This article has been accepted for inclusion in a future issue of this journal.
Content is final as presented, with the exception of pagination.
IEEE SYSTEMS JOURNAL
A Nonlinear H-infinity Control Approach to
Stabilization of Distributed Synchronous Generators
Gerasimos Rigatos, Senior Member, IEEE, Pierluigi Siano, Senior Member, IEEE,
Alexey Melkikh, Member, IEEE, and Nikolaos Zervos
Abstract—This paper proposes a new nonlinear H-infinity con-
trol method for stabilization and synchronization of distributed
interconnected synchronous generators. At first stage, local lin-
earization of the distributed generators’ model is performed round
its present operating point. The approximation error that is intro-
duced to the linearized model is due to truncation of higher-order
terms in the performed Taylor series expansion and is represented
as a disturbance. The control problem is now formulated as a min–
max differential game in which the control input tries to minimize
the state vector’s tracking error while the disturbances affecting
the model try to maximize it. Using the linearized description of
the distributed generators’ dynamics, an H-infinity feedback con-
troller is designed through the solution of a Riccati equation at each
step of the control algorithm. The inherent robustness properties
of H-infinity control assure that the disturbance effects will be
eliminated and the state variables of the individual power genera-
tors will converge to the desirable setpoints. The proposed method,
stands for a reliable solution to the problem of nonlinear control
and stabilization for interconnected synchronous generators. It is
also a novel approach, comparing to control of synchronous gener-
ators based on global linearization methods. Its efficiency is further
confirmed through simulation experiments.
Index Terms—Algebraic Riccati equations, asymptotic stabil-
ity, distributed synchronous generators (SG), min–max differential
games, nonlinear h-infinity control.
I. INTRODUCTION
HE rapid expansion and development of the electricity
T grid through the integration of renewable energy sources
has made imperative the development of control methods for
stabilization and synchronization of the distributed power gen-
eration units connected to it. The problem of control of dis-
tributed and interconnected power generators has been studied
in [1]–[6]. Nonlinear decentralized controllers for multimachine
and large-scale electric power systems have been proposed in
[7]–[14]. Control of distributed power generators using global
linearization approaches has been proposed in [15]–[20]. Opti-
mal control and optimization approaches to multimachine power
Manuscript received August 6, 2016; revised October 31, 2016, January 6,
2017, and February 24, 2017; accepted March 17, 2017. (Corresponding author:
Gerasimos Rigatos.)
G. Rigtos and N. Zervos are with the Unit of Industrial Automation, Indus-
trial Systems Institute, Rion Patras 26504, Greece (e-mail: grigat@ieee.org;
nzervos@isi.gr).
P. Siano is with the Department of Industrial Engineering, University of
Salerno, Fisciano 84084, Italy (e-mail: psiano@unisa.it).
A. Melkikh is with the Institute of Physics and Technology, Ural Federal
University, Yekaterinburg 620002, Russia (e-mail: melkikh2008@rambler.ru).
Digital Object Identifier 10.1109/JSYST.2017.2688422
1937-9234 © 2017 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.
1
systems (MMPSs) stabilization have been proposed in [21]–
[25]. The problem of robust control of synchronous generators
(SG), particularly with the use of H-infinity control methods has
been analyzed in [26]–[31].
In this paper, a new nonlinear H-infinity optimal control
method for distributed SGs is developed and specific results
are obtained for the problem of synchronization and stabiliza-
tion of a three-area MMPS. The nonlinear dynamical model
of the interconnected SGs undergoes linearization round tem-
porary operating points which are computed at each iteration
of the control algorithm. These operating points consist of the
present value of the system’s state vector and of the last value of
the control inputs vector that was exerted on it. This linearization
procedure is based on the concept of Taylor series expansion of
the system’s state-space model and on the computation of the
associated Jacobian matrices [32]–[34]. The modeling error due
to this approximate linearization is considered to be a perturba-
tion that is compensated by the robustness of the control scheme
[28], [29].
The feedback control input is based on the solution of an H-
infinity control problem [28], [29]. This stands for a min–max
differential game between the control signal and the disturbance
terms that affect the model of the distributed generators. The
feedback control gain is found from the solution of an alge-
braic Riccati equation at each iteration of the control algorithm
[28], [29]. The stability of the control scheme is demonstrated
through Lyapunov analysis. Actually, it is shown that the con-
trol loop satisfies an H-infinity tracking performance criterion
which signifies elevated robustness against model uncertainty
and perturbation inputs. Besides, under moderate conditions the
asymptotic stability of the control scheme is proven. Through
simulation experiments, it is further confirmed that the proposed
control method assures fast and accurate tracking of reference
setpoints.
The advantages of the control method are outlined as follows:
1) It is applied directly on the nonlinear dynamical model
of the distributed power generation system and not on an
equivalent linearized description of it.
2) It avoids the elaborated linearizing transformations (dif-
feomorphisms) which can be met in global linearization-
based control methods for distributed power generation
systems.
3) The controller is designed according to optimal control
principles which implies the best tradeoff between precise
tracking of the reference setpoints on the one side and
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE SYSTEMS JOURNAL

moderate variations of the control inputs on the other

side.
4) The control method exhibits significant robustness to para-
metric uncertainty, modeling errors as well as to external
perturbations.
5) The computational implementation of the control method
is simple since it requires only the solution of an algebraic
Riccati equation.
This paper is organized as follows. In Section II, the dynamic
model of the multimachine power system is analyzed and the
state-space model of the distributed power generators is ob-
tained. In Section III, linearization of the model of the distributed
SGs is performed by applying Taylor series expansion and by
computing the associated Jacobian matrices. In Section IV,
it is explained how H-infinity control can be applied to the
linearized model of the distributed power system through the
solution of an algebraic Riccati equation at each iteration of
the control algorithm. In Section V, the stability of the control
scheme is proven through Lyapunov analysis. In Section VI,
it is explained how robust estimation of the state vector of the
Fig. 1. Three-area distributed power generation model consisting of three
model of the distributed power generators can be performed with synchronous machines.
the use of the H-infinity Kalman Filter. In Section VII, simula-
tion experiments are performed to further confirm the stability where
and synchronization features of the proposed control scheme.
f1 (x) = x2 − ω0 (2)
Finally, in Section VIII concluding remarks are stated.
D1 Pm 1
II. DYNAMIC MODEL OF THE MMPS f2 (x) = − (x2 − ω0 ) + ω0
2J1 2J1
A. State-Space Description of the Distributed SGs Model ω0
− {G11 x23 + x3 [x6 G12 sin(x1 − x4 − a12 )
The dynamics of the mechanical and the electrical part, both 2J1
for the synchronous power generator in stand-alone functioning + x9 G13 sin(x1 − x7 − a13 )]
mode and for the model of the distributed interconnected SGs (3)
are described in Appendix A. Next, the state-space description 
of the model of the distributed SG is outlined. 1 1 xd 1 − xd 1
f3 (x) = −  x3 +  Vs cos(x1 ) (4)
The state variables of each SG are: x1 the turn angle of the Td 1 Td o 1 xd Σ
1
rotor denoted as Δδ, x2 the difference of the rotor’s angular
speed from the synchronous speed Δω = ω − ω0 , and x3 the f4 (x) = x5 − ω0 (5)

quadrature-axis transient voltage of the generator Eq which is D2 Pm 2
actually a variable representing the stator’s magnetic flux. This f5 (x) = − (x5 − ω0 ) + ω0
2J2 2J2
is explained in (46), (47), and (48), given in Appendix A.
ω0
Without loss of generality, the three-machine power system − {G22 x26 + x6 [x3 G21 sin(x4 − x1 − a21 )
of Fig. 1 is considered. For this power system, the state vector 2J2
x]inR9×1 : is x = [x11 , x12 , x13 , x21 , x22 , x23 , x31 , x32 , x33 ]T , where xji + x9 G23 sin(x4 − x7 − a23 )]
denotes the ith state variable of generator j. Denoting as ω0 , (6)
the synchronous angular velocity and using the definition of 

variables for the model of the SGs that is given in Appendix, 1 1 xd 2 − xd 2

f6 (x) = −  x6 +  Vs cos(x4 ) (7)
the system of the three distributed and interacting SGs can be Td 2 Td o 2 xd Σ
2
written in the matrix form
⎛ ⎞ f7 (x) = x8 − ω0 (8)
ẋ1 ⎛ ⎞ ⎛ ⎞
⎜ ẋ2 ⎟ f1 (x) 0 0 0 D2 Pm 3
⎜ ⎟ ⎜ f2 (x) ⎟ ⎜ 0 0 0 ⎟ f8 (x) = − (x8 − ω0 ) + ω0
⎜ ẋ3 ⎟ ⎜ ⎟ ⎜ ⎟ 2J3 2J3
⎜ ⎟ ⎜ f3 (x) ⎟ ⎜ g1 (x) 0 0 ⎟
⎜ ẋ4 ⎟ ⎜ ⎟ ⎜ ⎟⎛ ⎞ ω0
⎜ ⎟ ⎜ f4 (x) ⎟ ⎜ 0 0 0 ⎟ − 2
{G32 x9 + x9 [x3 G31 sin(x7 − x1 − a31 )
⎜ ẋ5 ⎟ ⎜ ⎟ ⎜ ⎟ u1 2J3
⎜ ⎟ = ⎜ f5 (x) ⎟ + ⎜ 0 0 0 ⎟ ⎝ u2 ⎠ (1)
⎜ ẋ6 ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ f6 (x) ⎟ ⎜ 0 g2 (x) 0 ⎟ u3 + x6 G32 sin(x7 − x4 − a32 )]
⎜ ẋ6 ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ f7 (x) ⎟ ⎜ 0 0 0 ⎟ (9)
⎜ ẋ7 ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎝ f8 (x) ⎠ ⎝ 0 0 0 ⎠
⎝ ẋ8 ⎠ 1 1 xd 3 − xd 3


f9 (x) 0 0 g3 (x) f9 (x) = −  x9 + Vs cos(x7 ) (10)

ẋ9 Td 3 Td o 3

xd Σ
3
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

RIGATOS et al.: NONLINEAR H-INFINITY CONTROL APPROACH TO STABILIZATION OF DISTRIBUTED SYNCHRONOUS GENERATORS 3

while one has that sin(x4 − x7 − a23 )]}, ∂ x 7 = 2J 2 x6 x9 G23 cos(x4 − x7 −

∂ f5 ω0

1 1 1 a23 ), ∂ x 8 = 0, ∂ x 9 = − 2J 2 x6 G23 sin(x4 − x7 − a23 ).

∂ f5 ∂ f5 ω0
g1 = g2 = g3 = u3 . (11)
Td o 1 Td o 2 Td o 3 For the sixth row of the Jacobian matrix A =
∇x [f (x) + g(x)u], one has: ∂∂ xf 61 = 0, ∂∂ xf 62 = 0, ∂∂ xf 63 =, ∂∂ xf 64 =

III. LINEARIZATION OF THE MODEL OF THE DISTRIBUTED SGS x d 2 −x d
− T d1 
xd
2
Vs sin(x4 ), ∂ f6
∂ x5 = 0, ∂ f6
∂ x6 = − T 1 , ∂ f6
∂ x7 = 0,
o2 d2
Local linearization is performed for the state-space model Σ 2

of the distributed power generators, round the operating point

∂ f6
∂ x8 = 0, ∂ x 9 = 0.
∂ f6

(x∗ , u∗ ) where x∗ is the present value of the system’s state vector For the seventh row of the Jacobian matrix A = ∇x [f (x) +
and u∗ is the last value of the control input that was exerted on g(x)u], one has: ∂∂ xf 71 = 0, ∂∂ xf 72 = 0, ∂∂ xf 73 = 0, ∂∂ xf 74 = 0, ∂∂ xf 75 = 0,
∂ x 6 = 0, ∂ x 7 = 0, ∂ x 8 = 1, ∂ x 9 = 0.
∂ f7 ∂ f7 ∂ f7 ∂ f7
the machine. Thus, one obtains the linearized description
For the eight row of the Jacobian matrix A =
˜
ẋ = Ax + Bu + d. (12) ∇x [f (x) + g(x)u] one has: ∂∂ xf 81 = 2J ω0
3
x9 [x3 G31 cos(x7 −
For the previous description of the distributed power generators’ x1 − a31 ) + x6 G32 cos(x7 − x4 − a32 )], ∂∂ xf 82 = 0, ∂∂ xf 83 =
model by the state-space equation of (1), it holds that A = − 2Jω0
x9 G31 sin(x7 − x1 − a31 ), ∂∂ xf 84 = 2J ω0
x9 x6 G32 cos(x7 −
3 3
∇x [f (x) + g(x)u] with x4 − a32 ), ∂ x 5 = 0, ∂ x 6 = − 2J 3 x9 G23 sin(x7 − x4 − a32 ),
∂ f8 ∂ f8 ω0

∂ x 7 = − 2J 3 x9 [x3 G31 cos(x7 − x1 − a31 ) + x6 G32 cos(x7 −

∂ f8
A = ∇x [f (x) + g(x)u] ω0

⎛ ∂f ⎞ x4 − a32 )], ∂∂ xf 88 = − 2J ∂ x 9 = − 2J 3 {G33 2x9 + [x3 G31

D3 ∂ f8 ω0
∂f 1 ∂f 1 1
··· ∂ f1 3,
⎜ ∂ x1 ∂ x2 ∂ x3 ∂ x9
⎟ sin(x7 − x1 − a31 ) + x6 G32 sin(x7 − x4 − a32 )]}.
⎜ ∂ f2 ∂ f2 ∂ f2
··· ∂ f2⎟ Moreover, it holds that B = ∇u [f (x) + g(x)u] with
⎜ ∂ x1 ∂ x2 ∂ x3 ∂ x9⎟
⎜ ⎟
⎜ ∂ f3 ∂ f3 ∂ f3
··· ∂ f3⎟
⎜ ∂ x1 ∂ x2 ∂ x3 ∂ x9⎟ B = ∇u [f (x) + g(x)u]
⎜ ⎟ ⎛ ⎞
= ⎜ ∂ f4 ∂ f4 ∂ f4
··· ∂ f4
⎟. (13)
⎜ ∂ x1 ∂ x2 ∂ x3 ∂ x9⎟ 0 0 0
⎜ ··· ··· ⎟ ⎜ 0
⎜ ··· ··· ··· ⎟ ⎜ 0 0 ⎟⎟
⎜ ⎟ ⎜ g1 (x) 0
⎜ ··· ··· ··· ··· ··· ⎟ ⎜ 0 ⎟⎟
⎝ ⎠ ⎜ 0
⎜ 0 0 ⎟⎟
∂ f9 ∂ f9 ∂ f9
··· ∂ f9
= ⎜ 0 ⎟
∂ x1 ∂ x2 ∂ x3 ∂ x9
⎜ 0 0 ⎟. (14)
⎜ 0 g2 (x) 0 ⎟
For the first row of the Jacobian matrix A = ∇x [f (x) + g(x)u], ⎜ ⎟
⎜ 0 0 0 ⎟
one has: ∂∂ xf 11 = 0, ∂∂ xf 12 = 1, ∂∂ xf 13 = 0, ∂∂ xf 14 = 0, ∂∂ xf 15 = 0, ∂∂ xf 16 = 0, ⎜ ⎟
⎝ 0 0 0 ⎠
∂ x 7 = 0, ∂ x 8 = 0, ∂ x 9 = 0.
∂ f1 ∂ f1 ∂ f1
0 0 g3 (x)
For the second row of the Jacobian matrix A =
∇x [f (x) + g(x)u], one has: ∂∂ xf 21 = − 2J ω0
x3 [x6 G12 cos(x1 −
1 IV. NONLINEAR H-INFINITY CONTROL
x4 − a13 )] + x9 G13 cos(x1 − x7 − a13 ), ∂∂ xf 22 = − 2J D1
1
, ∂∂ xf 23 =
−ω 0 The initial nonlinear model of the distributed power genera-
2J 1 {G11 2x3 + [x6 ]G12 sin(x1 − x4 − a12 ) + x9 G13 sin(x1 − tors is in the form
x7 − a13 )]}, ∂∂ xf 24 = 2J ω0
1
x3 x6 G12 cos(x1 − x4 − a12), ∂∂ xf 25 = 0,
ẋ = f (x, u) x∈Rn , u∈Rm . (15)
∂ x 6 = − 2J 1 x3 G12 sin(x1 − x4 − a12 ), ∂ x 7 = 2J 1 x3 x9 G13
∂ f2 ω0 ∂ f2 ω0

cos(x1 − x7 − a13 ), ∂∂ xf 28 = 0, ∂∂ xf 29 = − 2J ω0
1
x3 G13 sin(x1 − Linearization of the system (distributed power generators) is
x7 − a13 ). performed at each iteration of the control algorithm round its
For the third row of the Jacobian matrix A = ∇x [f (x) + present operating point (x∗ , u∗ ) = (x(t), u(t − Ts )). The lin-

x d 1 −x d earized equivalent model of the system is described by
g(x)u], one has: ∂ f3
∂ x1 = − T d1 
xd
1
Vs sin(x1 ), ∂ f3
∂ x2 = 0,
ẋ = Ax + Bu + Ld˜ x ∈ Rn , u ∈ Rm , d˜∈ Rq
o1
Σ 1
(16)
∂ f3
∂ x3 = − T 1 , ∂ f3
∂ x4 = 0, ∂ f3
∂ x5 = 0, ∂ f3
∂ x6 = 0, ∂ f3
∂ x7 = 0, ∂ f3
∂ x8 = 0,
d1
where matrices A and B are obtained from the computation of
∂ f3
∂ x9 = 0. the previously defined Jacobians and vector d˜ denotes distur-
For the fourth row of the Jacobian matrix A = ∇x [f (x) + bance terms due to linearization errors. The problem of distur-
g(x)u], one has: ∂∂ xf 41 = 0, ∂∂ xf 42 = 0, ∂∂ xf 43 = 0, ∂∂ xf 44 = 0, ∂∂ xf 45 = 1, bance rejection for the linearized model that is described by
∂ x 6 = 0, ∂ x 7 = 0, ∂ x 8 = 0, ∂ x 9 = 0.
∂ f4 ∂ f1 ∂ f1 ∂ f1

For the fifth row of the Jacobian matrix A = ẋ = Ax + Bu + Ld˜

∇x [f (x) + g(x)u], one has: ∂∂ xf 51 = 2J ω0
2 x6 [x3 G21 cos(x4 − y = Cx (17)
x1 − a21 ) + x3 G23 cos(x4 − x7 − a23 )], ∂∂ xf 52 = 0, ∂∂xf‘ 5‘3 =
where x ∈ Rn , u ∈ Rm , d˜∈ Rq , and y ∈ Rp , cannot be handled
− 2Jω0
x6 G21 sin(x4 − x1 − a21 ), ∂ x 4 = − 2J 2 x6 [x3 G21 cos
∂ f5 ω0
2 efficiently if the classical LQR control scheme is applied. This
(x4 − x1 − a21 ) + x9 G23 cos(x4 − x7 − a23 )], ∂∂ xf 55 = − 2J D2
2
, is because of the existence of the perturbation term d. ˜ The
∂ f5
= − ω0
{G22 2x 6 + [x G
3 21 sin(x 4 − x1 − a21 ) + x G
9 23
˜
disturbance term d apart from modeling (parametric) uncertainty
∂ x6 2J 2
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE SYSTEMS JOURNAL

tive control (MPC) or Nonlinear MPC (NMPC) have not been

equally explored [35]–[37]. For instance, it is acknowledged
that MPC has been primarily developed for linear dynamical
systems and its application to the nonlinear distributed power
grid lacks a global asymptotic stability proof and risks a genera-
tors’ tripping. On the other side, NMPC is based on an iterative
search for the optimal solution that is not always of assured
convergence while its performance is often dependent on ini-
tial parameterization. Consequently, the stability properties of
NMPC cannot be globally assured either.

V. LYAPUNOV STABILITY ANALYSIS

Through Lyapunov stability analysis it will be shown that the
proposed nonlinear control scheme assures H∞ tracking per-
Fig. 2. Diagram of the control scheme for the distributed SGs.
formance for the distributed SGs, and that in case of bounded
disturbance terms asymptotic convergence to the reference set-
points is succeeded.
and external perturbation terms can also represent noise terms
The tracking error dynamics for the distributed SGs is written
of any distribution.
in the form
In the H∞ control approach, a feedback control scheme is
designed for trajectory tracking by the system’s state vector and ė = Ae + Bu + Ld˜ (21)
simultaneous disturbance rejection, considering that the distur-
bance affects the system in the worst possible manner. The ef- where in the distributed SGs case L = I∈R9×9 with I being
fects that disturbances have on the distributed power generators the identity matrix. Variable d˜ denotes model uncertainties and
are incorporated in the following quadratic cost function: external disturbances of the motor’s model. The following Lya-
 punov equation is considered:
1 T T
J(t) = [y (t)y(t) + ruT (t)u(t)
2 0 1 T
V = e Pe (22)
2
˜
− ρ2 d˜T (t)d(t)]dt, r, ρ > 0. (18)
where e = x − xd is the tracking error. By differentiating with
Equation (18) denotes a min–max differential game taking place
respect to time, and by using the relation about the tracking error
between disturbance and control inputs. Actually, the control
dynamics given in Appendix C, one obtains
inputs try to minimize this cost function while the disturbance
inputs try to maximize it. Then, the optimal feedback control 1 T 1
V̇ = ė P e + eP ėT ⇒
law is given by (see Fig. 2) 2 2
u(t) = −Kx(t) (19) 1 ˜ T P + 1 eT P [Ae + Bu + Ld]
˜⇒
V̇ = [Ae + Bu + Ld]
2 2
with K = 1r B T P , while P is a positive semidefinite symmet- (23)
ric matrix which is obtained from the solution of the Riccati
1 T T
equation of the form V̇ = [e A + uT B T + d˜T LT ]P e

2
1 1 1
A P + PA + Q − P
T
BB − 2 LL
T T
P = 0. (20) + eT P [Ae + Bu + Ld] ˜⇒
r 2ρ 2
(24)
The transients of the control algorithm are determined by matrix 1 1 1
Q and also by gains r and ρ. The latter gain is the H-infinity V̇ = eT AT P e + uT B T P e + d˜T LT P e
2 2 2
attenuation coefficient and its minimum value that allows solu- 1 1 1
tion of (20) is the one that provides maximum robustness to the + eT P Ae + eT P Bu + eT P Ld. ˜ (25)
2 2 2
control algorithm.
The diagram of the considered control loop is depicted in The previous equation is rewritten as
Fig. 2. It is noteworthy that the proposed control method pro-

1 1 T T 1
vides a reliable and computational efficient solution to the non- V̇ = eT (AT P + P A)e + u B P e + eT P Bu
linear optimal control problem for distributed power generators, 2 2 2

which has not been solved in a conclusive manner up to now. The 1 ˜T T 1

computation of the feedback control law that is exerted on the + d L P e + eT P Ld˜ . (26)
2 2
power system is obtained following Bellman’s optimality prin-
ciple. Other approaches to solve the nonlinear optimal control Assumption: For given positive definite matrix Q and coeffi-
problem for the distributed power grid, such as model predic- cients r and ρ, there exists a positive definite matrix P , which
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

RIGATOS et al.: NONLINEAR H-INFINITY CONTROL APPROACH TO STABILIZATION OF DISTRIBUTED SYNCHRONOUS GENERATORS 5

is the solution of the following matrix equation: Equation (36) shows that the H∞ tracking performance criterion

is satisfied. The integration of V̇ from 0 to T gives
2 1  T   T
AT P + P A = −Q + P BB T − 2 LLT P. (27) 1 T 1
r ρ V̇ (t)dt ≤ − ||e||2Q dt + ρ2 ˜ 2 dt
||d||
0 2 0 2 0
Moreover, the following feedback control law is applied to the  T  T
system: ⇒ 2V (T ) + ||e||2Q dt ≤ 2V (0) + ρ2 ˜ 2 dt. (37)
||d||
0 0
1
u = − B T P e. (28) Moreover, if there exists a positive constant Md > 0, such that
r  ∞
˜ 2 dt ≤ Md
||d|| (38)
By substituting (27) and (28), one obtains
0

1 T 2 1 then one gets

V̇ = e [−Q + P BB − 2 LL
T T
P ]e 
2 r 2ρ ∞

||e||2Q dt ≤ 2V (0) + ρ2 Md . (39)
1 0
+ eT P B − B T P e + eT P Ld˜ ⇒ (29) ∞
r Thus, the integral 0 ||e||2Q dt is bounded. Moreover, V (T ) is

1 2 1 bounded and from the definition of the Lyapunov function V

V̇ = − eT Qe + P BB T P e − 2 eT P LLT P e in (22), it becomes clear that e(t) will be also bounded since
2 r 2ρ

e(t) ∈ Ωe = {e|eT P e≤2V (0) + ρ2 Md }.
1 T
− e P BB T P e + eT P Ld˜ (30) According to the above and with the use of Barbalat’s Lemma,
r one obtains limt→∞ e(t) = 0.
which after intermediate operations gives
VI. ROBUST STATE ESTIMATION WITH THE USE
1 1 OF THE H∞ KALMAN FILTER
V̇ = − eT Qe − 2 eT P LLT P e + eT P Ld˜ (31)
2 2ρ The control loop has to be implemented with the use of infor-
mation provided by a small number of sensors and by processing
or, equivalently,
only a small number of state variables. To reconstruct the miss-
1 1 ing information about the state vector of the distributed SGs, it is
V̇ = − eT Qe − 2 eT P LLT P e proposed to use a filtering scheme and based on it to apply state
2 2ρ
estimation-based control [20]. The recursion of the H∞ Kalman
1 1
+ eT P Ld˜ + d˜T LT P e. (32) Filter, for the model of the distributed SGs, can be formulated
2 2 in terms of a measurement update and a time update part
Lemma: The following inequality holds: Measurement update:

1 T ˜ 1˜ T 1 1 D(k) = [I − θW (k)P − (k) + C T (k)R(k)−1 C(k)P − (k)]−1

e Ld + dL P e − 2 eT P LLT P e ≤ ρ2 d˜T d.
˜ (33)
2 2 2ρ 2 K(k) = P − (k)D(k)C T (k)R(k)−1
Proof: The binomial (ρα − ρ1 b)2 is considered. Expanding x̂(k) = x̂− (k) + K(k)[y(k) − C x̂− (k)]. (40)
the left part of the above inequality, one gets Time update:
1 1 1 x̂− (k + 1) = A(k)x(k) + B(k)u(k)
ρ2 a2 + 2 b2 − 2ab ≥ 0 ⇒ ρ2 a2 + 2 b2 − ab ≥ 0
ρ 2 2ρ
P − (k + 1) = A(k)P − (k)D(k)AT (k) + Q(k) (41)
1 1 1 1 1 1
⇒ ab − 2 b2 ≤ ρ2 a2 ⇒ ab + ab − 2 b2 ≤ ρ2 a2 .
2ρ 2 2 2 2ρ 2 where it is assumed that parameter θ is sufficiently small
(34) to assure that the covariance matrix P − (k)−1 − θW (k) +
C T (k)R(k)−1 C(k) will be positive definite. When θ = 0, the
The following substitutions are carried out: a = d˜ and b = H∞ Kalman Filter becomes equivalent to the standard Kalman
eT P L and the previous relation becomes: Filter. One can measure only a part of the state vector of the dis-
tributed SGs, such as state variables xi = θi , i = 1, 2, 3 and can
1 ˜T T 1 1 1
d L P e + eT P Ld˜ − 2 eT P LLT P e ≤ ρ2 d˜T d.
˜ (35) estimate through filtering the rest of the state vector elements.
2 2 2ρ 2
VII. SIMULATION TESTS
Equation (35) is substituted in (32) and the inequality is en-
forced, thus giving A. Control of an Independently Functioning SG
1 1 Indicative results about the control of an independently func-
V̇ ≤ − eT Qe + ρ2 d˜T d.
˜ (36) tioning SG under the proposed nonlinear H-infinity control
2 2
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE SYSTEMS JOURNAL
Fig. 3. Rotational speeds ω of the SG and its control input: (a) when tracking
setpoint 1 and (b) when tracking setpoint 2.
Fig. 4. Tracking of reference setpoints by the rotational speeds ω i i = 1, 2, 3
of the distributed SGs: (a) setpoint 1 (p.u) and (b) setpoint 2 (p.u).
Fig. 5. Tracking of reference setpoints by the rotational speeds ω i i = 1, 2, 3
of the distributed SGs: (a) setpoint 3 (p.u) and (b) setpoint 4 (p.u).
method are given in Fig. 3. It has been confirmed that fast and
accurate tracking of the reference setpoints has been achieved.
For the implementation of the control method, the Jacobian ma-
trices of the state-space model of the power generator given
in (45) has been performed. The feedback control signal is
computed from the solution of the Riccati equation described
in (27).
B. Control of the Distributed SGs
The problem of synchronization and stabilization of the dis-
tributed SGs has been solved with the application of a nonlinear
H-infinity (optimal) control method. The performance of the
control method is evaluated through simulation experiments.
It can be observed that the control scheme assures fast and
accurate convergence of the rotation speeds of the individual
generators to the reference setpoints, as shown in Figs. 4–6. As
Fig. 6. Tracking of reference setpoints by the rotational speeds ω i i = 1, 2, 3
of the distributed SGs: (a) setpoint 5 (p.u) and (b) setpoint 6 (p.u).
TABLE I
RMSE OF THE POWER GENERATOTS TURN SPEED
test1 test2 test3 test4 test5 test6
ω1 3·10 −4 2·10 −4 2·10 −4 4·10 −4 4·10 −4 5·10 −4
ω2 3·10 −4 2·10 −4 2·10 −4 4·10 −4 4·10 −4 5·10 −4
ω3 3·10 −4 2·10 −4 2·10 −4 4·10 −4 4·10 −4 5·10 −4
it happens with all optimal control methods, the transients of the
state variables are affected by values of the controller’s gains,
that is r, ρ, and by values of the elements of matrix Q appearing
in (20). The values of these parameters affect also the variation
ranges of the control inputs. It can be noted that the H-infinity
algorithm exhibited remarkable robustness to uncertainty in
the model of the distributed power generators which is due to
approximate linearization. It also remained robust to measure-
ment noise and to external perturbations appearing in the model
of the distributed generators (e.g., variations in the mechanical
torque that excites the rotational motion of the machines).
The computational aspects of the proposed control method
have been also assessed. For the three-area MMPS, the dimen-
sion of the state vector was x ∈ R9 , while the dimension of
the feedback gain matrix P that was obtained from the solution
of the Riccati equation given in (20) was P ∈ R9×9 . Using an
i7 processor at 2.0 Ghz, the solution of the Riccati equation
was performed in milliseconds and definitely in time that was
significantly smaller than the sampling period of the control al-
gorithm. For a distributed power system, that comprises more
generators and which is of higher dimensionality, the real-time
implementation of the control algorithm is not going to be hin-
dered. It can be concluded that the proposed H-infinity control
method is computationally very efficient. The tracking accuracy
of the control method, with reference to the previously given
setpoints 1–6 is given in Table I.
The robustness of the control method to parametric variations
is shown in Table II. Actually, the change of the generators’ turn
speed RMSE is recorded as a result of the % change of parameter
Td o . It is noted that parameter Td o is the one that determines
the input gain matrix B in the generators’ linearized model. It
can be noticed that the RMSE is not significantly affected by
parametric changes thus the control scheme’s robustness can be
noticed.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
RIGATOS et al.: NONLINEAR H-INFINITY CONTROL APPROACH TO STABILIZATION OF DISTRIBUTED SYNCHRONOUS GENERATORS
TABLE II
ROBUSTNESS TO PARAMETRIC CHANGES
ΔTd o % 50 60 65 70 75
RMSE ω ·10 −4 3 4 5 6 7 8
Remark 1: Previous results on the problem of control of dis-
tributed power generators can be classified as follows.
1) Global linearization methods which apply a change of
state-variables (diffeomorphims) so as to arrive at an
equivalent linearized form of the system, the control prob-
lem is solved for the equivalent linearized model and next
to compute the control input that should be finally ex-
erted on the nonlinear system, one has to apply an inverse
transformation.
2) Approximate linearization methods which are based on
the decomposition of the model of the distributed power
generators into multiple local linear models, each one
remaining valid around a local operating point.
3) Empirical methods which assume linear dynamics for the
power generators or which apply heuristically tuned PID
controllers.
With reference with the aforementioned approaches the pro-
posed nonlinear H-infinity control method exhibits specific ad-
vantages.
1) It avoids the complicated transformations of global lin-
earization methods.
2) It avoids the computational burden of the solution of the
control problem for multimodel descriptions of the power
generation systems. In the latter approach, one has either
to solve multiple Riccati equations or must solve LMIs.
3) Unlike empirical methods, the nonlinear H-infinity control
scheme is of assured stability.
Remark 2: The previously analyzed control method is a
novel solution for the nonlinear optimal control problem un-
der model uncertainty and external perturbations. In contrast
to other attempts to solve the nonlinear optimal control prob-
lem through empirical initialization and iterative search of the
optimal solution (as in the case of the NMPC), the previous
control method is of assured and proven stability and exhibits
also very fast convergence to the optimal solution. The stages
of implementation of the H-infinity control method are clear.
The computational burden of the method is limited because
it requires only the repetitive solution of an algebraic Riccati
equation. Moreover, heuristics are avoided.
Remark 3: A challenge for optimal control methods for dis-
tributed renewable energy systems remains the proof of global
asymptotic stability and much research work has been done on
that [38]–[39]. The stability properties about the proposed non-
linear optimal control method is generic since it considers that
the state vector of the distributed power generation system can
be of any dimension. The presented nonlinear H-infinity control
method for the model of the distributed power generators is scal-
able and can be applied not only to the three-area three-machine
power system model but can be also generalized to the case of
7
other multiarea multimachine power generation schemes. Each
time a new SG is included in the distributed power generation
model, the order of its state-space description is increased by 3.
80
For the model of the three interconnected generators, and when
running the control software under moderate computing power,
it has been confirmed that the cycle time is a fraction of the sam-
pling period. This is a clear indication that the proposed control
algorithm can work efficiently even in the case of distributed
power generation systems of high dimensionality. The imple-
mentation of the control software in a real-time programming
environment and with the use of more powerful processors can
further improve the computational performance of the control
method.
VIII. CONCLUSION
In this paper, a new nonlinear optimal control method has been
developed for the model of distributed synchronous generators.
First, the dynamic model of the interconnected generators is
subjected to linearization round local operating points which
are defined by the present value of the state vector and the last
value of the control inputs vector that was exerted on the system.
This linearization makes use of Taylor series expansion and of
the computation of the associated Jacobian matrices. On the
linearized description of the system, an H-infinity controller is
applied. The feedback control gain is found by solving at each
iteration of the control algorithm an algebraic Riccati equation.
The stability of the H-infinity control scheme is proven
through Lyapunov analysis. It is shown that the control loop
satisfies the criterion of the H-infinity tracking performance and
this signifies elevated robustness against model uncertainty and
external perturbations. Moreover, under moderate conditions,
the asymptotic stability of the control loop is proven. Through
simulation experiments, it has been further confirmed that the H-
infinity control results in fast and accurate tracking of reference
setpoints by the state variables of the model of the distributed
power generators.
APPENDIX
A. Dynamics of the Synchronous Power Generator
It is considered that the third-order model of the synchronous
generator (SG) is connected to the power grid thus forming the
model of a Single Machine Infinite Bus (SMIB) system. The
PMSG mechanical dynamics can be represented as follows:
δ̇ = ω
D ω0
ω̇ = − (ω − ω0 ) + (Pm − Pe ) (42)
2J 2J
where δ is the turn angle of the generator’s rotor, ω is the rotation
speed of the rotor with respect to synchronous reference frame,
and ω0 is the synchronous speed of the generator (see Fig. 7).
Moreover, J is the moment of inertia of the rotor, Pe is the active
power of the generator, Pm is the mechanical input torque to the
generator which is associated with the mechanical input power,
D is the damping constant of the generator, and Te is the elec-
trical torque which is associated to the generated active power.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE SYSTEMS JOURNAL

of the SMIB model is obtained

δ̇ = ω − ω0

D Pm 1 V s Eq
ω̇ = − (ω − ω0 ) + ω0 − ω0  sin(Δδ)
2J 2J 2J xd Σ

 1  1 xd − xd 1
Ėq = −  Eq +  Vs cos(Δδ) + Ef (45)
Td Td o xd Σ Td o

 x
where Td = x dd Σ Td o is the time constant of the field winding.
Σ
The previously analyzed SMIB model of the PMSG is described
by a nonlinear state space model of the form
Fig. 7. Rotation frame for the magnetic field of the SG. ẋ = f (x) + g(x)u (46)
 T
where the state vector x is defined as x = Δδ Δω Eq . The
vector fields f (x) and g(x) are defined as
Additionally, the following variables are defined: Δδ = δ − δ0 ⎛ ⎞
and Δω = ω − ω0 with ω0 denoting the synchronous speed. The ω − ω0

⎜ D 1 Vs Eq ⎟
generator’s electrical dynamics is described as follows [20]: f (x) = ⎜ − (ω − ω0 ) + ω0 P2Jm − ω0 2J  sin(Δδ) ⎟
⎝ 2J 
xd Σ ⎠
 x −x
1 − T1 Eq + T1d o dx  d Vs cos(Δδ)

Ėq = (Ef − Eq ) (43) d dΣ
(47)
Td o
1
T
 g(x) = 0 0 Td o (48)
where Eq is the quadrature-axis transient voltage of the gener-
ator, Eq is the quadrature axis voltage of the generator, Td o is with control input u = Ef being the field voltage (equivalent
the direct axis open-circuit transient time constant of the gener- voltage in the excitation coil) and measurable output the turn
ator, and Ef is the equivalent voltage in the excitation coil. The angle of the rotor
algebraic equations of the SG are given by y = h(x) = δ − δ0 . (49)

xd Σ   Vs The rotation frame for the magnetic field of the SG is depicted

Eq =  E − (xd − xd )  cos(Δδ) in Fig. 7.
xd Σ q xd Σ
Vs B. Dynamic Model of Distributed SGs
Iq =  sin(Δδ)
xd Σ A MMPS with n SGs, in which the first machine is chosen
Eq
 as the reference machine can be described by the following
Vs
Id =  −  cos(Δδ) nonlinear dynamic model [20], [30]:
xd Σ xd Σ

δ̇i = ωi − ω0
V s Eq
Pe =  sin(Δδ) Di Pm i 1
xd Σ ω̇i = − (ωi − ω0 ) + ω0 − ω0
2Ji 2Ji 2Ji

V s Eq V2 
n
Qe =  cos(Δδ) − s 
· [Gii Eq i + Eq i
2  
Eq j Gij sin(δi − δj − αij )]
xd Σ xd Σ
 j =1,j = i
 
Vt = (Eq − Xd Id )2 + (Xd Iq )2 (44)  1  1 xd i − xd i 1


Ėq i = −  Eq i +  Vs i cos(Δδi ) + Ef i
Td i Td o i xd Σ i Td o i
 
where xd Σ = xd + xT + xL , xd Σ = xd + xT + xL , xd is the (50)
direct-axis synchronous reactance, xT is the reactance of the
 where the electric torque Pe i , which is associated with the active
transformer, xd is the direct axis transient reactance, xL is
power at the ith generator, is now given by
the reactance of the transmission line, Id and Iq are direct and
quadrature axis currents of the generator, Vs is the infinite bus  2 

n


voltage, Qe is the generator reactive power delivered to the Pe i = Gii Eq i + Eq i Eq j Gij sin(δi − δj − αij ) (51)
j =1,j = i
infinite bus, and Vt is the terminal voltage of the generator. Sub-
stituting the electrical equations of the PMSG given in (44) into for i = 1, 2, . . . , n. For a power grid that consists of n gen-
the equations of the electrical and mechanical dynamics of the erators, the aggregate state vector comprises the state vec-
rotor given in (42) and (43), respectively, the complete model tors of the local machines, i.e., x = [x1 , x2 , . . . , xn ]T , where
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

RIGATOS et al.: NONLINEAR H-INFINITY CONTROL APPROACH TO STABILIZATION OF DISTRIBUTED SYNCHRONOUS GENERATORS 9

xi = [xi1 , xi2 , xi3 ]T , with xi1 = Δδi , xi2 = Δωi , and xi3 = Eq i

to Δu, that is u∗ = u + Δu
are the state variables for the ith machine and i = 1, 2, . . . , n.
ẋd = Axd + Bu∗ + d2 . (62)
For the case of n = 3 interacting SGs, the dynamic model
of the machines consists of the following set of differential The dynamics of the controlled system described in (61) can be
equations: also written as

ẋ1 = x2 − ω0 (52) ẋ = Ax + Bu + Bu∗ − Bu∗ + d1 (63)

D1 Pm 1 and by denoting d3 = −Bu∗ + d1 as an aggregate disturbance

ẋ2 = − (x2 − ω0 ) + ω0 term, one obtains
2J1 2J1
ω0 ẋ = Ax + Bu + Bu∗ + d3 , (64)
− {G11 x23 + x3 [x6 G12 sin(x1 − x4 − a12 )
2J1
By subtracting (62) from (64), one has
+ x9 G13 sin(x1 − x7 − a13 )]
(53) ẋ − ẋd = A(x − xd ) + Bu + d3 − d2 . (65)

1 1 xd 1 − xd 1

1 By denoting the tracking error as e = x − xd and the aggregate
ẋ3 = −  x3 +  Vs cos(x1 ) + u1 (54) disturbance term as d˜ = d3 − d2 , the tracking error dynamics
Td 1 Td o 1 xd Σ Td o 1
1 becomes
ẋ4 = x5 − ω0 (55) ˜
ė = Ae + Bu + d. (66)
D2 Pm 2
ẋ5 = − (x5 − ω0 ) + ω0 The above linearized form of the distributed power generators’
2J2 2J2 model can be efficiently controlled after applying an H-infinity
ω0 2 feedback control scheme.
− {G22 x6 + x6 [x3 G21 sin(x4 − x1 − a21 )
2J2
+ x9 G23 sin(x4 − x7 − a23 )] REFERENCES
(56) [1] J. M. Guerrero, M. Chandorkar, T. L. Lee, and P. C. Loh, “Advanced

control architectures for intelligent microgrids—Part I: Decentralized and
1 1 xd 2 − xd 2 1 hierarchical control,” IEEE Trans. Ind. Electron., vol. 60, no. 4, pp. 1254–
ẋ6 = −  x6 +  Vs cos(x4 ) + u2 (57) 1262, Apr. 2013.
Td 2 Td o 2 xd Σ Td o 2 [2] L. Dong, Y. Zhang, and Z. Gao, “A robust decentralized load frequency
2
controlller for interconnected power systems,” ISA Trans., vol. 51, pp. 410–
ẋ7 = x8 − ω0 (58) 419, 2012.
[3] X. B. Chen, D. D. Siljiak, and S. S. Stankovic, “Decentralized H ∞ de-
D2 Pm 3 sign of automatic generation control,” in Proc. IFAC 15th World Congr.,
ẋ8 = − (x8 − ω0 ) + ω0 Barcelona, Spain, 2002.
2J3 2J3 [4] A. E. Leon, J. M. Mauricio, and J. A. Solsona, “Multi-machine power sys-
ω0 tem stability improvement using an observer-based nonlinear controller,”
− {G32 x29 + x9 [x3 G31 sin(x7 − x1 − a31 ) Elect. Power Syst. Res., vol. 89, pp. 202–214, 2012.
2J3
[5] K. Kalsi, J. Lian, and S. H. Zak, “Decentralized control of multi-machine
+ x6 G32 sin(x7 − x4 − a32 )] power systems,” in Proc. Amer. Control Conf., St. Louis, MO, USA, 2009,
pp. 2122–2127.
(59) [6] L. Jiang, Q. H. Wu, and J. Y. Wen, “Decentralized nonlinear adaptive con-
 trol for multimachine power systems via high-gain perturbation observer,”
1 1 xd 3 − xd 3 1 IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 51, no. 10, pp. 2052–2059,
ẋ9 = −  x9 +  Vs cos(x7 ) + u3 . Oct. 2004.
Td 3 Td o 3 xd Σ Td o 3 [7] W. Dib, R. Ortega, A. Barabanov, and F. Lamnabhi-Lagarrigue, “A
3
(60) globally convergent controller for multi-machine power systems using
structure-preserving models,” IEEE Trans. Automat. Control, vol. 54,
no. 9, pp. 2179–2185, Sep. 2009.
C. Equivalent Linearized Dynamics of the Distributed Power [8] W. Dib, R. Ortega, A. Astolfi, and D. Hill, “Improving transient stabil-
ity of multimachine power systems: Synchronization via immersion of a
Generators pendular system,” in Proc. Amer. Control Conf., San Francisco, CA, USA,
Jul. 2011, pp. 1408–1413.
After linearization round its current operating point, the dy- [9] S. Gros, “A distributed algorithm for NMPC-based wind farm control,”
namic model of the distributed power generators is written as in Proc. 53rd IEEE Conf. Decis. Control, Los Angeles, CA, USA, Dec.
15–17, 2014, pp. 4844–4849.
[10] Z. Xi, G. Feng, D. Cheng, and Q. Lu, “Nonlinear decentralized saturated
ẋ = Ax + Bu + d1 . (61) controller design for power systems,” IEEE Trans. Control Syst. Technol.,
vol. 11, no. 4, pp. 539–547, Jul. 2003.
Parameter d1 stands for the linearization error in the inverter’s [11] Q. Lu, S. Mei, W. Hu, F. Wu, Y. Ni, and T. Shen, “Nonlinear decentralized
disturbance attenuation excitation control via new recursive design for
model appearing in (61). The reference setpoints for the in- multi-machine power systems,” IEEE Trans. Power Syst., vol. 16, no. 4,
verter’s state vector are denoted by xd = [xd1 , . . . , xd4 ]. Tracking pp. 729–736, Nov. 2001.
of this trajectory is achieved after applying the control input u∗ . [12] Y. Wang, D. Cheng, C. Li, and Y. Ge, “Dissipative hamiltonian realization
and energy-based -disturbance attenuation control of multimachine power
At every time instant the control input u∗ is assumed to differ systems,” IEEE Trans. Automat. Control, vol. 48, no. 8, pp. 1428–1433,
from the control input u appearing in (61) by an amount equal Aug. 2003.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10
[13] Z. Xia, D. Chen, Q. Lu, and S. Mei, “Nonlinear decentralized controller de-
sign for multimachine power systems using hamiltonian function method,”
Automatica, vol. 38 pp. 527–534, 2002.
[14] Y. Guo, D. J. Hill, Y. Wang, “Nonlinear decentralized control of large-scale
power systems,” Automatica, vol. 36, pp. 1275–1289, 2000.
[15] N. Yadaiah and N. Venkata Ramara, “Linearization of multi-machine
power system: Modelling and control—A survey,” Electr. Power Energy
Syst., vol. 29, pp. 297–311, 2007.
[16] G. Rigatos, P. Siano, and N. Zervos, “Derivative-free nonlinear Kalman
filtering for PMSG sensorless control,” in Mechatronics Engineering: Re-
search Development and Education, M. Habib and P. Davis, Eds. London,
U.K.: ISTE, 2012.
[17] G. Rigatos, P. Siano, and N. Zervos, “Sensorless control of distributed
power generator with the derivative-free nonlinear Kalman filter,” IEEE
Trans. Ind. Electron., vol. 61, no. 11, pp. 6369–6382, Nov. 2014.
[18] G. Rigatos, P. Siano, and N. Zervos, “PMSG sensorless control with the
use of the derivative-free nonlinear kalman filter,” in Proc. IEEE Int. Conf.
Clean Elect. Power, Alghero, Italy, Jun. 2013, pp. 673–678.
[19] M. A. Mahmud, M. J. Hossain, H. R. Pota, and N. K. Roy, “Robust
nonlinear excitation controller design for multimachine power systems,”
IEEE PES Gen. Meeting—Conf. Expo., National Harbor, MD, USA, Jul.
2014.
[20] G. Rigatos, Nonlinear control and filtering using differential flatness ap-
proaches: Applications to electromechanicsl systems. New York, NY,
USA: Springer, 2015.
[21] M. Gros, “A distributed algorithm for NMPC-based wind farm control,”
in Proc. 53rd IEEE Conf. Decis. Control Los Angeles, CA, USA, Dec.
2014, pp. 4844–4849.
[22] C. Zhao and S. Low, “Optimal decentralized primary frequency control in
power networks,” in Proc. 53rd IEEE Conf. Decis. Control, Los Angeles,
California, USA, Dec. 15–17, 2014, pp. 2467–2473.
[23] C. Zhao, U. Topcu, N. Li, and S. Low, “Design and stability of load-
side primary frequency control in power systems,” IEEE Trans. Automat.
Control, vol. 59, no. 5, pp. 1177–1189, May 2014.
[24] H. Nazaripouya and S. Mehraeen, “Modeling and nonlinear optimal con-
trol of weak/islanded grids using FACTS device in a game theoretic ap-
proach,” IEEE Trans. Control Syst. Technol., vol. 24, no. 1, pp. 158–171,
Jan. 2016.
[25] X. Xie, W. Cui, and Y. Han, “Simultaneously tuning decentralized nonlin-
ear optimal excitation controllers in multimachine power systems,” Elect.
Power Syst. Res., vol. 74, pp. 371–378, 2005.
[26] M. Zribi, M. S. Mahmoud, M. Karkoub, and T. T. Lie, “H-infinity con-
trollers for linearised time-delay power systems,” Inst. Electr. Eng. Proc.-
Gener., Transmiss. Distrib., vol. 147, no. 3, pp. 401–408, Nov. 2000.
[27] M. J. Alden and X. Wang, “Robust H-infinity control of time delayed
power systems,” Syst. Sci. Control Eng., vol. 3, pp. 253–261, 2015.
[28] G. Rigatos, P. Siano, P. Wira, and F. Profumo, “Nonlinear H-infinity
feedback control for asynchronous motors of electric trains,” J. Intell.
Ind. Syst., vol. 1, no. 2, pp. 85–98, 2015.
[29] G. Rigatos, P. Siano, and C. Cecati, “A new nonlinear H-infinity feedback
control approach for three-phase voltage source converters,” Elect. Power
Compon. Syst., vol. 44, no. 3, pp. 302–312, 2015.
[30] G. Rigatos, Modelling and Control for Intelligent Industrial Systems:
Advanced Algorithms in Robotics and Industrial Engineering. New York,
NY, USA: Springer, 2011.
[31] G. Rigatos, Intelligent Renewable Energy Systems: Modelling and Con-
trol. New York, NY, USA: Springer, 2016.
[32] M. Basseville and I. Nikiforov, Detection of abrupt changes: Theory and
Applications. Englewood Cliffs, NJ, USA: Prentice-Hall, 1993.
[33] G. G. Rigatos and S. G. Tzafestas, “Extended Kalman filtering for fuzzy
modelling and multi-sensor fusion,” Math. Comput. Model. Dyn. Syst.,
vol. 13, pp. 251–266, 2007.
[34] G. Rigatos and Q. Zhang, “Fuzzy model validation using the local statis-
tical approach,” Fuzzy Sets Syst., vol. 60, no. 7, pp. 882–904, 2009.
[35] I. Rossi, L. Banta, A. Cuneo, M. Ferrari, and A. Traverso, “Real-time man-
agement solutions for a smart polygeneration microgrid,” Energy Convers.
Manage., vol. 112, pp. 11–20, 2016.
[36] Y. Zhang, T. Zhang, R. Wang, Y. Liu, and B. Guo, “Optimal operation
of a smart residential microgrid based on model predictive control by
considering uncertainties and storage impacts,” Solar Energy, vol. 122,
pp. 152–165, 2015.
[37] G. Li and M. Belmont, “Model predictive control of sea wave energy
converterse part II: The case of an array of devices,” Renew. Energy,
vol. 68, pp. 540–549, 2014.
IEEE SYSTEMS JOURNAL
[38] H. Nazaripouya and S. Mehraeen, “Modelling and nonlinear optimal con-
trol of weak/islanded grids using FACTS device in a gametheoretic ap-
proach,” IEEE Trans. Control Syst. Technol., vol. 24, no.1, pp. 158.171,
Jan. 2016
[39] X. Liu, Y. Zhang, and K. Y. Lee, “Robust distributed MPC for load fre-
quency control of uncertain power systems,” Control Eng. Pract., vol. 56,
pp. 136–147, 2016.
[40] M. D. Doan, T. Keviczky, and B. De Schutter, “Application of distributed
and hierarchical model predictive control in hydro power valleys,” in Proc.
4th IFAC Nonlinear Model Predictive Control Conf. Int. Fed. Automatic
Control, Noordwijkerhout, The Netherlands. Aug. 2012.
Gerasimos Rigatos (S’99–M’00–SM’15) received the Diploma degree in elec-
trical engineering and Ph.D. degree in control systems from the National Techni-
cal University of Athens (NTUA), Athens, Greece, 1995 and 2000, respectively.
He is currently a Researcher (Grade B’) with the Industrial Systems Institute,
Patras, Greece. He has held research and teaching positions with INRIA-IRISA
France in 2001, Université Paris XI in 2007, and Harper-Adams University Col-
lege, U.K. from 2011 to 2012 and in Greek universities. His research interests
include the areas of control and robotics, mechatronics, electric power systems,
computational intelligence, fault diagnosis, and optimization.
Dr. Rigatos is involved in editorial tasks for several international journals
and is also a member and CEng of the IET and a member of IMACS.
Pierluigi Siano (M’09–SM’14) received the M.Sc. degree in electronic engi-
neering and Ph.D. degree in information and electrical engineering from the
University of Salerno, Salerno, Italy, in 2001 and 2006, respectively.
He is an Aggregate Professor of electrical energy engineering with the De-
partment of Industrial Engineering, University of Salerno. In 2013, he received
the Italian National Scientific Qualification as a Full Professor with the com-
petition sector electrical energy engineering. His research interests include the
integration of distributed energy resources in smart distribution systems and on
planning and management of power systems. He has co-authored more than 160
papers including more than 70 international journals.
Dr. Siano is an Associate Editor for the IEEE TRANSACTIONS ON INDUSTRIAL
INFORMATICS and an Editor–in–Chief of the Journal of Intelligent Industrial Sys-
tems (Springer).
Alexey Melkikh (M’04) received the Diploma degree in physics (with a spe-
cialization in nuclear-power stations technology) and Ph.D. degree in molecular
physics from the Ural Polytechnical Institute, Yekaterinburg, Russia, in 1989
and 1995, respectively.
Since 2007, he has been a Professor with the Ural Federal University, Yeka-
terinburg, Russia. His research interests include the areas of control, chaotic
dynamics, games theory, evolution, and foundations of quantum mechanics.
Nikolaos Zervos received the Diploma degree in electrical and mechanical
engineering from the National Technical University of Athens, Athens, Greece,
the M.Sc. degree in systems and computing science from Carleton University,
Ottawa, ON, Canada, and the Ph.D. degree in electrical engineering from the
University of Toronto, Toronto, ON, Canada.
He is currently a Researcher (Grade A’) with the Industrial Systems Institute,
Patras, Greece. He was with Bell Laboratories, initially with AT&T and then
with Lucent Technologies, as an Acting Technical Manager of multimedia access
communication networks. He is one of the world’s experts in bandwidth-efficient
digital transmission and author of several patents in the areas of data transmission
and digital signal processing.