Nonlinear Excitation Control for Transient

Stability of Multi–Machine Power Systems

using Structure–Preserving Models ⋆
Wissam Dib ∗ Romeo Ortega ∗ David Hill ∗∗
Françoise Lamnabhi-Lagarrigue ∗
∗
Laboratoire des Signaux et Systèmes, CNRS-SUPELEC ,
Gif-sur-Yvette, 91192, France (e-mail: name@lss.supelec.fr).
∗∗
Department of Information Engineering, The Australian National
University, ACT 0200, Australia (e-mail: david.hill@anu.edu.au)

Abstract: The approach adopted in this paper for the problem of transient stability of multi-
machine power systems sees the entire network as the (structure-preserving) interconnection of
the network components, described by well known models. These structure-preserving models
preserve the identity of the network components and allow for a more realistic treatment of the
loads. Our main contribution is the explicit computation of a control law which provides transient
stabilization, i.e, prevents the system from loss of synchronism after a large perturbation. The
theory is illustrated by means of simulation on a 3-machine power system.

Keywords: Nonlinear control, power system, differential and algebraic system, stability.

1. INTRODUCTION Lyapunov function in that paper is obtained by adding

Classical research on transient stabilization of power sys-
tems has relied on the use of aggregated reduced network
models that represent the system as an n–port described
by a set of ordinary differential equations. Several exci-
tation controllers that establish Lyapunov stability of the
desired equilibrium of these models have been reported.
These nonlinear controller design techniques include feed-
back linearization Wang (1993), damping injection, as
well as, the more general, interconnection and damping
assignment passivity–based control, see Ortega (2005).
Aggregated models erase the identity of the network com-
ponents and impose an unrealistic treatment of the loads.
In this paper, we abandon the aggregated n–port view of
the network and consider the more natural and widely pop-
ular structure–preserving models (SPM), first proposed
in Bergen et al. (1981). Since these models consist of
differential algebraic equations (DAE) they require the
development of some suitably tailored tools for controller
synthesis and stability analysis. Another original feature
of the present work is that we do not aim at Lyapunov
stability, but establish instead a global convergence result.
In Giusto (2006) SPM were used to identify—in terms
of feasibility of a LMI—a class of power systems with
nonlinear (so-called ZIP) loads and leaky lines for which
a linear time–invariant controller renders the overall lin-
earized system dissipative with a (locally) positive definite
storage function, thus ensuring stability of the desired
equilibrium. Unfortunately, a full–fledged nonlinear analy-
sis of the problem was not possible due to the difficulty in
handling the complicated interdependence of the variables
appearing in the algebraic constraints of the DAEs. The
⋆ This work has been partially supported by EGIDE, PHC-FAST
Project, and by the Australian National University.
a quadratic term in the rotor angle to the classical en-
ergy function of Varaiya (1985). This quadratic term is
needed to compensate for a linear term (in rotor angle)
appearing in the energy function of Varaiya (1985) and
render the new storage function positive definite. To obtain
our global convergence result we observed in Dib (2008)
that removing the linear term from the energy function
of Varaiya (1985) and increasing the quadratic term in
bus voltages yields a function whose time derivative can
be arbitrarily assigned with a globally defined static state
feedback. Furthermore, although this new function is not
positive definite, it is bounded from below and has some
suitable radial unboundedness properties—features that
are essential to establish boundedness of trajectories. How-
ever due to the complexity of the calculations, the energy
function used describes only the synchronous machines.
Goal of this work is to show that, by exploiting the results
in Dib (2008), it is possible to construct a globally con-
vergent controller that renders attractive the ball centered
on the stable equilibrium point, with an energy function
that describes the entire electric network. The only critical
assumption required to establish this result is that the
loads are constant impedances, see Dib (2009).
It is worth mentioning that the transient stabilization
problem (in the sense of the definition given in Willems
(1974)) does not coincide with the problem of stabilization
of power system at the desired operating point. From a
practical point of view the solution of the latter problem is
preferable but its realization just after some fault can cause
the large control efforts acting on the system. Therefore,
one can use the controller which provides transient stabi-
lization, i.e., prevents the system from loss of synchronism,
as an intermediate control law and then solve the problem
of power system stabilization at the desired operating where the state variables xj := col(δj , ωj , Ej ) ∈ S ×
point which better fits the current electric power demand R × R denote the rotor angle, the rotor speed and the
and supply. quadrature axis internal e.m.f., respectively, and EFj is
the field voltage. The latter is split in two terms, EF⋆ j + vj ,
The structure of the paper is as follows. Section 2 presents
the mathematical model of the various elements compris- the first is constant and fixes the equilibrium value, while
ing the power system. Then, we formulate the control the second one is the control action. The parameters are
problem in Section 3. Section 4 contains our main “global” denoted as in Varaiya (1985), and are fairly standard. To
convergence result. Section 5 includes the application of simplify notation, we define the constants
x′dj − xqj
⊤
the proposed technique to a classical example. We wrap xdj

1
up the paper with some concluding remarks in Section 6. Y2j := , YEj := ′ , Lvj := 0 0 ,
2xqj x′dj xdj (xdj − x′dj ) τj
Caveat Due to space limitations, we have omitted the  
proofs of the Lemmas that can be found in Dib (2008). x′dj + xqj 1 1 0
YV j := , YFj := , Bu (yj ) :=  ,
Notation All vectors in the paper are column vectors, 2xqj x′dj xdj − x′dj 0 V1j
∂
even the gradient of a scalar function: ∇x = ∂x . For any    
n ∂f
function f : R → R, we define ∇zj f (z) := ∂zj (z). To 0 1
0 0 0 0
 Mj   
 Dj
simplify notation we introduce the sets Jj := − 1 0 0 = −Jj⊤ , Rj := 0 Mj2 0  ≥ 0.
  
 Mj
Mn := Sn × Rn × Rn × Sn × Rn>0 , n ∈ n̄ := {1, .., n},
  
0 0 0 0 0 τj Y1F
where Rn>0 := {x ∈ Rn | xj > 0} and S is the unitary circle. j

Notice that, in the operating region of interest, the mag-

2. STRUCTURE–PRESERVING MODELLING nitude of the bus voltages Vj is bounded away from zero
hence the division by Vj is well defined.
In this section we recall the well–known structure–
preserving model reported in Varaiya (1985). To simplify Fact 1. The synchronous machine model (2) defines an op-
the presentation of our results we assume a simplified erator ΣM M
j : (uj , yj ) described by the implicit perturbed
network topology where attached to each bus there is a port-Hamiltonian system

machine and a load. 1 Each bus, and their corresponding
ẋj =(Jj − Rj )∇xj SjM (xj , yj ) + Lvj vj + ξj ,

machine and load, have an associated identifier j ∈ n̄ := (3)
{1, .., n}. Buses are interconnected through transmission  0 =−∇ S M (x , y ) + B (y )uM − ΨM (y ),
yj j j j u j j j j
lines that are identified by the double subindex jk ∈ Ω ⊂
n̄ × n̄, indicating that the line jk connects the bus j ∈ n̄ with storage function SjM : M1 → R,
with the bus k ∈ n̄; the set avoids obvious repetitions,
e.g., if jk ∈ Ω then kj ∈ / Ω. All elements share as port 1 1 1
SjM := Mj ωj2 + YEj Ej2 + [∆j + YVj ]Vj2 − YFj EF⋆ j Ej
variables the angle θj and the magnitude Vj of the bus 2 2 2
voltage phasor yj = col(θj , Vj ) ∈ S × R>0 . Associated Y2j 2 Ej Vj
− cos 2(θj − δj )Vj − ′ cos(θj − δj ), (4)
to each bus are the active and reactive powers entering 2 xdj
the machine, the load or the transmission lines, that are
P ⊤
denoted where ξj := [0, Mmjj , 0]⊤ , ΨM
j (yj ) := [0, −∆j Vj ] , and
 M  L  
Pj Pj Pjk ∆j ≥ 0 is a key design parameter.
uM = , u L
= , u jk = , (1)
j QM
j
j QLj Qjk
respectively. Following standard convention, we take ac- Remark 1. The functions SjM defined in (4) should be
tive and reactive powers as positive when entering their contrasted with the energy functions used in Tsolas (1985),
corresponding component. see also Giusto (2006). On one hand, the latter includes
an additional term −Pmj δj . 2 On the other hand, we have
2.1 Synchronous machines model included a term ∆j Vj2 that, as will come clear below, is
essential for the construction of the control law.
Each synchronous machine is described by a set of third
order DAE’s, Varaiya (1985): Remark 2. To handle the linear term −Pmj δj in a
δ̇j =ωj , Lyapunov–like analysis we must take care of some deli-
cate theoretical issues that have, unfortunately, been over-
Mj ω̇j =Pmj − Dj ωj + PjM , looked in the literature and we discuss in detail here—
xd xdj −x′d see also discussion in Ortega (2005). Since this function
τj Ėj =− x′ j Ej + j
Vj cos(δj − θj ) + EFj ,
dj
x′d
j is not defined in S, but in R, if we look at the system as
Ej Vj evolving in Mn it will be a discontinuous function and
PjM =− sin(δj − θj ) − Y2j Vj2 sin(2(δj − θj )),
x′d
j (standard) Lyapunov arguments will not hold true. To
Ej Vj avoid this difficulty, we should consider that δj evolves
QM 2
j =(YVj − Y2j cos(2(δj − θj )))Vj − cos(δj − θj ),
x′d
j in R, instead of S. In this case, the function SjM is not
(2) lower bounded anymore, stymying the establishment of the
1 As will become clear below the derivations are also applicable 2 Removing this term induces the appearance of the constant term
for other network topologies—at the expense of a more cluttered ξj in (3), but we show in the sequel that its effect can be compensated
notation. with a suitable selection of the control law.
property of trajectory boundedness needed for LaSalle– 3. CONTROL PROBLEM
based arguments. 3
To obtain the overall model, it is convenient to group all
2.2 Loads model the algebraic constraints (3), (6), (9), (11) into the function
g : Mn → R2n :
Loads are described by the standard ZIP model: g(x, y) := ∇y S0 (x, y) + Ψ(y), (12)
PjL =PZj Vj2 + PIj Vj + P0j , where Ψ : Sn × Rn>0 → R2n is given by Ψ(y) := col(Ψj (y)),
(5) with
QL 2
j =QZj Vj + QIj Vj + Q0j ,
X
Ψj (y) = Ψjk (yj , yk ) + ΨL M
j (yj ) + Ψj (yj ),
which explicitly represent the contribution of each type of
k∈Ωj
load, i.e, constant impedance, current and power.
and S0 : Mn → R
Fact 2. The ZIP load model (5) defines an operator ΣL j : X X X
(uL , y j ) described by the implicit PCH system S0 (x, y) := SjM (xj ) + SjL (yj ) + Sjk (yj , yk ).
j
j j jk∈Ω
0 = −∇yj SjL (yj ) + Bu (yj )uL L
j − Ψj (yj ), (6) (13)
with storage function SjL : R × R>0 → R, A compact description of the system is obtained defining
QZj 2 the block–diagonal matrices
SjL (yj ) := P0j θj + V + QIj Vj + Q0j ln(Vj ), (7) 1
2 j Lv := diag{col(0, 0, )}, J := diag{Ji }, R := diag{Ri }.
where τj
⊤
ΨL 2 This allows us to rewrite the overall system as

j (yj ) := PZj Vj + PIj Vj 0 . 
ẋ=(J − R)∇x S0 (x, y) + Lv v + ξ

2.3 Transmission lines model (14)
 0=∇ S (x, y) + Ψ(y),
y 0
The transmission lines are modeled with the standard
lumped Π circuit, Anderson and Fouad (1977): where x := col(xj ), y := col(yj ), v := col(vj ) and
ξ := col(ξj ).
Pjk = Gjk Vj2 + Bjk Vj Vk sin(θj − θk )
3.1 Problem formulation
−Gjk Vj Vk cos(θj − θk ),
(8)
c
Qjk = (Bjk − Bjk )Vj2 − Bjk Vj Vk cos(θj − θk ) Assumption A1. There exists an isolated asymptotically
stable open loop equilibrium (x⋆ , y ⋆ ) of the system (14).
−Gjk Vj Vk sin(θj − θk ),
where jk ∈ Ω, while Gjk , Bjk and Bjk c
denote the lines Transient Stabilization Problem. Consider the sys-
conductance, series and shunt susceptance, respectively. tem (14) satisfying Assumptions A1. Find a control law
v = v̂(x, y) such that there exists ǫ > 0 with :
Fact 3. The transmission line model (8) defines an oper-
ator Σjk : (ujk , yj , yk ) described by an implicit (memory– Bǫ = {(x(t), y(t)); k((x(t), y(t)) − (x⋆ , y ⋆ )k ≤ ǫ}
less) PCH system
is globally attractive in closed–loop for all (x(t), y(t)) ∈
0 = −∇yj Sjk (yj , yk ) + Bu (yj )ujk − Ψjk (yj , yk ), (9) Mn , ∀t ≥ 0.
with storage function Sjk : R × R>0 × R × R>0 → R,
c
(Bjk − Bjk ) 2 3.2 Proposed solution strategy
Sjk (yj , yk ) := (Vj +Vk2 )−Bjk Vj Vk cos(θj −θk ),
2
(10) The solution to the problem stated in section 3.1 proceeds
where   along the following steps:
−V 2 + Vj Vk cos(θj − θk ) (1) Give an explicit solution of the power balance equa-
Ψjk (yj , yk ) := −Gjk  j . tions g(x, y) = 0.
Vk sin(θj − θk ) (2) Construction of a control signal that, assigning the
derivative of the Hamiltonian function S0 (x, y), en-
2.4 Bus equations sures the boundedness of all trajectories.
(3) Prove that the resulting controller is well defined and
From Kirchhoff’s laws, at each bus we have convergence is guaranteed.
0= k∈Ωj Pjk + PjM + PjL ,
P
The second step can be carried out for the model with
(11) the general ZIP loads (5). Invoking the existence of an
0= k∈Ωj Qjk + QM L
P
j + Qj , isolated local minimum of Assumption A1 and using
where Ωj := {k ∈ n̄ | ∃ jk ∈ Ω}, the set of buses that are some continuity arguments we can, therefore, conclude
linked to the bus j through some transmission line. that the proposed controller renders the neighborhood of
equilibrium locally attractive. This kind of local results
3 This unfortunate mistake is made in many papers. For instance, in are easily obtained using linearization, and known in the
Proposition 1 of Hao (2007), where the interesting idea of damping power systems community as small–signal stability. In this
injection for structure preserving models is proposed, boundedness paper we are interested in the nonlinear transient stabil-
of trajectories is never established—nor assumed. ity phenomenon, i.e., the large–signal stability problem,
therefore, the last step is indispensable. To complete it, 4.2 “Global” assignment of Ṡ0 (x, ŷ(x))
the first step is essential—unfortunately, this imposes the
restrictive requirement of constant impedance loads. Consider the system (14) satisfying Assumptions A1
and A2. Therefore, one important property of the en-
Assumption A2. Assume that the loads is described ergy function S0 (x, y) is that it is quadratic in Zjk :=
only by a constant impedance, therefore a simplified model col(ωj , Ej , Vj , Vk ) and, furthermore, bounded from below.
of the loads is: (Consequently, if S0 (x, y) is non–increasing, we can con-
clude that all signals are bounded—because Zjk will be
PjL =PZj Vj2 , bounded and θj and δj live in compact sets.) To prove
(15)
QL 2 this fact, let us define the function
j =QZj Vj . X
Remark 3. This simplification allows us to transform the S0j (x, y) := SjM (xj ) + SjL (yj ) + Sjk (yj , yk )
algebraic constraints into a set of linear equations for k∈Ωj
which we can give conditions for solvability. which describes S0 (x, y) at bus j and write it in the form
1 ⊤ ⊤
S0j = Zjk Tj (θj − δj , {θj − θk }k∈Ωj )Zjk + Zjk bj , (19)
2
4. MAIN RESULT where we have defined
 
This section contains our main “global” convergence re- Mj 0 0 0
 
sult, which is derived proceeding along the steps delineated  0 YEj cj 0 
 
in Subsection 3.2. Tj :=  ,
 0 c a d 
 j j j

4.1 Solution of g(x, y) = 0 0 0 dj Bjd

bj := col(0, −YFj EF⋆ j , 0, 0), Bjd := k∈Ωj (Bjk − Bjkc
P
) > 0,
In this subsection we present an explicit solution to the
algebraic constraints g(x, y) = 0 satisfying Assumption − cos(θj − δj ) X
A2, a result which is of interest on its own. To simplify the cj := ′ , dj := − Bjk cos(θj − θk ),
presentation we define, for j ∈ n̄, the complex variables xdj
k∈Ωj
Vj := Vj eiθj ∈ C, n
V := col(Vj )j∈n̄ ∈ C , (16) and aj := ∆j + YVj + QZj + Bjd − Y2j cos 2(θj − δj ).
and
Using the fact that (for all ∆j ≥ 0)
n
E := col(Ej )j∈n̄ ∈ R , δ := diag{δj }j∈n̄ ∈ R n×n
. ∆j + YVj + Qzj + Bjd > Y2j ,
Lemma 1. Consider the algebraic equations g(x, y) = 0 of it is possible to show that, uniformly in θj − δj , and
the power systems model (14) defined by (2), (8), (11), {θj − θk }k∈Ωj , there exists ǫj > 0 such that Tj ≥ ǫj I.
and (15). If Consequently, after some basic bounding, we prove that
1 X (YFj EF⋆ j )2
+ Q Z > c
Bjk , j ∈ n̄, (17) S0j ≥ − . (20)
x′dj j 2ǫj
k∈Ωj
Moreover, besides being lower bounded and quadratic (in
g(x, y) = 0 has a “globally” defined solution. That is, Zjk ) we prove in the paper another fundamental property
there exists a function ŷ : Sn × Rn → Sn × Rn>0 such that of this function, namely, that its derivative restricted to the
g(x, ŷ(x)) = 0. Furthermore, this function can be written set g(x, y) = 0, can be arbitrarily assigned with a suitable
in the form selection of the control v. Towards this end, compute
V = W (δ)E, (18)
Ṡ0 (x, y) = −∇⊤ ˜ ⊤ −1
v + ∇⊤
x S0 R∇x S0 + ξ(x, y) + ∇E S0 τ y S0 ẏ
where W : Rn×n → Cn×n is bounded and invertible, with (21)
elements are rational functions of cos(δj ) and sin(δj ). See P
Dib (2009) for more details. where ξ̃(x, y) := j∈n̄ ωj Pmj ∈ R and τ := diag{τj }j∈n̄ ∈
Rn×n . Lemma 1 shows that the set g(x, y) = 0 is equivalent
to V = W (δ)E. Therefore, to evaluate ẏ it is convenient
Remark 4. Note that condition (17) is always verified and
to express the Hamiltonian function (13) in terms of the
realistic, since we are considering the low voltage terminals
complex variables V defined in (16). Hence, noticing that
of the generators, which is usually connected through a
single step-up transformer to the network. Also, it is clear Vj2 cos 2(θj − δj ) = Re{e−2iδj Vj2 },
that the construction of ŷ directly follows from (16) and we define
(18), and is omitted for brevity.
1 ⊤
w M w + E ⊤ YE E + VH (YV + ∆) V


S0C (x, V ) :=
Remark 5. Since the solution of the algebraic equations 2
g(x, y) = 0 is globally defined, and it will be used in 1
− Re{EH X V} − Re{V⊤ Y2 e−2iδ V} − E ⊤ YF EF
the design of the controller, we overcome the classical 2
assumption made about the invertibility of ∇y g(x, y), i.e. 1 H 1 H d
+ V Qz V + V (B − B) V,
the continuity of the trajectories restricted to g(x, y) = 0 2 2
Hao (2007). where
 M := diag{Mj }, YE := diag{YEj }, YV := diag{YVj }, zero, we can easily select a control law v to cancel the
1 sign indefinite term Ξ and make Ṡ0 ≤ 0 or even assign an
∆ := diag{∆j }, X := diag{ ′ }, Y2 := diag{Y2j }, arbitrary function to Ṡ0 .
xdj
Proposition 1. Consider the power systems model (14)
YF := diag{YFj }, EF := col(EF⋆ j ), QZ := diag{QZj },
with Assumptions A1 and A2 and the Hamiltonian func-
B d := diag{Bjd }, B := {Bjk }, G := {Gjk }, tion (13). There exists ∆minj > 0 such that, for all ∆j ≥
min
are defined in Rn×n with j, k ∈ n̄, and Bjk = Gjk = 0 if ∆j we have
k 6∈ Ωj . L⊤ (x, ŷ(x))E > 0 for all x ∈ Sn × Rn × Rn>0 ,
The following lemma is instrumental to compute the
required derivative. where L⊤ is given in (23). Therefore, the globally defined
control law
Lemma 2. Consider the quadratic function f : R×R → R, M
v = −2 ⊤ τ E (24)
f (µ1 , µ2 ) = c1 µ21 + 2c3 µ1 µ2 + c2 µ22 , L E
with ci ∈ R and µ1 , µ2 : R → R. Define z = µ1 + iµ2 ∈ C ensures Ṡ0 ≤ −∇⊤x S0 R∇x S0 − β where M ≥ |Ξ| belongs
and the function fC : C → R such that to class K and β > 0.
fC (z) = f (µ1 , µ2 ). Proposition 2. Consider the power systems model (14)
Then, with Assumptions A1 and A2 and the Hamiltonian func-
 ∗  tion (13). There exists ∆min > 0 such that, for all ∆j ≥
∂fC j
f˙ = Re ż , ∆jmin
we have
∂z
where ∂f ∂fC ∂fC L⊤ (x, ŷ(x))E > 0 for all x ∈ Sn × Rn × Rn>0 ,
∂z := ∂µ1 + i ∂µ2 .
C

where L⊤ is given in (23). Therefore, for any function

From the lemma it is clear that, to compute
n the time
o α : Mn → R, the globally defined control law
H
derivative of S0C , we require the term Re (∇V S0C ) V̇ . 1 
−λ [S0 (x, y) − S0⋆ ] + ∇⊤

v= ⊤ x S0 R∇x S0 − Ξ τ E
It is easy to see that L E
(25)
∇V S0C = (YV + ∆ + Qz + B d − B) V −Y2 e2iδ V∗ −X E, ensures
where (·)∗ denotes complex conjugation. From Dib (2008), d
we know that the algebraic constraints g(x, y) = 0 written [S0 (x, y) − S0⋆ ] = −λ[S0 (x, y) − S0⋆ ],
dt
in terms of the complex variables are equivalent to the where S0⋆ := S0 (x⋆ , y ⋆ ), and λ > 0.
following equation
A0 V∗ −X E∗ −Y2 e−2iδ V = 0. Proof. By definition,
L⊤ E = Re (∇V S0C )H W (δ)E + ∇⊤

with A0 := YV + QZ − iPZ + B d − iGd − B + iG ∈ Cn×n , E S0 E. (26)
PZ := diag{PZj }, and Gd := diag{Gdj }. Substituting the Let us consider the first term. Since V = W (δ)E and
complex conjugate of the latter in ∇V S0C above we get ∇V S0C = D V we have
∇V S0C = (YV + ∆ + Qz + B d − B − A∗0 ) V, D + D∗
Re (∇V S0C )H W E = Re VH D V = VH
 
V
and by definition of A0 , we get ∇V S0C = D V, where 2
D := ∆ + i(−PZ − Gd + G) ∈ Cn×n . The matrix D is symmetric (not Hermitian self conjugate).
Therefore,
Let us now compute V̇. In Lemma 1 it is shown that D + D∗
V = W E, where W : Rn×n → Cn×n is bounded and = ∆,
invertible. Therefore, 2
The quadratic form above can then be made arbitrarily
V̇ = Ẇ E + W Ė large by choosing a large ∆ > 0.
The function Ẇ depends on δ and ω, but is independent Using again V = W E and invertibility of W we see that
of v, while Ė will bring along terms on v. We now come the second term in (26) is also a quadratic function of V,
back to Ṡ0 , that takes the form that can be written in the form
n o
⋆
Ṡ0 = −∇⊤ S0 R∇x S0 +ξ+∇
x
˜ ⊤ S0 τ −1 v+Re (∇V S0 )H V̇ .
E C
∇⊤ H H
E S0 E = V S(δ) V + Re(V s(δ)EF ),

That, replacing the computations above, can be compactly for some suitable matrices S, s : Rn×n → Cn×n . From
written as boundedness of W −1 we have that S and s are also bounded
and we can conclude that, throughout Mn , the first term
Ṡ0 = −∇⊤ ⊤
x S0 R∇x S0 + Ξ(δ, w, E, V) + L (δ, w, E, V)τ
−1
v in (26) can be made strictly greater than the second.
(22) Therefore, the denominator in (24) and (25) is always
where we defined the (real valued) functions larger than zero, completing the claim.
n h io
H −1
Ξ := ξ̃ + Re (∇V S0C ) Ẇ E −W (τ YF ) ∇E S0 , Remark 6. Proposition 1 ensures that Ṡ0 ≤ 0, therefore
n n
H
oo since S0 is bounded from below, this implies that all signals
L⊤ := ∇⊤ S
E 0 + Re (∇ S
V 0C ) W . (23) are bounded and the aim is achieved. In addition, a prac-
tically interesting property of the control law (24) is that
Let us take a brief respite to analyze (22). It is clear that, it is “almost” decentralized. Moreover, although all signals
wherever the vector L(δ, w, E, V) is bounded away from are bounded with Proposition 1, we propose in Proposition
2 the controller (25) in order to force all trajectories to
converge to the level set of the Hamiltonian function that
contains the stable equilibrium point, and henceforth all
trajectories will be closer to the stable equilibrium point.
Remark 7. An interesting interpretation concerning the
design parameter ∆ is that it can be assimilated to Qz
in the energy function. This means that by increasing the
reactive power in the network, we are able the stabilize the
system after a large perturbation using the specific control
law (24) or (25).
5. A BENCHMARK SIMULATION EXAMPLE
We consider here the classical 3-machines, 9-buses system
considered in Anderson and Fouad (1977). We assume
that the active and reactive components of the loads have
constant impedance characteristics. Computations were
done with the software package PSAT Milano (2005).
Using the controller of Proposition 2, we analyze the
response of (14) to a short circuit which consists of a zero-
impedance three phase fault on the transmission line. The
fault is introduced at t = 0.5 s and removed after a certain
time (called the clearing time and denoted tcl ), after which
the system is back to its pre-disturbance topology.
To tune the controller there is a compromise between the
choices of ∆ that, as indicated in Proposition 2, should
be big enough to ensure that the denominator of the
controller will stay away from zero, and λ that determines
the speed of convergence to the desired level set. Indeed,
∆ appears in S0 as ∆V 2 where V represents, in some way,
the perturbation. Then, the bigger ∆ is, the bigger S0 will
be in the transient phase, and we have to decrease λ to
eliminate impulsive responses in the controller during the
perturbation.
Fig. 1 depicts the transient response of the nonlinear
systems with tcl = 0.1 s, ∆j = 10, j = 1, 2, 3, and λ = 3.
As it can be seen, the controller is able to significantly
improve the system’s stability and provide damping.
0.35
0.3
0.25
0.2
0.15
rad/s
0.1
0.05
0 Automat. Contr., vol. 50, no. 1, pp. 60–75, Jan. 2005.
−0.05
−0.1
0 2 4
Time (s)
6 8 10
Fig. 1. Speed w1 . Open loop (dashed), closed loop (contin-
uous).
6. CONCLUSIONS AND FUTURE WORK
We have presented in this paper an excitation controller
to improve the transient stabilization properties of multi–
machine power systems described by SPM with leaky
lines including capacitive effects. Our main contribution
is the explicit computation of a control law that ensures
boundedness of all trajectories. Although the proposed
controller does not provide asymptotic convergence to
the desired equilibrium point the goal of transient sta-
bilization problem is successfully achieved. To the best
of our knowledge, no equivalent result is available in the
literature at this level of generality. The usefulness of the
technique for synthesis was illustrated with its application
to a classical example. Similarly to most developments
reported by the control theory community on the tran-
sient stability problem, it is clear that the complexity of
the proposed controller—as well as its high sensitivity to
the system parameters and the assumption of full state
measurement—severely stymies the practical application
of this result. This kind of work pertains, however, to
the realm of fundamental research where basic issues like
existence of solutions are addressed.
REFERENCES
P. M. Anderson and A. A. Fouad. Power Systems Control
and Stability. The Iowa State University Press, 1977.
A. Bergen, and D. Hill. A structure preserving model for
power system stability analysis. IEEE Trans on Power
App. Syst., vol. PAS-100, no. 1, pp. 25–35, 1981.
W. Dib, R. Ortega, A. Barabanov, and F. Lamnabhi-
Lagarrigue. A “Globally” Convergent Controller for
Transient Stability of Multi–Machine Power Systems
using Structure–Preserving Models. 45th IEEE Conf.
on Decision and Control, Cancun, December 2008.
W. Dib, A. Barabanov, R. Ortega, and F. Lamnabhi-
Lagarrigue. An Explicit Solution of the Power Balance
Equations of Structure Preserving Power System Mod-
els. To be published in IEEE Trans. on Power Systems,
2009.
A. Giusto, R. Ortega and A. Stankovic. On transient
stabilization of power systems: a power–shaping solution
for structure–preserving models. 45th IEEE Conference
on Decision and Control, San Diego, CA, USA, 2006.
J. Hao, C. Chen, L. Shi and J. Wang. Nonlinear decen-
tralized disturbance attenuation excitation control for
power systems with nonlinear loads based on the hamil-
tonian theory. IEEE Trans. on Energy Conversion, vol.
22, no. 2, pp. 316–324, June 2007.
F. Milano. An open source power system analysis toolbox.
IEEE Trans on Power Systems, vol. 20, no. 3, pp. 1199–
1206, August 2005.
R. Ortega, M. Galaz, A. Astolfi, Y. Sun and T. Shen.
Transient stabilization of multimachine power systems
with nontrivial transfer conductances. IEEE Trans.
N. A. Tsolas, A. Arapostathis and P. P. Varaiya. A
structure preserving energy function for power system
transient stability analysis. IEEE Trans. Circuits Syst.,
vol. 32, pp. 1041–1049, 1985.
P. Varaiya, F. Wu and R. Chen. Direct methods for
transient stability analysis of power systems: Recent
results. Proc IEEE, vol. 73, no. 12, pp. 1703–1714, 1985.
Y. Wang, D.J. Hill, R.H. Middleton and L. Gao. Transient
stability enhancement and voltage regulation of power
system. IEEE Trans on Power Systems, vol. 8, pp. 620–
627, 1993.
J. L. Willems. A partial stability approach to the problem
of transient power system stability. Int. J. of Control,
vol. 19, no. 1, pp. 1–14, 1974.