
(0.04 0.1]) are reasonable in practice.
•As it is wellknown, OLTC alters the power transformer
turns ratio in a number of predeﬁned steps and the value
of ¯
EMV (t) can vary in percentage with respect to the
nominal voltage Enom as follows
Enom ≤  ¯
EMV (t) ≤ (1 + β)Enom ,∀t, β ∈(0 0.1] (13)
•According to the CEI016 Italian Electrotechnical Com
mittee norm CEI016 [2015], the complex power Si(t)
generated by the ith generator with storage should be
constrained as (Figure 3)
¯
Si(t) ≤ Sni(14a)
−PMAX di≤Pb
i(t)≤ −PMAX ci(14b)
where Sniis the maximum reachable apparent power
(capability). Moreover PMAX di, PM AX ciare the dis
charge/charge power limits.
•Obviously, the quantity of storable energy in each local
battery is limited because the capacity of the battery is
constrained by a quantity x, i.e.
xb
i(t)≤x(15)
Furthermore, an additional constraint bounding the min
imum level of stored energy is taken into account
xb
i(t)≥x(16)
where xis the minimum amount of energy that should be
stored in the battery. Moreover, according to the kinetic
battery model, only a certain amount of stored energy is
immediately available for charging or discharging, being
the remaining chemically bound.
The control problem of the above described power grid
relies on the achievement of the following goals
Preprints of the 20th IFAC World Congress
Toulouse, France, July 914, 2017
10422
Meausurement signal
Control signal
Reference signal
Fig. 4. CG supervision scheme for smart grid
O1) At main bus level, it is worth to maintain the voltage
EMV (t) close to the nominal voltage Enom . Modiﬁ
cation of EM V (t) should take place with a smooth
behavior;
O2) At nodes level, the main goal is to transfer the
available quantity of power Pr
i(t) by acting on the
active power Pi(t). In other words, this goal translates
into minimizing the power transferred Pb
i(t) to/from
the local battery;
O3) At the same time, one would like to minimize the
amount of absorbed/injected reactive power Qi(t)
The above management policies can be expressed as a
multiobjective optimal control problem. In fact, objective
O1can be achieved by minimizing the following perfor
mance indicators:
J1a(t),kEMV (t)−Enom k2
2,W1a,(17a)
J1b(t),kEMV (t)−EM V (t−1)k2
2,W1b,(17b)
where k·k2,W denotes the 2norm weighted by a symmetric
and positive matrix W.
The second goal is enforced by deﬁning the following
performance index
Ji
2(t),kPr
i(t)−Pi(t)k2
2,W2(18a)
whereas the third goal can be translated into the mini
mization of the local cost
Ji
3(t),kQi(t)k2
2,W3,(19a)
3. COMMAND GOVERNOR DESIGN
In this section a CG approach to solve the above stated
problem is presented. The centralized CG solution is
customized to the present application where a unique
centralized CG device is in charge of supervising the
OLTC device, the smart generators and the storage devices
distributed along the feeders. The CG scheme of interest
here is depicted in Figure 4. There, a power grid network is
supervised by a CG device via the command g(t) already
deﬁned in (9). In particular, at each time t, on the basis
of the measured state x(t) and nominal reference Enom(t),
the command g(t) is computed in order to minimize the
performance indexes (17)(18) subject to the above stated
pointwiseintime setmembership constraints (12) −(14),
compactly represented as
c(t)∈ C,∀t∈Z+,(20)
to be held true along the system trajectories generated by
the CG, where
c(t) = [ ¯
E1,..., ¯
EN, xb
1,··· , xb
N,¯
EMV ,¯
J1,..., ¯
JN]T(21)
The main idea behind the CG approach is that of selecting
and applying at each time step a virtual command g(t)≡
w,∀t, chosen in such a way that, if constantly applied to
the system over a semiinﬁnite horizon k∈Z+, from the
initial state x(t), the future predictions (virtual evolutions)
of the cvariable along the virtual time kdo not violate
constraints c(k, x(t), w, d(k)) ∈ C,∀k∈Z+, for all possible
bounded disturbance d(k)∈ D realizations. Because dis
unknown, this is obtained by selecting the command win
a robust way in the following set
V(x),{w: ¯c(k, x, w)∈ Ck,∀k∈Z+}(22)
It is worth mentioning that, if (8) is asymptotically stable,
the set V(x) is ﬁnitely determined (see Gilbert et al.
[1995]).
/MV
T
V1V2V3
V6V7V8
SDG1SDG2SDG3
L11L12L13
L21L22L23
C
VMV
~ ~ ~
1C2C3
C6C7C
HV
V4V5
SDG4SDG5
L15
~
C4C5
V9V10
L25
C9C
L24
L14
810
~
B1B2
Fig. 5. Simulated network scheme
Finally, the CG design problem is solved by choosing at
each time instant ta command g(t), which is the solution
of the following optimization problem:
g(t) = arg min
w∈V(x(t)) J(t) (23)
where J(t),J1a(t) + J1b(t) +
N+1
X
i=2 Ji
2(t) + Ji
3(t)
See also Bemporad et al. [1998] for computational and
implementation details.
4. ILLUSTRATIVE EXAMPLE
4.1 Case Study Description
Consider in Fig. 5 a radial power grid obtained by suitably
simplifying and modeling a typical portion of the MV
Italian distribution network. The grid under consideration
consists of:
•A HV/MV station, connected to a 132 kV HV busbar
through a 132/20 kV, 40 M V A, transformer, equipped
with On Load Tap Changer (OLTC), at the primary side
and to 20 kV MV busbar, at their secondary side;
•The ﬁrst feeder F1consists in 5 nodes over which loads
characterized by nominal power P cF1= 900kW are
supplied by local generators SDG with local storage units.
The nominal power Pnof each SDG is 700 k W , with a
local storage capacity equal to 400k W h.
•The second feeder F2is composed by 5 nodes having
loads with nominal power P cF2= 850k W . Moreover the
ﬁrst two nodes of F2are equipped with storage devices
characterized by nominal power Pn= 300kW and storage
capacity equal to 400kW h. From a mathematical point
of view that nodes are able to transfer active power Pi(t)
to/from the network without any contribution of external
renewable source (i.e. Pr
i(t) = 0 in (4)).
Preprints of the 20th IFAC World Congress
Toulouse, France, July 914, 2017
10423
All the transmission lines are supposed to be 3km long
and characterized by an impedance ˙
ZL= 5.424 + j9.216
as all capacitive eﬀects are neglected.
4.2 Simulations
We are interested in showing the eﬀectiveness the proposed
approach in a critical, although realistic, scenario where
the power provided from the local renewable source is
such that the amount of injected active power for SDG
is high and close to the maximum achievable limit. An
example of such a situation is depicted in Fig. 6 where
the same proﬁle of achievable active power Pr
ifor each
generator is assumed. In the simulations, the disturbance
Time[h]
0 6 12 18 24
Pr [p.u.]
0
0.2
0.4
0.6
0.8
1
i
Fig. 6. Achievable active power Pr
iproﬁle during the
simulation
d(t) representing the timevarying loads has been consid
ered as in Figure 7. The simulations have been performed
by means of the SimPowerSystems ToolboxTM . Moreover,
for the sake of simplicity, signals related to voltages and
powers will be considered in p.u. with respect to Enom and
Pnrespectively.
We have tested the presented CGbased approach in two
diﬀerent setups. The former is denoted as Setup 1 and
it is characterized by these weights on the cost functions
W1a=W1b= 1, W2= 0.01IN,W3=IN. The latter
choice is oriented to exploit reactive power transfer rather
than storage usage in regulating voltage along the lines.
In the second setup, denoted as Setup 2, the considered
optimization weights are W1a=W1b= 1, W2=IN,
W3= 0.01INand are tuned in order to prefer a larger
storage usage along the simulation.
0 6 12 18 24
0
0.1
0.2
0.3
0.4 δI1
δI2
δI3
δI4
δI5
δIev1
0 6 12 18 24
0
0.1
0.2
0.3
0.4
δI6
δI7
δI8
δI9
δI10
δIev2
δIev3
[p.u.][p.u.]
Time[h]
Time[h]
Fig. 7. Timevarying loads d(t)
Under the scenario of Figures 67, if the smart function
alities on the SDG were disabled, the constraints on the
voltage are violated several times along the simulation.
In this respect, Figure 8 depicts the voltage evolutions
on each node without the action of the CG device, when
the generators on the Feeder 1, without any manipulation,
inject the active power according to the generation proﬁles
depicted in Figure 6 and the timevarying loads shown in
Figure 7.
Time[h]
0 5 10 15 20
[p.u.]
0.8
1
1.2
Time[h]
0 5 10 15 20
[p.u.]
0.7
0.8
0.9
1
1.1
E
12
E
22
E
32
E
42
E
52
24
E62
E72
E82
E92
E10
2
Fig. 8. Voltages proﬁle on each node without reactive
power management
0 5 10 15 20
0.9
0.95
1
1.05
1.1
[p.u.]
0 5 10 15 20
0.9
0.95
1
1.05
1.1
[p.u.]
E
12
E 2
E
32
E4
E 2

5
E6
2
E 
E8
2
E92
E10
2
2
2


Time[h]
Time[h]
Fig. 9. Voltage proﬁle on each node with CG actions in
Setup 1 case.
Under the same scenario, a relevant improvement on
the voltage proﬁles can be achieved when the reactive
power management is performed on the SDGs, as depicted
in both Figures 910. The control voltage problem is
successfully solved in both setups, in particular Figures 11
12 highlight the diﬀerent reactive power management on
Node 1 when diﬀerent weights are considered. As expected,
in Setup 1 the storage devices actions never occur and
the voltage is kept within the prescribed bounds thanks
to reactive power absorption. Alternatively, this task is
accomplished in Setup 2 by storing a certain amount of
energy in the local batteries, operation that leads to the
consequent ”safe” reduction of active power injected into
the network.
Preprints of the 20th IFAC World Congress
Toulouse, France, July 914, 2017
10424
0 5 10 15 20
0.9
1
1.1
[p.u]
0 5 10 15 20
0.9
1
1.1
[p.u]
E 1
2
E 2
E32
E 4

E 2

5
E6
2
E 
E8
2
E92
E10
2
2
2


2
7
Time[h]
Time[h]
Fig. 10. Voltage proﬁle on each node with CG actions in
Setup 2 case.
0 10 20
P1(t) [p.u]
0
0.5
1Setup 1
Setup 2
0 10 20
0.6
0.4
0.2
0
Q1(t) [p.u]
Time [h] Time [h]
Fig. 11. Injected/absorbed power on Node 1: (left) Active
Power P1(t), (right) reactive power Q1(t).
0 5 10 15 20
Time [h]
0
0.2
0.4
0.6
0.8
1
SOC
Setup 1
Setup 2
Fig. 12. SOC of the battery on Node 1.
5. CONCLUSIONS
This work has been focused on the voltage control problem
in MV distribution grids in the presence of high penetra
tion of DG and storage facilities.
The presented solution is based on the wellknow Com
mand Governor (CG) approach, here used to design a
supervisory scheme which was demonstrated in previous
works of the authors to be eﬀective in coordinating dis
tributed generators and OLTC transformers in order to
guarantee a safe integration of green power generators
and storage units into the distribution grid in spite of
the variability and intermittent behavior of the renewable
sources.
In the ﬁnal simulation example the CGbased scheme has
shown good capabilities in reactive power management
and a certain degree of ﬂexibility during the design phase.
In fact, it is possible to establish, by properly selecting
optimization weights, if the CG action should prefer a high
reactive power usage instead of a stronger storage facilities
exploitation in controlling voltage along the feeders.
REFERENCES
L.V. Barboza, G.P. Dimuro, R.H.S. Reiser, “Towards
interval analysis of the load uncertainty in power electric
systems”, 8th International Conference on Probabilistic
Methods Applied to Power Systems, Ames, Iowa, 2004,
pp. 538544.
A. Bemporad, A. Casavola and E. Mosca “Nonlinear
control of constrained linear systems via predictive ref
erence management”. IEEE Transaction on Automatic
Control,42(3), 340–349, 1997.
A.Bemporad, A.Casavola and E.Mosca, “A Predictive
Reference Governor for Constrained Control Systems”,
Computers in Industry, Vol. 36, pp. 5564, 1998.
B. Bletterie, A. Gor˘sek, T. Fawzy, D. Premm, W. Deprez,
F. Truyens, A. Woyte, B. Blazi˘c, B. Uljani˘c, “Devel
opment of innovative voltage control for distribution
networks with high photovoltaic penetration”, Progress
in Photovoltaics: Research and Applications, 2012, vol.
20(6), pp. 747–759.
A.Casavola, G. Franz`e, N. Carelli, “ Voltage regulation
in networked electrical power systems for distributed
generation: a constrained supervisory approach”. Pro
ceedings of ”NOLCOS 2007”, Pretoria, South Africa,
2007.
A. Casavola, G. Franz`e, D. Menniti and N. Sorrentino,
“Voltage regulation in distribution networks in the pres
ence of distributed generation: A voltage setpoint re
conﬁguration approach”. Electric Power Systems Re
search 81: (1), January 2011.
A. Casavola, F. Tedesco, M. Vizza, ”Command Governor
Strategies for the Online Management of Reactive Power
in Smart Grids With Distributed Generation.”, IEEE
Transactions on Automation Science and Engineering
(2017).
Comitato Elettrotecnico Italiano Norm CEI016
“Reference technical rules for the connection of
active and passive consumers to the HV and MV
electrical networks of distribution Company (2015)”,
url:http://www.ceiweb.it/it/comunicati/news/569
normacei016.html.
Ferry V. A., Sannino A., Jaap Daalder J., “Voltage control
with onload tap changers in medium voltage feeders
in presence of distributed generation”, Electric power
systems research, 2007, vol. 77(10), pp. 1314–1322.
E.G. Gilbert, I. Kolmanovsky and K. Tin Tan. “Discrete
time Reference Governors and the Nonlinear Control
of Systems with State and Control Constraints”. Int.
Journ. on Robust and Nonlinear Control, 5, pp. 487504,
1995.
L. H. Macedo, J. F. Franco, M. J. Rider, and R. Romero,
“Optimal operation of distribution networks considering
energy storage devices”, IEEE Trans. Smart Grid, 2015,
Vol. 6(6), pp. 2825–2836.
McEvoy A., Markvart T., Casta˜ner L., Practical handbook
of photovoltaics: fundamentals and applications Else
vier, 2003.
Manwell J.F., McGowan J.G., “Lead acid battery storage
model for hybrid energy systems”, Solar Energy, 1993,
vol.50(5), pp. 399–405
T. Reinaldo, L.A.C. Lopes, “Impact of active power cur
tailment on overvoltage prevention and energy produc
tion of PV inverters connected to low voltage residential
feeders”, Renewable Energy, 2011, vol. 36, pp.35663574.
Preprints of the 20th IFAC World Congress
Toulouse, France, July 914, 2017
10425