Nonlinear Robust Coordinated PSS-AVR Control for a

Synchronous Generator connected to an Infinite Bus

Cristiano Maria Verrelli, Riccardo Marino, Patrizio Tomei, Gilney Damm

To cite this version:

Cristiano Maria Verrelli, Riccardo Marino, Patrizio Tomei, Gilney Damm. Nonlinear Robust Coordi-
nated PSS-AVR Control for a Synchronous Generator connected to an Infinite Bus. IEEE Transactions
on Automatic Control, 2022, 67 (3), pp.1414 - 1422. �10.1109/TAC.2021.3062174�. �hal-03482768v2�

HAL Id: hal-03482768

https://hal.science/hal-03482768v2
Submitted on 10 Jun 2022

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

Nonlinear Robust Coordinated PSS-AVR Control

for a Synchronous Generator connected to
an Infinite Bus
C. M. Verrelli, R. Marino, P. Tomei, G. Damm

Abstract—The transient stabilization and voltage regulation as LM. Even though satisfactory results are achieved by
problem for a synchronous generator connected to an infinite the ”Desensitized Four Loops Regulator” DFLR in [3] and
bus is addressed, in this technical note, in the presence of [19] (here referred to as robust coordinated PSS-AVR in the
all uncertain physical parameters. First, new interpretations
for the Desensitized Four Loops Regulator, here referred to whole paper), two open problems remain: P1 : in order to
as robust coordinated PSS-AVR (Power System Stabilizer & obtain satisfactory performance over a wide range of operating
Automatic Voltage Regulator), are derived in terms of minimum conditions, different behaviours of nonlinear power systems in
phase properties with respect to a suitably chosen output for different operating conditions require different control objec-
the linearized error system. Secondly, on the basis of such an tives and consequently different linear control gain scheduling;
output, a nonlinear generalization of the robust coordinated
PSS-AVR is designed with the aim of enlarging the stability P2 : the absence of nonlinear feedback terms in non-switching
region and improving the transients about the unpredictable controllers may prevent the enlargement of the machine
operating conditions determined by the faults. Its linear action stability region. Two different strategies are then in order.
coincides with the one provided by the robust coordinated Switching Linear Controllers: the first strategy consists in
PSS-AVR: the compelling simplicity of control structure (just using different linearization-based controllers with switching
one integrator is involved) and robust tuning procedure of the
linear design are definitely inherited, with, additionally, no use actions depending on the different operating conditions (see
of the mechanical input power. A numerical analysis along [10] and references therein for applications of gain scheduling,
with realistic simulations confirm improved stability when the fuzzy control and multi-model techniques). Nonlinear Non-
proposed nonlinear controller is applied. Switching Controllers: the second strategy consists in resorting
Keywords: Nonlinear control; Synchronous generators; to nonlinear non-switching controllers that incorporate nonlin-
Transient stability; Voltage regulation; Robust coordinated ear feedback terms, simultaneously allowing for the automatic
PSS-AVR. gain scheduling and enlargement of the machine stability
region. The structures of such controllers however are rather
complex and require additional measurements and parameters.
I. I NTRODUCTION Examples are provided: i) a state feedback solution is proposed
in [4] (see also [16], [18]), in the presence of uncertainties in
The problem of power systems stability [1], [20] is still
almost all system parameters; ii) output feedback solutions are
widely studied, according to the latest related theoretical
reported in [22] and [5], in the presence of transmission line
results in [2] and [24], [25], [26], [27]. Transient stability is to
impedance and mechanical power uncertainties. In contrast to
be enhanced in the presence of perturbations such as changes
the robust coordinated PSS-AVR strategy, the strategy adopted
in load, outages of powers plants, failures in transformer
in the aforementioned papers (see also [12]) does not consider
substation and power lines. Electromechanical oscillations -
the terminal voltage as the natural controlled output. Rather, it
namely, local, inter-machine, and inter-area oscillations - have
artificially transforms the power angle into the main controlled
to be damped, while terminal voltage is to be regulated to
output. As a consequence, the (available) terminal voltage
its reference value. Linear and local controllers are currently
regulation error is not directly used in the feedback action,
employed. In practice, Linear Automatic Voltage Regulators
with either the power angle being consequently assumed to
(AVRs) are used in conjunction with Power System Stabilizers
be directly measured or the knowledge of critical parameters
(PSSs) (see [3], [6], [7], [13], [14], [19]). Those controllers
being required.
are typically numerically designed and tuned on the basis
This technical note presents two contributions. The first contri-
of the classical reduced model of a synchronous generator
bution provides new interpretations for the ”Desensitized Four
connected to an infinite bus, whose dynamics are linearized
Loops Regulator” of [3], by viewing it as a robust stabilizer
around the (unknown) desired constant operating condition
for the stably invertible system LM with respect to a suitably
that guarantees terminal voltage regulation. Such a linear
chosen output function. The second contribution takes advan-
model will be referred, throughout this technical note, to
tage from the previously defined output and presents a new
C.M. Verrelli, R. Marino, P. Tomei are with the Electronic Engineering De- nonlinear controller that generalizes the robust coordinated
partment of the University of Rome Tor Vergata, Via del Politecnico 1, 00133 PSS-AVR - namely, its linear action coincides with the one
Rome, Italy, E-mail (of the corresponding author): [email protected].
G. Damm is with COSYS-LISIS, Univ Gustave Eiffel, IFSTTAR, F-77447 provided by the robust coordinated PSS-AVR -, with, addi-
Marne-la-Vallée, France. tionally, no use of the mechanical input power. In fact, such

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

a nonlinear generalization just endows the robust coordinated voltage. The reader is referred to [5] for the physical meaning
PSS-AVR with: i) a nonlinear term that relies on the power of D, H and the six positive parameters θi > 0, 1 ≤ i ≤ 6.
angle (being computed as in [22] or being measured via Model (1) turns out to be the well-known single-machine-
now available wide-area measurement technologies), allowing infinite-bus representation, which, though of low dynamic
for an automatic gain scheduling; ii) nonlinear robustifying order, well captures the main (small-signal) characteristics of
terms on the back-stepping-based electrical power tracking a synchronous generator in a power network in the presence of
error, with the aim - whose effectiveness is illustrated by a parameter variations (see [3], [19]). Now, in order to avoid the
numerical analysis - of enlarging the machine stability region. physical singularities at δ = 0 and δ = π, the operation of the
The simplicity of control structure and tuning procedure of system is restricted to an arbitrarily large compact (connected)
the linear design is maintained: just one integrator is involved, set B contained in the open set:
while all model parameters are uncertain and are not estimated  
by the controller. From a theoretical point of view, such a D = (δ, ω, Pe ) : (δ, ω, Pe ) ∈ (0, π) × R × R+ .
nonlinear contribution simultaneously improves: i) the results
in [8], [10], [23], [28], since transient stabilization and voltage
B. Problem formulation
regulation tasks are simultaneously accomplished, with no use
of robust design strategies that require the existence of suitable Definition 1 (transient stabilization & voltage regula-
stabilizing matrices and with no need of model-knowledge- tion): Consider the synchronous generator described by the
based feedback linearizing precompensation-actions or switch- dynamic model (1), with its operation being restricted to
ing between transient stabilizers and voltage regulators; ii) B. Assume that all the model parameters are constant and
the output feedback results in [22] (see also [5] and refer- uncertain. The (robust) transient stabilization and voltage
ences therein), since, even when the power angle profile is regulation problem for the synchronous generator consists
computed as in [22], only the values for the generator direct in exponentially regulating to zero: i) the relative speed ω
axis reactance and the voltage of the infinite bus are here (transient stabilization); ii) the error between terminal voltage
required to be available, whereas several model parameters Vt and its reference value Vtr = 1 (p.u.) (voltage regulation).
(including the damping coefficient) are assumed to be known The problem formulation of Definition 1 is well-posed, since
in [22]; iii) the state feedback results in [4] (and all references the desired operating condition (at which voltage regulation is
therein), since several parameter estimates are not required, achieved) exists. It is given by δ = δs , ω = 0, Pe = Pm ,
thus reducing the control complexity. Realistic simulations where (see [5] for the physical meaning of the parameters):
concerning a benchmark scenario show that the nonlinear " s #
Vs 2 Xd Vs 2 Vtr 2
controller preserves the stability when a short circuit occurs, δs = arccotg − + −1
whereas instability arises when the standard PSS-AVR is used. Xs Pm (Xs + Xd ) Xs 2 Pm 2
belongs to the open set (0, π/2) and guarantees terminal
II. P ROBLEM FORMULATION AND DFLR CONTROL voltage regulation, namely:
In this section we present the generator dynamic model, " # 21
2
formulate the problem and introduce the robust coordinated θ4 Pm
+ θ5 + θ6 Pm cotg(δs ) = Vtr . (2)
PSS-AVR control (DFLR control). sin2 (δs )
It corresponds to a stable equilibrium for system (1) fed by the
A. Dynamic model open loop control uf s = θ1 Pm θ3−1 sin(δ
h
−1
s )  . According to
The third order dynamic model of a synchronous generator [4], there is also another value δu ∈ π/2, π for δ (see also
connected to an infinite bus (see [5] for its description and [2]) that guarantees terminal voltage regulation. It however
[21], [2] for more general models) is given by: corresponds to an unstable equilibrium for system (1) fed by
δ̇ = ω the open loop control uf u = θ1 Pm θ3−1 sin(δu )−1 .
D ωs ωs
ω̇ = − ω− Pe + Pm (1)
H H H C. Robust coordinated PSS-AVR control (DFLR)
Ṗe = −θ1 Pe + θ2 ω sin2 (δ) + Pe ωcotg(δ) + θ3 sin(δ)uf Let us introduce the following regulation errors: ω̃ = ω,
" # 12 P̃e = Pe − Pm , Ṽt = Vt − Vtr . The classical Desensitized
θ4 Pe2
Vt = + θ5 + θ6 Pe cotg(δ) , Four Loops Regulator (DFLR) (or, equivalently, the robust
sin2 (δ) coordinated PSS-AVR) in [3] [see also (2)-(3) in [19]] can
in which: δ (rad) denotes the generator power angle with be then written as:
Z t
respect to the infinite bus rotating at synchronous speed ωs ;
uf (t) = kω ω̃(t) − kp P̃e (t) − kv Ṽt (t) − kI Ṽt (τ )dτ, (3)
ω = ωg − ωs (rad/s) is the difference between the generator 0
angular speed ωg and the synchronous speed ωs ; Pe (p.u.)
in terms of suitable control parameters kω , kp , kv , kI . The
denotes the active electrical power delivered by the generator
above DFLR is apparently constituted by the combination of:
to the infinite bus; uf (p.u.) is the input to the Silicon
i) the (transient) stabilizing control action
Controlled Rectifier amplifier of the generator; Pm (p.u.) is
the mechanical input power, Vt (p.u.) is the generator terminal uf,ts (t) = kω ω̃(t) − kp P̃e (t) (4)

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

that is represented, according to the second equation of (1), [δs , 0, Pm ]T . Let µ, ν satisfy the constraints of Lemma 1. Then
by the transfer function (acting on the Laplace transform of the application of the (µ, ν)-DFLR (ky , kyI > 0):
the relative speed ω̃): Z t
uf,µ,ν (t) = −ky y(t) − kyI y(τ )dτ (9)
Wts (s) = kω + kp Hωs −1 s + DH −1 ;


(5) 0

ii) the voltage regulating control action achieves transient stabilization and voltage regulation, pro-
Z t vided that ky is sufficiently large.
uf,vr (t) = −kv Ṽt (t) − kI Ṽt (τ )dτ (6) Proof: Equivalently express the generalized (µ, ν)-DFLR
0 (9) as:
that is represented by the transfer function (acting on the
Laplace transform of the terminal voltage regulation error Ṽt ): uf,µ,ν (t) = −ky y(t) + ûf s (t)
û˙ f s (t) = −kyI y(t), ûf s (0) = 0. (10)
Wvr (s) = − (kv + kI /s) . (7)
Consider the transfer function WM (s). For the sake of com-
As shown by [3], such DFLR can be put, by using straight- pactness: i) denote by b1 , b2 , b3 the positive coefficients
forward approximations that do not significantly deteriorate of its numerator and let them constitute the components
the performance of the controller, in the standard AVR-PSS of a column vector b, with the zeroes of the polynomial
structure IEEE ST1A + PSS1A reported in Figures 2 and 4 p(s) = b1 s2 + b2 s + b3 all belonging to C− ; ii) denote by a1 ,
of [3], with (5) and (7) mimicking the actions of the PSS and a2 , a3 the coefficients of its monic denominator and let them
AVR, respectively. constitute the components of the column vector a. Accordingly
take the realization of WM (s) in observability form:
III. N EW INTERPRETATIONS FOR THE DFLR  
0 1 0
Define, besides the already defined regulation errors (ω̃ = ω, ẋ =  0 0 1  x + ay + bu , Ac x + ay + bu
P̃e = Pe − Pm , Ṽt = Vt − Vtr ), the power angle regulation 0 0 0
error: δ̃ = δ−δs . Write the first order approximated expression
y = [1, 0, 0]x. (11)
of Ṽt (around the desired operating condition [δs , 0, Pm ]T )
as Ṽt = −β0 δ̃ + β1 P̃e , where β0 and β1 turn out to be Recalling that b1 > 0, the linear change of coordinates
positive over the generator operating region (see [28]). We [η T , y] , [η1 , η2 , y] - with ηi = xi+1 −bi+1 b1 −1 x1 , 1 ≤ i ≤ 2,
preliminarily state Lemma 1, whose results will be used in y = x1 - can be performed and stability arguments similar to
the reminder of this technical note. those used in [17], once specialized to the case of constant
Lemma 1: The linearized dynamics LM of the synchronous disturbance compensation, can be used to prove the thesis.
generator around the desired operating condition [δs , 0, Pm ]T The following Corollary demonstrates that the above (µ, ν)-
is of relative degree one and minimum phase with respect to DFLR (9) is related to a reparameterization of the DFLR (3).
the output Corollary 1: Consider the linearized dynamics LM of
y = Ṽt − µω + ν P̃e , (8) the synchronous generator around the operating condition
[δs , 0, Pm ]T . Let P̃e0 , ω̃0 , Ṽt0 denote the initial conditions for
where µ and ν are reals satisfying ωs µ + (β1 + ν)D > 0 & P̃e (t), ω̃(t), Ṽt (t) at time t = 0. Define the constant:
β1 + ν > 0.   !
Proof: Starting from (1), consider the linearized dynamics νD β1 Ṽt0 νH
σ0 = kyI µ + P̃e0 − + kyI ω̃0 . (12)
LM in terms of the state variables: δ̃, ω̃, P̃e and the output ωs β0 β0 ωs
y. By mimicking the first steps in [22], perform the change of
coordinates: z1 = δ̃, z2 = ω̃, z3 = −DH −1 ω̃ − ωs H −1 P̃e Then, the (µ, ν)-DFLR (10) with ûf s (0) = σ0 becomes
and thus obtain the system expressed in the new coordi- the DFLR (3), with kv , kω , kp satisfying: kω =
ky µ +
nates. It is of relative degree one and in reachability form: kyI ωs−1 Hν, kp = ky ν −
kyI β1 β0−1 µ + νDωs −1 , kv =
its transfer function (between the output y and the input ky + kyI β0−1 µ + νDωs −1 , kI = kyI .
u , ωs θ3 H −1 sin(δs )(uf − uf s )) can be derived to be: Proof: Use the δ̃ and ω̃ dynamics for the computation of
  the y-integral and write δ̃ as β0−1 (β1 P̃e − Ṽt ).
(β1 +ν)H 2 (β1 +ν)D
ωs s + µ + ωs s + β0 Once Lemma 1 and Corollary 1 have been established, a
WM (s) = 3 direct guideline in the control parameters tuning for (3) can
D
− α2 s2 + ωHs α1 − H D
α2 s + ωHs α0



s + H
be easily obtained: it involves just one parameter to be set at
where α0 , α1 , α2 are the reals: α0 = θ3 uf s cos(δs ), α1 = last, in accordance with the following corollary.
θ2 sin2 (δs ) + Pm cotg(δs ), α2 = −θ1 depending on the Corollary 2 (limit behaviour): Consider the linearized
generator operating condition. Owing to the positive nature dynamics LM of the synchronous generator around the
of the coefficients characterizing the numerator of WM (s), operating condition [δs , 0, Pm ]T . Then, the application of
the minimum phase property is guaranteed. the (µ, ν)-DFLR (9) with kyI /ky being constantly set to z,
We now state Theorem 1. for ky tending to +∞, makes: i) two of the closed loop
Theorem 1: Consider the linearized dynamics LM of eigenvalues converge to the two open loop zeros of WM (s)
the synchronous generator around the operating condition (not explicitly depending on Xd0 , Td0 , Kc ); ii) one of the closed

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

loop eigenvalues converge to −z; iii) one of the closed loop generalization of the robust coordinated PSS-AVR is given
eigenvalues tend to −∞ (along the real axis). by:
Proof: Consider LM, expressed, in accordance with the
−kp (Pe − Per ) − kr Φ(δ, Pe )(Pe − Per ) + λρ̂˙
proof of Lemma 1, in terms of the coordinates z1 , z2 , z3 and uf =
rewrite uf,µ,ν in terms of such coordinates. Define the new sin(δ)
 2 2 2
variable Φ(δ, Pe ) = sin (δ) + [Pe cotg(δ)]
Hz4 = −ωs α0 z1 − (ωs α1 − Dα2 ) z2 − (D − Hα2 ) z3 Per = kpω ω + ρ̂ (15)
−ωs θ3 sin(δs )(uf − uf s ) (13) ρ̂˙ = −µv Ṽt + µω ω,
where: kp , kr , λ, kpω , µv are suitable positive control parame-
and compute its dynamics. The proof relies on the root
ters; µω is a suitable non-negative control parameter satisfying
locus analysis and on the structure of the corresponding
max{D, µω } > 0.
characteristic polynomial for the error system in z1 , z2 , z3 , z4 -
Before stating the main result of this section (namely,
coordinates, which reads
Theorem 2), the following comments are in order.
qz (s) = m1 (s) + ky ωs sin(δs )θ3 H −1 m2 (s), (14) Comment A. The nonlinear robust coordinated PSS-AVR
(15) involves the injection of the power angle to provide
where the roots of the polynomials m1 (s) and m2 (s) denote an automatic gain scheduling and to handle the severe dis-
the open loop poles and zeroes, respectively (recall WM (s) turbances and contingencies characterizing the operation of
and the Laplace transform of uf,µ,ν in (9)). Such open loop power systems. We can provide two interpretations for the
zeroes, in accordance with the definition of β0 , β1 and θ4 , θ6 , above controller (15): i) the state feedback one, when the
δs , do not explicitly depend on Xd0 , Td0 , Kc . power angle is measured via now available wide-area mea-
The proof of such a corollary clarifies that non-zero µ surement technologies (see [9]); ii) the output feedback one,
and ν in (9) are assigned to guarantee satisfactory transient when the values of δ and Xs are computed through the (δ, Xs )-
performance, and in particular, to satisfactorily damp down formulas derived in [5] (Q is the reactive power):
speed oscillations: if µ = ν = 0, then, for arbitrarily large " s #
ky (and kyI /ky being constantly set to z), two closed loop Vs 2 Xd Vs 2 Vt 2
eigenvalues are attracted by the open loop zeroes of WM (s), δ = arccotg − ± −1
Xs Pe (Xs + Xd ) Xs 2 Pe 2
which, for such µ = ν = 0, guarantee a non-satisfactory p
damping since D is typically very small. On the other hand, −QVs2 ± Q2 Vs4 − (Q2 + Pe2 )(Vs2 − Vt2 )Vs2
Xs =
Corollary 2 suggests to choose µ, ν, kyI /ky to robustly place Q2 + Pe2
the open loop zeroes in the left half complex plane C− that just require the knowledge of the generator direct axis
(independently of Xd0 , Td0 , Kc ), and then to largely increase reactance and the voltage of the infinite bus2 . With this respect,
ky to achieve transient stabilization and voltage regulation1 . recall that Xs may undergo sudden variations when electrical
The different parameterization of the DFLR (3) rather allows, perturbations, such as faults on the transmission line, occur.
to arbitrarily impose, once (3) is applied to the error system Comment B. Set µω = 0 and ρ̂(0) = Pm . Then, when higher
in z1 , z2 , z3 , z4 -coordinates, the four closed loop eigenvalues order robustifying actions are neglected [w.r.t. the regulation
through the four control parameters kω , kp , kv , kI (to be and estimation errors δ̃, ω̃, P̃e , Pm − ρ̂], (15) reduces to (k̄p =
simultaneously chosen depending on the specific operating kp + kr Φ(δs , Pm )):
condition and all the system parameters). Z t
1 h i
uf = − k̄p P̃e + k̄p kpω ω − λµv Ṽt − k̄p µv Ṽt (τ )dτ
IV. N ONLINEAR GENERALIZATION OF THE DFLR sin(δ) 0

In this section, we consider the nonlinear model (1) in that coincides with the DFLR (3):
Z t
place of the linearized one LM. We accordingly present the 1 h i
design of a new nonlinear controller that generalizes, to the uf = − k̄p P̃e + k̄p kpω ω − λµv Ṽt − k̄p µv Ṽt (τ )dτ
sin(δs ) 0
nonlinear model (1), the design of the robust coordinated
once δ appearing at the denominator is replaced by δs and
PSS-AVR of Sections II-III, while additionally avoiding the
incorporated into the control gains. In other words, the non-
use of the mechanical input power. The proposed nonlinear
linear robust coordinated PSS-AVR (15) exhibits different
1 It is interesting to notice that the industrial coordinated PSS-AVR of Figure linearizations around different operating conditions. Those dif-
5 of [3] can be, in turn, obtained through the application of a slight variant of ferent linear actions can be seen as equivalent actions of local
the approach underlying the proof of Theorem 1. Such a variant is based on
Lemma 4 in [15]. In particular, while the controller in Theorem 1 (described
robust coordinated PSS-AVR linear controllers with different
by a proper rational function) first feeds back the output error y to stabilize control parameters. The controller (15) can be thus viewed as
the (minimum phase) error system and then includes an integral action on y a nonlinear version of a simple combination between a robust
being responsible of generating a converging estimate of the uncertain constant
uf s , such a different controller relying on Lemma 4 in [15] (described by
coordinated PSS-AVR and an automatic smooth nonlinear gain
a strictly proper transfer function) first includes an integrator in its transfer scheduling procedure. The operating status of the controller
function and then stabilizes the resulting error feedback through a proper
rational function. In this respect, notice that Lemma 4 in [15] is correct for 2 Uncertainties about the infinite bus voltage value are usually present in
d ≤ 4, whereas, for d ≥ 5, a typo concerning g0 in place of qd appears the practice. In the realistic simulations of Section V, just the nominal value
(which the counterexample of [11] to Lemma 4 of [15] just relies on). of the infinite bus voltage will be used.

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

is automatically changed when operating conditions vary, so is derived, where ηz (z1 , z2 , z3 , ep ) , ηV (δ̃, ω̃, P̃m , ep ). In
that a smooth transfer between the actions of local robust accordance with the companion form structure of the matrix
coordinated PSS-AVR linear controllers is performed. characterizing the above error subsystem, it is possible to
We are now able to present Theorem 2. choose µv , µω , kpω (affecting λω ) in order to robustly place
Theorem 2: Consider the nonlinear robust coordinated all its eigenvalues - namely, the roots of the polynomial ΠA (s)
PSS-AVR (15). Besides δ̃ = δ − δs and ω̃ = ω, define the new - in the left half complex plane C− (independently of Xd0 , Td0 ,
error coordinates: ep = Pe −Per −Pess , P̃m = Pm −Pess − ρ̂ Kc ). Meanwhile, notice how µω is to be necessarily positive
in terms of the suitable constant when D is zero. On the other hand, the ep -dynamics can be
−θ1 Pm derived as:
Pess = . (16)
θ3 [kp + kr Φ(δs , Pm )] ėp = −θ1 Pe + θ2 ω sin2 (δ) + Pe ωcotg(δ) + θ3 sin(δ)uf
h
− kpω z3 − ωs H −1 ep + µv β0 z1 − µv β1 ep


Then transient stabilization and voltage regulation is achieved.
In particular, the origin of the error system in the (δ̃, ω̃, −(D+
ωs kpω )H −1 ω̃+ωs H −1 P̃m , ep )-coordinates - corresponding to +µv β1 Hωs−1 z3 + µv β1 Hωs−1 λω z2 − µv β1 kpω z2
i
the desired operating condition [δs , 0, Pm ]T -, is exponentially −µv ηz (z1 , z2 , z3 , ep ) + µω z2 , (19)
stable, for sufficiently large kp and kr and for suitable µω ,
µv , kpω robustly placing in C− the roots of the polynomial with Pe again satisfying Pe = ep − P̃m + kpω ω̃ + Pm . Finally
(not explicitly depending on Xd0 , Td0 , Kc ): ΠA (s) = s3 + get:
((D + ωs kpω )H −1 + β1 µv )s2 + (β1 DH −1 µv + ωs H −1 µω )s + ėp = −θ1 ep − θ1 Pm + θ1 Hωs−1 (z3 + λω z2 ) − θ1 kpω z2
ωs β0 H −1 µv .
Proof: First compute the δ̃- and the ω̃- dynamics by +θ2 z2 sin2 (δ) + Pe z2 cotg(δ) + θ3 sin(δ)uf
defining λω = (D + ωs kpω )H −1 . According to the first part −µv β0 z1 − µv β1 Dωs−1 z2 − µω z2
of Section III, write − kpω + µv β1 Hωs−1 z3 + kpω ωs H −1 + µv β1 ep

Ṽt = −β0 δ̃ + β1 P̃e + hV (δ̃, P̃e ), (17) +µv ηz (z1 , z2 , z3 , ep ), (20)

where hV (·, ·) collects the higher order terms in the regulation whereas, from (15), write
errors δ̃ and P̃e . By writing P̃e = ep + kpω ω̃ − P̃m and by θ1 Pm
uf sin(δ) = −kp ep − kr Φ(δ, Pe )ep + λρ̂˙ + − kr Pess Λp ,
recalling the definition of Per , the Ṽt -expression, in the new θ3
error coordinates, becomes:
2
2 2
with Λp = sin2 (δ) − sin2 (δs ) + (Pe cotg(δ)) −
Ṽt = −β0 δ̃ + β1 kpω ω̃ − β1 P̃m + β1 ep + ηV (δ̃, ω̃, P̃m , ep ) 2
(Pm cotg(δs )) . On the basis of the Λp -expression and the
where ηV (δ̃, ω̃, P̃m , ep ) , hV (δ̃, P̃e ) collects, by definition, Pess -definition, write
the higher order terms in the new errors δ̃, ω̃, P̃m , ep . The kr θ1 Pm
−kr Pess Λp = Λp , αp Λp , (21)
injection of the output y into the P̃m -dynamics (no knowledge θ3 [kp + kr Φ(δs , Pm )]
of the mechanical input power is here assumed, so that ν = 0 with αp belonging to
the open connected set Cp =
has to be taken in (8)) is thus able to modify the last row 0, θ1 Pm θ3−1 Φ(δs , Pm )−1 . Now, in the RHS of (18), let: Az
of the matrix characterizing the (δ̃, ω̃, P̃m )- error subsystem. T
be the matrix multiplying the vector z = [z1 , z2 , z3 ] ; Bp be
Accordingly write the error dynamics in terms of the new the column vector multiplying ep ; Bη be the column vector
error variables: z1 = δ̃, z2 = ω̃, z3 = −λω ω̃ + ωs H −1 P̃m . multiplying ηz (z1 , z2 , z3 , ep ). Set the control parameters to
The aim is to recover the previously presented arguments satisfy the related assumption of Theorem 2 (concerning the
concerning the error system in z-coordinates, once: i) the value roots of ΠA (s)) and let Pz be the symmetric positive definite
of D has been artificially modified through the kpω -based pre- matrix satisfying the Lyapunov equation Pz Az + AT z Pz = −I
feedback action (with the aim of obviating the absence of ν); in terms of the identity matrix I ∈ M(3, R). Furthermore,
ii) ρ̂˙ has been endowed with the role of uf and y has been consider the composite Lyapunov function
fed back with the suitable negative sign in order to stabilize,
1
with the desired eigenvalues, the resulting third-order error W = z T Pz z + e2p (22)
subsystem. In this sense, the polynomial ΠA (s) does constitute 2
the mirrored counterpart of the characteristic polynomial of that satisfies, along the trajectories of the closed loop system,
the matrix characterizing LM in z-coordinates, once uf is the equality:
˙ After some tedious computations, the perturbed
replaced by ρ̂. Ẇ = −kzk2 + 2z T Pz Bp ep + 2z T Pz Bη ηz (z1 , z2 , z3 , ep )
triangular system:
− θ1 + kp θ3 − kpω ωs H −1 − µv β1 e2p
 

ż1 = z2 ˙ 3 ep − kr θ3 Φ(δ, Pe )e2

+αp Λp θ3 ep + λρ̂θ
ωs p
ż2 = z3 − ep +θ2 sin2 (δ)z2 ep + Pe cotg(δ)z2 ep
H
ż3 = −ωs β0 H −1 µv z1 − (β1 DH −1 µv + ωs H −1 µω )z2 +µv ηz (z1 , z2 , z3 , ep )ep − µv β0 z1 ep
−(λω + β1 µv )z3 + ωs H −1 (λω + β1 µv )ep + θ1 Dωs−1 − µω − µv β1 Dωs−1 z2 ep
 

+ωs H −1 µv ηz (z1 , z2 , z3 , ep ) + θ1 Hωs−1 − kpω − µv β1 Hωs−1 z3 ep .

 
(18) (23)

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

Recall that a2 − b2 = (a − b)(a + b) for a, b ∈ R; then the other hand, a back-stepping action that is reminiscent of
write Pe cotg(δ) − Pm cotg(δs ) = Pe cotg(δ) − Pm cotg(δ) + the typical two-time-scale arguments has been adopted4 . It
Pm cotg(δ) − Pm cotg(δs ), as well as P P̃e = ep + kpω z2 − is confined to the natural definition of the electrical power
3
Hωs−1 (z3 + λω z2 ). Then obtain |Λp | ≤ i=1 γpi |zi | + γ4 |ep | reference, so that all the uncertainties in the corresponding
over B ⊂ D, with γi (i = 1, 2, 3, 4) being positive reals electrical power tracking error dynamics are compensated by
depending on the diameter of B but not increasing while kp the integrator placed into the upper mechanical subsystem.
and kr increase. On the other hand, write Nonlinear robustifying actions can be just included at last,
h through the use of a composite Lyapunov function. In con-
λρ̂˙ = −λµv − β0 z1 + β1 kpω z2 − β1 Hωs−1 (z3 + λω z2 ) trast to the fully linearized analyses which Theorem 1 and
Corollary 2 rely on, genuinely nonlinear stability tools (applied
i
+β1 ep + ηz (z1 , z2 , z3 , ep ) + λµω z2 .
to the nonlinear model (1) with just local validity) allow
Equation (23) thus leads to the inequality3 us to neglect, in the Lyapunov analysis of Theorem 2, just
the nonlinearities concerning the terminal voltage equation,
Ẇ ≤ −kzk2 + 2kzkkPz kkBp k|ep | whereas the well-known nonlinearities affecting the electrical
+2kzkkPz kkBη k|ηz (z1 , z2 , z3 , ep )| power dynamics are explicitly faced by the nonlinear features
of the controller (aiming at enlarging the stability region) and
− θ1 + kp θ3 − kpω ωs H −1 − µv β1 + λµv β1 θ3 e2p
 
by one last control parameter to be set at last.
+αp |Λp |θ3 |ep | − kr θ3 Φ(δ, Pe )e2p Comment D. According to Comment B, set µω = 0,
+θ2 sin2 (δ)|z2 ||ep | + |Pe cotg(δ)||z2 ||ep | ρ̂(0) = Pm , replace δ appearing at the denominator by
+µv |ηz (z1 , z2 , z3 , ep )||ep | + µv β0 |z1 ||ep | δs and neglect higher order robustifying actions in (15).
Then, (15) coincides with DFLR (3) [under equivalent gains],
+µv β0 λθ3 |z1 ||ep | + θ1 Dωs−1 − µω − µv β1 Dωs−1 which admits an attraction domain Rl (whose estimate may
+µv β1 Dλθ3 ωs−1 + λθ3 µω |z2 ||ep | be provided by linearization arguments), whereas a possibly
different attraction domain Rnl exists (whose estimate is given
+ θ1 Hωs−1 − kpω − µv β1 Hωs−1 by the proof of Theorem 2) for the corresponding nonlinear
version (15). The effectiveness of (15) in including points in
+µv β1 Hωs−1 λθ3 |z3 ||ep | the state space of (1) - that do not belong to Rl - into its
+µv θ3 λ|ηz (z1 , z2 , z3 , ep )||ep |. attraction domain Rnl can be inspected through a numerical
analysis. Consider the synchronous generator in [22] (no input
By finally completing the squares and by setting sufficiently saturations are considered), whose physical parameter values
large kp and kr (under given µω , µv , kpω ), we obtain, in are: ωs = 314.159 rad/s, D = 5 p.u., H = 8 s, Td0 = 6.9 s,
terms of suitable positive reals gp and gpb , the last compact Kc = 1, Xd = 1.863 p.u., Xd0 = 0.257 p.u., XT = 0.127
inequality: p.u., XL = 0.4853 p.u.. The nominal operating condition
δs =1.256 rad, Pm =0.9 p.u., ω=0 rad/s guarantees Vt = Vtr = 1
Ẇ ≤ −kzk2 /4 − gp e2p + gpb |ηz (z1 , z2 , z3 , ep )|(|ep |
p.u., with Vs = 1 p.u.. The common control parameters for
+kzk), (24) both the controllers (15) and (3) [under equivalent gains]
where ηz (z1 , z2 , z3 , ep ) = hV (δ̃, P̃e ) just collects the higher are (in SI units): kpω = 1.5, λ = 3, kp = 1.5, kr = 9,
order terms of Ṽt in (17). Such a Ṽt is injected into the µv = 18 (µω = 0), while the controller initial conditions
error dynamics by the control (15), through the gain µv . are compatible with the operating point (with the feedforward
Exponential convergence of the (z, ep )-errors is thus achieved contribution uf s being included in the controllers). Figure 1 -
for sufficiently small initial (z, ep )-errors: z1 (0) = δ̃(0), concerning the numerical integration of the equations involved
z2 (0) = ω̃(0), z3 (0) = −λω ω̃(0)+ωs H −1 (Pm − ρ̂(0)−Pess ), in the considered scenario - shows that there exists a set of
ep (0) = Pe (0) − Pm − kpω ω̃(0) + (Pm − ρ̂(0) − Pess ), that initial conditions, namely δ(0) = 0.3 rad, ω(0) = −3 rad/s,
make, for any t and in accordance with (24): i) Ẇ be negative Pe (0) = 2.2 p.u., that belongs to the stability region Rnl
definite; ii) generator states and references belong to B. With guaranteed by (15) but not on Rl .
this respect, recall that: i) γi (i = 1, 2, 3, 4) are positive Comment E. The nonlinear robust coordinated PSS-AVR
reals that do not increase while kp and kr increase; ii) Pess (15) can be seen as the evolution of the nonlinear decentralized
turns out to decrease as kp and kr increase. On the other robust adaptive controller in [21], which was exclusively
hand, achieving exponential convergence of the (z, ep )-errors designed for transient stabilization, under a rather weak set of
to zero guarantees exponential convergence of δ̃(t) = δ(t)−δs , assumptions on the remote network dynamics. The inclusion
ω̃(t) = ω(t) to zero, along with exponential convergence to in (15) of the terminal voltage regulation error to drive the
zero of P̃e (t) and Pm − ρ̂(t) − Pess . mechanical input power estimate (acting as the integral term
Comment C. The design of (15) and the corresponding of (3)) allows (15) to achieve voltage regulation even at
stability analysis have taken advantage from the y-based permanently perturbed operating points (see the subsequent
arguments of Section III, as the P̃m -dynamics has shown. On Figure 2 concerning a 70%-decrease of Pm and a double

3 Notice that, when λθ = 1, several coefficients disappear from the non- 4 The use of the proposed technique can be extended to similar scenarios
3
definite sign terms. This explains the feedforward role of the λ-term in (15). concerning nonlinear systems with extended matching uncertainties.

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

connected to a 5000 MW power plant through a 700 km line.

The physical saturation for the generators exciters, which is
an important limitation for the control performance, is the
same for both the controllers. The parameters for the standard
PSS-AVR are set as: gain= 2, wash-out time constant 0.7,
lead-lag time constants Tnum = 60e−3 , Tden = 0.5. The
parameters of the output feedback control (15) are chosen as:
kpω = 70, λ = 90, kp = 10, kr = 10, µv = 10, µω = 10.
Figure 3 shows how the standard PSS-AVR is not able to
keep the system stability: δ goes outbounds. On the other
hand, when (15) is used in the test machine - with the infinite
bus remaining with the standard PSS-AVR -, systems stability
is assured even in the presence of 5% noise5 on the active
Fig. 1. Power angle δ, relative speed ω, electrical power Pe , terminal voltage
Vt (blue: DFLR; red: proposed nonlinear controller). and reactive electrical powers Pe and Q, as shown in Figure
4: all states are kept inside their operation bounds. According
to Figure 5, the standard PSS-AVR input saturates while
33%-increase of the nominal XL -value for the synchonous being not able to bring the generator back to synchronism
generator already used in Comment D., under the action of to the main grid, whereas the (15)- control input saturates
the nonlinear controller (15) and the equivalently tuned DFLR as well but it is then able to bring the generator back to its
(5) with (in SI units): kpω = 1.5, λ = 3, kp = 1.5, kr = 9, equilibrium point.
µv = 18 (µω = 0), and with the controllers initial conditions
being compatible with the operating point. Here the DFLR
performance slightly degrades with respect to (15).

Fig. 3. Standard PSS-AVR: (a) power angle δ in degrees; (b) angular speeds
ω1,2 in p.u.; (c) output voltages Vt1,2 .

Fig. 2. Power angle δ, relative speed ω, electrical power Pe , terminal voltage

Vt (blue: DFLR (3); red: nonlinear controller (15)).

V. S IMULATION RESULTS IN S IMSCAPE E LECTRICAL

This section illustrates the previous theoretical derivations
concerning the output feedback version of the control
algorithm (15), in comparison to the standard PSS-AVR
structure (see Section II.C). The quantities Xs and δ required
by (15) have been computed via the (δ, Xs )-formulas of [5],
Fig. 4. Output feedback control (15) under noise effects: (a) power angle δ
here using Xd and just the nominal value of the infinite in degrees; (b) angular speeds ω1,2 in p.u.; (c) output voltages Vt1,2 in p.u.
bus voltage Vs . The simulations have been carried out
with reference to the Matlab/Simulink Simscape Electrical
benchmark (including detailed electrical models and several VI. C ONCLUSIONS
additional elements), in which the duration of the short
circuit - placed just after the transformer connecting the test The transient stabilization and voltage regulation problem
machine to the transmission line - has been set to 90 ms for a synchronous generator connected to an infinite bus
and no SVC has been considered. In particular, simulations (see Definition 1) has been addressed when all the physical
concern the Single Machine Infinite Bus real case at parameters are uncertain. New interpretations for the DFLR (3)
https://fr.mathworks.com/help/physmod/sps/examples/svc-and- 5 Usually (in similar scenarios) all current structures work pretty well under
pss-phasor-model.html, represented by a 1000 MW machine, noise.

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAC.2021.3062174, IEEE
Transactions on Automatic Control

[13] K.T. Law, D.J. Hill, N.R. Godfrey, Robust co-ordinated AVR-PSS
design, IEEE Transactions on Power Systems, 9: 1218-1225, 1994.
[14] J. Machowski, S. Robak, J.W. Bialek, J.R. Bumby, N. Abi-Samra, De-
centralized stability-enhancing control of synchronous generator, IEEE
Transactions on Power Systems, 15: 1336-1344, 2000.
[15] I. Mareels, A simple selftuning controller for stably invertible systems,
Systems & Control Letters, 4: 5-16, 1984.
Fig. 5. Standard PSS-AVR (left); output feedback control (15) under noise [16] R. Marino, T. Shen, C.M. Verrelli, Robust adaptive transient stabiliza-
effects (right): control input in p.u. tion of a synchronous generator with parameter uncertainty, European
Journal of Control, 12: 135-155, 2006.
[17] R. Marino, P. Tomei, C.M. Verrelli, Learning control for nonlinear
systems in output feedback form, Systems and Control Letters, 61: 1242-
have been first derived in Section III (see Lemma 1, Theorem 1247, 2012.
1 and Corollaries 1-2). Such new interpretations have led to [18] R. Marino, C.M. Verrelli, A global state feedback output regulating
the nonlinear control (15), whose linear action coincides with control for uncertain systems in strict feedback form, Systems and
Control Letters, 58: 682-690, 2009.
the one provided by (3): simplicity of control structure (just [19] H. Quinot, H. Bourlés, T. Margotin, Robust coordinated AVR+PSS
one integrator is involved) and robust tuning procedure of for damping large scale power systems, IEEE Transactions on Power
the linear design (see Theorem 2) have been preserved, with, Systems, 14: 1446-1451, 1999.
[20] P.W. Sauer, M.A. Pai, Power System Dynamics and Stability, Prentice
additionally, no use of the mechanical input power. Control Hall: Upper Saddle River, 1998.
(15) has endowed the robust coordinated PSS-AVR (3) with: [21] C.M. Verrelli, G. Damm, Robust transient stabilization problem for a
i) a nonlinear term that relies on the power angle, allowing synchronous generator in a power network, International Journal of
Control, 83: 816-828, 2010.
for an automatic gain scheduling; ii) nonlinear robustifying [22] C.M. Verrelli, G. Damm, Output feedback transient stabilization and
terms on the back-stepping-based electrical power tracking voltage regulation of a synchronous generator, International Journal of
error, with the aim of enlarging the machine stability region Robust and Nonlinear Control, 22: 1495-1504, 2012.
[23] Y. Wang, G. Guo, D.J. Hill, Robust decentralized nonlinear controller
(as illustrated by numerical analysis). Realistic simulations design for multimachine power systems, Automatica, 33: 1725-1733,
concerning a benchmark scenario have finally shown that the 1997.
nonlinear controller preserves the stability when a 90 ms- [24] E. Weitenberg, C. De Persis, Robustness to noise of distributed averaging
integral controllers in power networks, Systems and Control Letters, 119:
long short circuit occurs, whereas instability arises when the 1-7, 2018.
standard PSS-AVR is used. [25] E. Weitenberg, C. De Persis, N. Monshizadeh, Exponential convergence
under distributed averaging integral frequency control, Automatica, 98:
R EFERENCES 103-113, 2018.
[26] E. Weitenberg, Y. Jiang, C. Zhao, E. Mallada, F. Dörfler, C. De Persis,
[1] P.M. Anderson, A.A. Fouad, Power System Control and Stability, Wiley- Robust decentralized frequency control: a leaky integrator approach,
IEEE Press: New York, 2003. European Control Conference, 764-769, June 12-15, 2018, Limassol,
[2] N. Barabanov, J. Schiffer, R. Ortega, D. Efimov, Conditions for almost Cyprus.
global attractivity of a synchronous generator connected to an infinite [27] E. Weitenberg, Y. Jiang, C. Zhao, E. Mallada, C. De Persis, F. Dörfler,
bus, IEEE Transactions on Automatic Control, 62: 4905-4916, 2017. Robust decentralized secondary frequency control in power systems:
[3] H. Bourlés, S. Peres, T. Margotin, M.P. Houri, Analysis and design of merits and trade offs, IEEE Transactions on Automatic Control, 64(10):
a robust coordinated AVR/PSS, IEEE Transactions on Power Systems, 3967-3982, 2019.
13: 568-575, 1998. [28] C. Zhu, R. Zhou, and Y. Wang, A new nonlinear voltage controller for
[4] G. Damm, F. Lamnabhi-Lagarrigue, R. Marino, C.M. Verrelli, Transient power systems, Electrical Power and Energy Systems, 19: 19-27, 1997.
stabilization and voltage regulation of a synchronous generator, in
Taming Heterogeneity and Complexity of Embedded Control, Editors:
F. Lamnabhi-Lagarrigue, S. Laghrouche, A. Loria and E. Panteley,
International Scientific & Technical Encyclopedia (ISTE), London, 211-
225, 2007.
[5] G. Damm, R. Marino, F. Lamnabhi-Lagarrigue, Adaptive nonlinear
output feedback for transient stabilization and voltage regulation of
power generators with unknown parameters, International Journal of
Robust and Nonlinear Control, 14: 833-855, 2004.
[6] G.J.W. Dudgeon, W.E. Leithead, A. Dyśko, J. O’Reilly, J.R. McDonald,
The effective role of AVR and PSS in power systems: frequency response
analysis, IEEE Transactions on Power Systems, 22: 1986-1994, 2007.
[7] A. Dyśko, W.E. Leithead, J. O’Reilly, Enhanced power system stability
by coordinated PSS design, IEEE Transactions on Power Systems, 25:
413-422, 2010.
[8] L. Gao, L. Chen, Y. Fan, H. Ma, A nonlinear control design for power
systems, Automatica, 28: 975-979, 1992.
[9] J. C. Gonzalez-Torres, G. Damm, V. Costan, A. Benchaib, F. Lamnabhi-
Lagarrigue, A novel distributed supplementary control of Multi-Terminal
VSC-HVDC grids for rotor angle stability enhancement of AC/DC
systems’, IEEE Transactions on Power Systems, doi: 10.1109/TP-
WRS.2020.3030538, October, 2020.
[10] Y. Guo, D.J. Hill, Y. Wang, Global transient stability and voltage
regulation for power systems, IEEE Transactions on Power Systems,
16: 678-688, 2001.
[11] J.B. Hoagg, D.S. Bernstein, Direct adaptive dynamic compensation for
minimum phase systems with unknown relative degree, IEEE Transac-
tions on Automatic Control, 52: 610-621, 2007.
[12] X. Jiao, Y. Sun, T. Shen, Adaptive controller design for a synchronous
generator with unknown perturbation in mechanical power, International
Journal of Control, Automation, and Systems, 3: 308-314, 2005.

0018-9286 (c) 2021 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.
Authorized licensed use limited to: IFSTTAR. Downloaded on March 05,2021 at 09:44:04 UTC from IEEE Xplore. Restrictions apply.