EE549 - Power System Dynamics and Control

Small Signal Stability

S. Sivasubramani

Associate Professor
Electrical Engineering Department
Indian Institute of Technology Patna
www.iitp.ac.in/~siva

Siva (IIT P) EE549 1 / 97

Small Signal Stability

It is the ability of the power system to maintain synchronism when

subjected to small disturbances.

Small disturbances can be a small change in load or generation or line

tripping.
Since the disturbance is small, the nonlinear differential algebraic equa-
tions that describe the system may be linearized around the steady
operating point.
Instability can be of two forms.
1 steady increase in rotor angle due to lack of synchronizing torque
2 rotor oscillations of increasing amplitude due to lack of sufficient
damping torque.
In modern power systems, the instability is mainly due to insufficient
damping.

Siva (IIT P) EE549 2 / 97

There are three modes of oscillations due to small disturbance.
1 Local modes of oscillations are due to a single generator or group of
generators oscillating against the rest of the system.
2 Intra-plant modes of oscillations are due to oscillations among the gen-
erators in the same plant. The typical frequencies of oscillations of
local and intra plant modes are in the range of 1 Hz to 2 Hz.
3 Inter area modes of oscillations are due to a group of generators in
one area oscillating together against another group of generators in
a different area. The typical frequency range of inter area mode of
oscillations are 0.1 Hz to 0.8 Hz.

Siva (IIT P) EE549 3 / 97

There are three modes of oscillations due to small disturbance.
1 Local modes of oscillations are due to a single generator or group of
generators oscillating against the rest of the system.
2 Intra-plant modes of oscillations are due to oscillations among the gen-
erators in the same plant. The typical frequencies of oscillations of
local and intra plant modes are in the range of 1 Hz to 2 Hz.
3 Inter area modes of oscillations are due to a group of generators in
one area oscillating together against another group of generators in
a different area. The typical frequency range of inter area mode of
oscillations are 0.1 Hz to 0.8 Hz.

Local, intra-plant and inter-area are electromechanical modes of os-

cillations in which the rotor angle and the speed of the generators
oscillate.
Apart from electromechanical modes of oscillations there can be control
modes, due to lack of proper tuning of controller like voltage regula-
tor of an excitation system, and torsional mode of oscillations due to
oscillations in the turbine-generator shafts.
Siva (IIT P) EE549 3 / 97
Fundamental Concepts of Stability of Dynamic Systems

The behavior of a dynamic system such as a power system may be

represented by a set of n first order nonlinear ODEs as follows:

ẋi = fi (x1 , x2 , · · · , xn ; u1 , u2 , · · · , ur ; t) i = 1, 2, · · · , n (1)

where n is the order of the system and r is the number of inputs. It can be
written as
ẋ = f(x, u, t) (2)
where      
x1 u1 f1
 x2  u2   f2 
x=. u=. f =.
     
 ..   ..   .. 
xn ur fn

Siva (IIT P) EE549 4 / 97

If the state variables are not explicit functions of time, the system is said
to be autonomous.
ẋ = f(x, u) (3)
The output vector can also be written as

y = g(x, u) (4)

where   

y1 g1
 y2   g2 
y= .  g= . 
   
 ..   .. 
ym gm

Siva (IIT P) EE549 5 / 97

Equilibrium Points

At equilibrium,
ẋ = f(x0 , u0 ) = 0 (5)
where x0 is the state vector x at the equilibrium point.
If f is a set of linear functions, the system is linear. Linear systems
have only one equilibrium state.
If f is a set of nonlinear functions, the system is nonlinear. Nonlinear
systems may have more than one equilibrium state.

Siva (IIT P) EE549 6 / 97

Linearization

Let x0 be the initial state vector and u0 be the input vector corresponding
to the equilibrium point.

ẋ = f(x0 , u0 ) = 0 (6)

Let us perturb the system from the above state.

x = x0 + ∆x u = u0 + ∆u

where ∆ denotes a small deviation.

At the new state,

ẋ = x˙0 + ∆ẋ
(7)
= f(x0 + ∆x, u0 + ∆u)

Siva (IIT P) EE549 7 / 97

As the perturbations are assumed to be small, the nonlinear functions
f(x, u) can be expressed in terms of Taylor’s series expansion. By
neglecting the higher order terms,

ẋi =ẋi0 + ∆ẋi = f (x0 + ∆x, u0 + ∆u)

∂fi ∂fi
fi (x0 , u0 ) + ∆x1 + · · · + ∆xn
∂x1 ∂xn
∂fi ∂fi
+ ∆u1 + · · · + ∆ur
∂u1 ∂ur
Since ẋi0 = fi (x0 , u0 ),
∂fi ∂fi ∂fi ∂fi
∆ẋi = ∆x1 + · · · + ∆xn + ∆u1 + · · · + ∆ur
∂x1 ∂xn ∂u1 ∂ur
with i = 1, 2, · · · , n. Similarly,
∂gj ∂gj ∂gj ∂gj
∆yj = ∆x1 + · · · + ∆xn + ∆u1 + · · · + ∆ur
∂x1 ∂xn ∂u1 ∂ur
with j = 1, 2, · · · , m.
Siva (IIT P) EE549 8 / 97
Therefore, the linearized forms of equations (3) and (4) are

∆ẋ = A∆x + B∆u (8)

∆y = C∆x + D∆u (9)

where
 ∂f1 ∂f1   ∂f1 ∂f1 
∂x1 ··· ∂xn ∂u1 ··· ∂ur
A =  ... .. ..  B =  ... .. .. 
 
. .  . . 
∂fn ∂fn ∂fn ∂fn
∂x1 ··· ∂xn ∂u1 ··· ∂ur
 ∂g1 ∂gm   ∂g1 ∂g1  (10)
∂x1 ··· ∂xn ∂u1 ··· ∂ur
C =  ... .. ..  D =  ... .. .. 
 
. .  . . 
∂gm ∂gm ∂gm ∂gm
∂x1 ··· ∂xn ∂u1 ··· ∂ur

Siva (IIT P) EE549 9 / 97

 ∆x = state vector of dimension n
∆y = output vector of dimension m
∆u = input vector of dimension r
A = state or plant matrix of size n × n
B = control or input matrix of size n × n
C = output matrix of size m × n
D = feed forward matrix of size m × r

By taking the Laplace transform of the equations (8) and (9),

s∆x(s) − ∆x(0) = A∆x(s) + B∆u(s) (11)

∆y(s) = C∆x(s) + D∆u(s) (12)

Siva (IIT P) EE549 10 / 97

Rearranging the equation (11),

(sI − A)∆x(s) = ∆x(0) + B∆u(s)

Hence,
∆x(s) = (sI − A)−1 [∆x(0) + B∆u(s)] (13)
and
∆y(s) = C(sI − A)−1 [∆x(0) + B∆u(s)] + D∆u(s) (14)
The poles of ∆x(s) and ∆y(s) are the roots of the equation.

det(sI − A) = 0 (15)

The values of s which satisfy the above equation are known as the
eigenvalues of matrix A. Equation (15) is referred to as the characteristic
equation of A.

Siva (IIT P) EE549 11 / 97

Eigenvalues

The eigenvalues of a matrix are given by the values of the scalar parameter
λ for which there exist non-trivial solutions to the following equation.

Aφ = λφ (16)

where

A is an n × n matrix
φ is an n × 1 vector

To find the eigenvalues,

(A − λI)φ = 0 (17)
For a non-trivial solution,
|A − λI| = 0 (18)

Siva (IIT P) EE549 12 / 97

The eigenvalues may be real or complex.
If A is real, complex eigenvalues always occur in conjugate pairs.
Similar matrices have identical eigenvalues.
A matrix and its transpose have the same eigenvalues.

Siva (IIT P) EE549 13 / 97

Eigenvectors

For any eigenvalue λi ,

Aφi = λi φi i = 1, 2, · · · , n (19)

The eigenvector φi is the right eigenvector and has the form

 
φ1i
φ2i 
φi =  
· · ·
φni

Similarly, the n-row vector ψi which satisfies

ψi A = λi ψi i = 1, 2, · · · , n (20)

is called the left eigenvector for the eigenvalue λi .

Siva (IIT P) EE549 14 / 97

The left and right eigenvectors corresponding to different eigenvalues
are orthogonal.
ψ j φi = 0
For the same eigenvalue,
ψi φi = Ci
where Ci is a non-zero constant.
If the eigenvectors are normalized,

ψi φi = 1

Siva (IIT P) EE549 15 / 97

Let the left and right eigenvector matrices be defined as
 
Φ = φ1 φ2 · · · φn
T (21)
Ψ = ψ1 T ψ2 T · · · ψn T


From the properties of eigenvectors,

ΦΨ = I (22)

Ψ = Φ−1 (23)
Let an n dimensional diagonal matrix of eigenvalues be defined as
 
λ1 ··· 0
 .. .. .
Λ= . . ..  (24)
0 · · · λn

Siva (IIT P) EE549 16 / 97

In terms of these matrices, the equation (19) may be expressed as follows:

AΦ = ΦΛ (25)

The above equation can be written as

Λ = Φ−1 AΦ = ΨAΦ (26)

Φ, Ψ and Λ are of size n × n matrices and are called as modal matrices.

Siva (IIT P) EE549 17 / 97

Stability of a Homogeneous System

Let us take a homogeneous system which does not have an external input.
Therefore, the response is a natural response.

∆ẋ = A∆x (27)

Since there is cross coupling between states, it is difficult to isolate

those parameters that influence the response in a significant way.
To eliminate the cross coupling, let us consider a new state vector z.
∆x = Φz (28)
where Φ is the modal matrix containing right eigenvectors of A as its
columns. From (27) and (28),

Φż = AΦz (29)

Siva (IIT P) EE549 18 / 97

The new state equation can be written as

ż = Φ−1 AΦz (30)

In view of (26), the above equation becomes

ż = Λz (31)

The important difference between (27) and (31) is that Λ is a diagonal

matrix whereas A, in general, is non-diagonal.
Equation (31) is a set of decoupled first order equations.

żi = λi zi i = 1, 2, · · · , n (32)

The above equation is a simple first order differential equation whose

solution is
zi (t) = zi (0)e λi t (33)
where zi (0) is the initial value of zi .

Siva (IIT P) EE549 19 / 97

From (33), the stability of the system can be assessed as follows:
1 If all the eigen values are on the left half of s-plane that is the real part
of the eigenvalues should be having a negative value then the system
is stable. This is called as asymptotic stability.
2 If at least one of the eigen values is on the right side of s-plane that is
the real part of the eigen value is positive then the system is unstable.
This is called as aperiodic instability.
3 The eigen values of a system with a real state matrix have complex
conjugate pairs of eigen values. If at least one of the complex conjugate
pair lies on the imaginary axis of the s-plane then the system becomes
oscillatory. This is called as oscillatory instability.
4 If at least one eigenvalues is on the origin then the system stability
cannot be assessed.

Siva (IIT P) EE549 20 / 97

From (28), the response in terms of the original state vector is

∆x(t) = Φz(t)
 
z1 (t)
z2 (t)
 (34)
 
= φ1 φ2 · · · φn  . 
 .. 
zn (t)

From (33),
n
X
∆x(t) = φi zi (0)e λi t (35)
i=1

From (34), we have

z(t) = Φ−1 ∆x(t)

(36)
= Ψ∆x(t)

Siva (IIT P) EE549 21 / 97

Therefore
zi (t) = ψi ∆x(t) (37)
With t = 0,
zi (0) = ψi ∆x(0) (38)
Equation (35) can be written as
n
X
∆x(t) = φi ci e λi t (39)
i=1

where ci = ψi ∆x(0). The time response of i th state is given by

∆xi (t) = φi1 c1 e λ1 t + φi2 c2 e λ2 t + · · · + φin cn e λn t (40)

Siva (IIT P) EE549 22 / 97

 ∆x(t) = Φz(t)
 
= φ1 φ2 · · · φn z(t)

and
z(t) = Ψ∆x(t)
T
= ψ1 T ψ2 T · · · ψn T

∆x(t)

∆x1 , ∆x2 , · · · , ∆xn are the original state variables. z1 , z2 , · · · , zn are

the transformed state variables.
The right eigenvector gives the mode shape, i.e., the relative activity
of the state variables when a particular mode is excited.
The left eigenvector identifies which combination of the original state
variables displays only the i th mode.

Siva (IIT P) EE549 23 / 97

The combined effect of activity and contribution of a state variable in a
particular mode can be represented as a matrix called as participation
factor. The participation factor matrix is given as
 
P = p1 p2 · · · pn

where    
p1i φ1i ψi1
p2i  φ2i ψi2 
pi =  .  =  . 
   
 ..   .. 
pni φni ψin

The element pki = φki ψik is termed as the participation factor. It is a

measure of k th state variable in the i th mode and vice versa.
Since the eigenvectors are normalized,
n
X n
X
pki = 1; pki = 1
i=1 k=1

Siva (IIT P) EE549 24 / 97

Controllability and Observability

Equations (8) and (9) are

∆ẋ = A∆x + B∆u

∆y = C∆x + D∆u

Expressing the above equations in terms of the transformed state variables

z,

Φż = AΦz + B∆u

∆y = CΦz + D∆u

The state equations in the decoupled form can be written as

ż = Λz + B0 ∆u (41)
0
∆y = C z + D∆u (42)

Siva (IIT P) EE549 25 / 97

where

B0 = Φ−1 B (43)
0
C = CΦ (44)

if the i th row of B0 is zero, the inputs have no effect on the i th mode.

Hence, the i th mode is uncontrollable.
if the i th column of C0 is zero, zi does not contribute to the formation
of outputs. Hence, the i th mode is unobservable.
B0 of size n × r is referred to as the mode controllability matrix.
C0 of size m × n is referred to as the mode observability matrix.

Siva (IIT P) EE549 26 / 97

Single Machine Connected to an Infinite Bus (SMIB)

π
(Id + Iq )e (δ− 2 ) Re Xe
V∞ ∠0◦
It ∠γ

π
(Vd + Vq )e (δ− 2 ) = Vt e α

Figure: SMIB

Siva (IIT P) EE549 27 / 97

Small Signal Stability of a SMIB

Let us analyze with varying degrees of models.

1 Constant flux linkage model - Classical Model
2 Flux decay model - Without AVR
3 Flux decay model - With AVR
4 Flux decay model - With AVR and PSS

Siva (IIT P) EE549 28 / 97

Small Signal Stability of a SMIB - Classical Model

Let the input mechanical torque Tm be constant.

Therefore, turbine and speed governor dynamics are not needed.
Flux linkages are assumed to constant. Hence, field and exciter
dynamics are neglected.
Let the generator be represented using the classical model.

Siva (IIT P) EE549 29 / 97

The differential equations of this model are

dδ
= ω − ωs (45)
dt
2H dω
= Tm − Te − D(ω − ωbase ) (46)
ωs dt
The equivalent circuit of a SMIB system using this model is shown here.

Rs Xd0 Vt ∠α Re Xe
It ∠γ

+
E 0 ∠δ V∞ ∠0◦
−

Figure: SMIB equivalent circuit using the classical model

Siva (IIT P) EE549 30 / 97

With resistances neglected, the air-gap power (Pe ) is equal to the terminal
power (Pt ). In per unit,

E 0 V∞ sin δ
Te = Pe = (47)
XT

where XT = Xd0 + Xe . Linearizing around the equilibrium (δ0 ) for a small

disturbance gives

d∆δ
= ∆ω (48)
dt
2H d∆ω
= −∆Te − D∆ω (49)
ωs dt

From (47),
E 0 V∞ cos δ0
∆Te = ∆δ = Ks ∆δ (50)
XT

Siva (IIT P) EE549 31 / 97

 0
where Ks = E V∞ cos δ0
XT is the synchronizing coefficient. Expressing them in
the state space form
    
∆δ̇ 0 1 ∆δ
= ωs ωs (51)
∆ω̇ − 2H Ks − 2H D ∆ω

On taking the Laplace Transform and finding the characteristics equation,

ωs D ωs K s
s2 + s+ =0 (52)
2H 2H
By comparing it with the general form, s 2 + 2ζωn s + ωn2 , the undamped
natural frequency is r
ωs Ks
ωn = rad/s (53)
2H
and the damping ratio is
√
1 ωs D 1 D ωs
ζ= = √ (54)
2 2Hωn 2 2HKs
Siva (IIT P) EE549 32 / 97
The time responses of the state variables are

∆δ(t) = φ11 c1 e λ1 t + φ12 c2 e λ2 t rad (55)

λ1 t λ2 t
∆ω(t) = φ21 c1 e + φ22 c2 e rad/s (56)

where φ is the right eigenvector matrix of the system matrix A and λ1 and
λ2 are eigenvalues of it. c1 and c2 are calculated as follows:
   
c1 ∆δ(0)
=ψ (57)
c2 ∆ω(0)

ψ is the left eigenvector matrix of A

Siva (IIT P) EE549 33 / 97

Example 12.2 from P. Kundur : Figure show the system representation
applicable to a thermal generating station consisting of four 555 MVA, 24
kV, 60 Hz units.

The reactances are in per unit on 2220 MVA, 24 kV base on LT side.

Analyze the small signal stability characteristics of the system about the
stead operating condition following the loss of line 2. The post fault
system condition in per unit on the 2220 MVA, 24 kV base is as follows:

P = 0.9 Q = 0.3 Vt = 1.0∠36◦ V∞ = 0.995∠0◦

Siva (IIT P) EE549 34 / 97
Use the classical model. Xd0 = 0.3 , H = 3.5 s and D = 0. Determine the
time response if at t = 0, ∆δ = 5◦ and ∆ω = 0

0.3 It ∠γ 0.15 0.5

Vt = 1.0∠36◦

E 0 ∠δ 0.995∠0◦

S∗ 0.9 − 0.3
I¯t = ∗ = = 0.95∠17.56◦
V 1.0∠−36◦
Ē 0 = V̄t + I¯t × 0.3 = 1.0∠36◦ + 0.95∠17.56◦ × 0.3 = 1.123∠49.91◦
The angle of E 0 with respect to the infinite bus is

δ0 = 49.91◦
Siva (IIT P) EE549 35 / 97
The synchronizing torque coefficient is
E 0 V∞ 1.123 × 0.995
Ks = cos δ0 = cos 49.91◦ = 0.7574
XT 0.95
The system matrix A from the linearized equations is
   
0 1 0 1
A= ωs ωs =
− 2H Ks − 2H D −40.79 0

The eigenvalues are

λ1 , λ2 = ±6.39
The right eigenvector modal matrix is
 
−0.1547 +0.1547
Φ=
0.99 0.99

The left eigenvector modal matrix is

 
−1 3.2323 0.5061
Ψ=Φ =
−3.2323 0.5061
Siva (IIT P) EE549 36 / 97
The participation matrix is
   
φ11 ψ11 φ12 ψ21 0.5000 0.5000
P= =
φ21 ψ12 φ22 ψ22 0.5000 0.5000
The time response is
∆δ(t) = φ11 c1 e λ1 t + φ12 c2 e λ2 t rad
λ1 t λ2 t
∆ω(t) = φ21 c1 e + φ22 c2 e rad/s
c1 and c2 are calculated as follows:
        
c1 ∆δ(0) 3.2323 0.5061 0.0873 0.2822
=ψ = =
c2 ∆ω(0) −3.2323 0.5061 0 −0.2822
The time response of rotor angle deviation is
∆δ(t) = 0.0873 cos(6.35t) rad
The time response of speed deviation is
∆ω(t) = −0.5644 sin(6.35t) rad/s
Siva (IIT P) EE549 37 / 97
Small Signal Stability of a SMIB - Including Field Circuit
Dynamics

Let the input mechanical torque Tm be constant.

Therefore, turbine and speed governor dynamics are not needed.
Exciter output voltage is constant. Hence, exciter dynamics are not
needed.
Let the generator be represented using the flux decay model.

Siva (IIT P) EE549 38 / 97

The DAE (Differential Algebraic Equations) of the flux decay model
excluding exciter dynamics are

0
dEq0
Td0 = −Eq0 − (Xd − Xd0 )Id + Efd (58)
dt
dδ
= ω − ωs (59)
dt
2H dω
= Tm − Te − D(ω − ωbase ) (60)
ωs dt

Vq + Rs Iq = −Xd0 Id + Eq0 (61)

Vd + Rs Id = Xq Iq (62)

Siva (IIT P) EE549 39 / 97

Where
Te = Real {((Xq − Xd0 )Iq + Eq0 )(Id + Iq )∗ }
= (Xq − Xd0 )Iq Id + Eq0 Iq

From the figure SMIB,

π
(δ− π2 ) (Vd + Vq )e (δ− 2 ) − V∞ e ∠0
(Id + Iq )e = (63)
Re + Xe

On separating the real and imaginary parts from (63),

Id Re − Iq Xe = Vd − V∞ sin δ (64)
Iq Re + Id Xe = Vq − V∞ cos δ (65)

Siva (IIT P) EE549 40 / 97

Let the steady state operating point values of (Eq0 , δ, Efd , Id , Iq , Vd , Vq ) be
0 ,δ ,E
(Eq0 0 fd0 , Id0 , Iq0 , Vd0 , Vq0 ). Linearizing equations (61), (62), (64) and
(65) for a small perturbation ∆ around the steady operating point and
neglecting Rs give
      
∆Vd 0 Xq ∆Id 0
= + (66)
∆Vq −Xd0 0 ∆Iq ∆Eq0
      
∆Vd R −Xe ∆Id V∞ cos δ0
= e + ∆δ (67)
∆Vq Xe Re ∆Iq −V∞ sin δ0

Siva (IIT P) EE549 41 / 97

Let the steady state operating point values of (Eq0 , δ, Efd , Id , Iq , Vd , Vq ) be
0 ,δ ,E
(Eq0 0 fd0 , Id0 , Iq0 , Vd0 , Vq0 ). Linearizing equations (61), (62), (64) and
(65) for a small perturbation ∆ around the steady operating point and
neglecting Rs give
      
∆Vd 0 Xq ∆Id 0
= + (66)
∆Vq −Xd0 0 ∆Iq ∆Eq0
      
∆Vd R −Xe ∆Id V∞ cos δ0
= e + ∆δ (67)
∆Vq Xe Re ∆Iq −V∞ sin δ0

By equating (66) and (67),

         
0 Xq ∆Id 0 R −Xe ∆Id V∞ cos δ0
0 + 0 = e + ∆δ
−Xd 0 ∆Iq ∆Eq Xe Re ∆Iq −V∞ sin δ0
(68)

Siva (IIT P) EE549 41 / 97

Solving for ∆Id and ∆Iq from (68),
   −1     
∆Id Re −(Xe + Xq ) 0 −V∞ cos δ0
= + ∆δ
∆Iq (Xe + Xd0 ) Re ∆Eq0 V∞ sin δ0
 
1 Re (Xe + Xq )
= 2
Re + (Xe + Xq )(Xe + Xd0 ) −(Xe + Xd0 ) Re
  
−V∞ cos δ0 0 ∆δ
V∞ sin δ0 1 ∆Eq0
1
= 2
Re + (Xe + Xq )(Xe + Xd0 )
  
V∞ (Xe + Xq ) sin δ0 − Re V∞ cos δ0 (Xe + Xq ) ∆δ
V∞ (Xe + Xd0 ) cos δ0 + Re V∞ sin δ0 Re ∆Eq0
(69)

Siva (IIT P) EE549 42 / 97

On linearizing differential equations (58)-(60),
0
∆E˙q − T10 0 0 ∆Eq0
   
d0
 ∆δ̇  =  0 0 1   ∆δ 
ωbase ωbase ∆ω
∆ω̇ − 2H Iq0 0 − 2H D
 
− T10 (Xd − Xd0 ) 0  
d0  ∆Id
+ (70)

0 0
∆Iq

− ω2H
base
(Xq − Xd0 )Iq0 − ω2H
base
((Xq − Xd0 )Id0 + Eq00 )

 1
0

0
Td0
 
∆Efd
+ 0
 0 
ωbase ∆Tm
0 2H

Siva (IIT P) EE549 43 / 97

On linearizing differential equations (58)-(60),
0
∆E˙q − T10 0 0 ∆Eq0
   
d0
 ∆δ̇  =  0 0 1   ∆δ 
ωbase ωbase ∆ω
∆ω̇ − 2H Iq0 0 − 2H D
 
− T10 (Xd − Xd0 ) 0  
d0  ∆Id
+ (70)

0 0
∆Iq

− ω2H
base
(Xq − Xd0 )Iq0 − ω2H
base
((Xq − Xd0 )Id0 + Eq00 )

 1
0

0
Td0
 
∆Efd
+ 0
 0 
ωbase ∆Tm
0 2H

Substituting (69) in (70) and simplifying lead to

0
    1
∆E˙q − K31T 0 − TK04 ∆Eq0 0
 
0 0
Td0
 
 ∆δ̇  = 
d0 d0 ∆Efd
∆δ +

 0 0 1    0 0 
∆Tm
ωbase
∆ω̇ − ω2H
base
K2 − ω2H
base
K1 − ω2H
base
D ∆ω 0 2H
(71)
Siva (IIT P) EE549 43 / 97
where
1
K1 = − ×
Re2 + (Xe + Xq )(Xe + Xd0 )
[Iq0 V∞ (Xd0 − Xq )((Xe + Xq ) sin δ0 − Re cos δ0 )
+V∞ ((Xd0 − Xq )Id0 − Eq0 0
)((Xe + Xd0 ) cos δ0 + Re sin δ0 )]
1
K2 = 2 [Iq0 ((Re2 + (Xe + Xq )(Xe + Xd0 ))−
Re + (Xe + Xq )(Xe + Xd0 ) (72)
0 0 0
(Xd − Xq )(Xe + Xq )) − Re (Xd − Xq )Id0 + Re Eq0 ]
1 (Xd − Xd0 )(Xe + Xq )
=1+ 2
K3 Re + (Xe + Xq )(Xe + Xd0 )
V∞ (Xd − Xd0 )
K4 = 2 [(Xe + Xq ) sin δ0 − Re cos δ0 ]
Re + (Xe + Xq )(Xe + Xd0 )

Siva (IIT P) EE549 44 / 97

Example 12.3 from P. Kundur : This example analyzes the small signal
stability of the example 12.2 including filed circuit dynamics.

The parameters of each of the four generators of the plant are in per unit
on its rating are as follows:

Xd = 1.81 Xq = 1.76 Xd0 = 0.3 Xl = 0.16

0
Rs = 0.003 Tdo = 8 s H = 3.5 D=0

Siva (IIT P) EE549 45 / 97

If the plant output is

P = 0.9 Q = 0.3 Vt = 1.0∠36◦ V∞ = 0.995∠0◦

compute the following:

1 Eigenvalues of A and the corresponding eigenvectors and participation
matrix.
2 Synchronizing torque coefficient Ks and damping torque coefficient
Kd

Siva (IIT P) EE549 46 / 97

 0.003 1.76 It ∠γ 0.15 0.5
Vt = 1.0∠36◦

E 0 ∠δ 0.995∠0◦

S∗ 0.9 − 0.3
I¯t = ∗ = = 0.95∠17.56◦
V 1.0∠−36◦
Ē 0 = V̄t + I¯t × (Rs + Xq ) = 1.1253∠49.84◦
The angle of E 0 with respect to the infinite bus is

δ0 = 81.99◦

Siva (IIT P) EE549 47 / 97

 Vd0 = Vt sin(δ0 − α) = 0.7192
Vq0 = Vt cos(δ0 − α) = 0.6948
Id0 = It sin(δ0 − γ) = 0.8569
Iq0 = It cos(δ0 − γ) = 0.4101
0
Eq0 = Vq0 + Iq0 Rs + Id0 Xd0 = 0.9531
0
Efd0 = Eq0 + (Xd − Xd0 )Id = 2.2471
The system matrix A is
 
−0.3237 −0.1957 0
A= 0 0 1.0000
−56.0058 −40.2725 0

The eigenvalues are

λ1 = −0.0515; λ2 , λ3 = −0.1361 ± 6.34

Siva (IIT P) EE549 48 / 97

Small Signal Stability of a SMIB - Including Exciter
Dynamics

Let the input mechanical torque Tm be constant.

Therefore, turbine and speed governor dynamics are not needed.
Exciter dynamics, field flux variations along with rotor dynamics are
considered.

Siva (IIT P) EE549 49 / 97

A simplified fast acting static exciter is now considered which is
represented by a first order control block with a gain KA and time
constant TA as shown here.

Figure: Simplified Fast Acting Static Exciter

The differential equation representing the dynamics of the exciter is

dEfd
TA = −Efd + KA (Vref − Vt ) (73)
dt

Siva (IIT P) EE549 50 / 97

Linearizing (73) gives
d∆Efd
TA = −∆Efd + KA (∆Vref − ∆Vt ) (74)
dt
The terminal voltage in terms of dq0 parameters is
q
Vt = Vd2 + Vq2 (75)

Linearizing (75) gives

  
Vd0 Vq0 Vd0 Vq0 ∆Vd
∆Vt = ∆Vd + ∆Vq = (76)
Vt Vt Vt Vt ∆Vq

Siva (IIT P) EE549 51 / 97

Linearizing (73) gives
d∆Efd
TA = −∆Efd + KA (∆Vref − ∆Vt ) (74)
dt
The terminal voltage in terms of dq0 parameters is
q
Vt = Vd2 + Vq2 (75)

Linearizing (75) gives

  
Vd0 Vq0 Vd0 Vq0 ∆Vd
∆Vt = ∆Vd + ∆Vq = (76)
Vt Vt Vt Vt ∆Vq

Substituting (69) in (66) gives

   
∆Vd 1 0 Xq
= 2
∆Vq Re + (Xe + Xq )(Xe + Xd0 ) −Xd0 0
     (77)
V∞ (Xe + Xq ) sin δ0 − Re V∞ cos δ0 (Xe + Xq ) ∆δ 0
+
V∞ (Xe + Xd0 ) cos δ0 + Re V∞ sin δ0 Re ∆Eq0 ∆Eq0

Siva (IIT P) EE549 51 / 97

Substituting (77) in (76) and simplifying give

∆Vt = K5 ∆δ + K6 ∆Eq0 (78)

where
1 Vd0
K5 = ( Xq (Re V∞ sin δ0 + V∞ (Xe + Xd0 ) cos δ0 )
Re2 + (Xe + Xq )(Xe + Xd0 ) Vt
Vq0 0
+ X (Re V∞ cos δ0 − V∞ (Xe + Xq ) sin δ0 ))
Vt d
1 Vd0 Vq0 0 Vq0
K6 = 0 ( Xq Re − Xd (Xe + Xq )) +
Re2 + (Xe + Xq )(Xe + Xd ) Vt Vt Vt

Siva (IIT P) EE549 52 / 97

Equation (74) can be substituted in (71), with the linearized terminal
voltage defined in (78), and can be written as
0
− K31T 0 − TK04 1
   
∆E˙q 0 ∆Eq0

T 0
d0 d0 d0
 ∆δ̇   0 0 1 0  ∆δ 
   
 =  ωbase ωbase ωbase  
 ∆ω̇  − 2H K2 − 2H K1 − 2H D 0   ∆ω 

∆Ėfd − KTA AK6 − KTA AK5 0 − T1A ∆Efd
  (79)
0 0
 
0 0  ∆Vref
+
0 ω2H
 
base 
∆Tm
KA
TA 0

Siva (IIT P) EE549 53 / 97

Equation (74) can be substituted in (71), with the linearized terminal
voltage defined in (78), and can be written as
0
− K31T 0 − TK04 1
   
∆E˙q 0 ∆Eq0

T 0
d0 d0 d0
 ∆δ̇   0 0 1 0  ∆δ 
   
 =  ωbase ωbase ωbase  
 ∆ω̇  − 2H K2 − 2H K1 − 2H D 0   ∆ω 

∆Ėfd − KTA AK6 − KTA AK5 0 − T1A ∆Efd
  (79)
0 0
 
0 0  ∆Vref
+
0 ω2H
 
base 
∆Tm
KA
TA 0

The constants K1 and K6 are called as Heffron-Phillips constants and

the state space model is called Heffron-Phillips model.
It can be observed from (79) that the nonlinear SMIB system has
been linearized and now represented in state space with the state
variables being ∆Eq0 ∆δ ∆ω ∆Efd .


The stability of the system now depends on the eigen values of the
state matrix.
Siva (IIT P) EE549 53 / 97
The state space representation of the system given in (79) can be
represented as a control block diagram as shown here.

Figure: Heffron-Phillips model

Siva (IIT P) EE549 54 / 97

Taking the Laplace transform of (79) gives

1 K4 1
s∆Eq0 (s) = − 0 ∆Eq0 (s) − 0 ∆δ(s) + 0 ∆Efd (s)
K3 Td0 Td0 Td0
s∆δ(s) = ∆ω(s)
ωbase ωbase ωbase
s∆ω(s) = − K2 ∆Eq0 (s) − K1 ∆δ(s) − D∆ω(s)
2H 2H 2H
ωbase
+ ∆Tm (s)
2H
KA K6 KA K5 1 KA
s∆Efd (s) = − ∆Eq0 (s) − ∆δ(s) − ∆Efd (s) + ∆Vref (s)
TA TA TA TA
(80)

Siva (IIT P) EE549 55 / 97

Simplifying (80) gives

K4 K3 K3
∆Eq0 (s) = − 0 ∆δ(s) + 0 ∆Efd (s) (81)
1 + sK3 Td0 1 + sK3 Td0
KA KA
∆Efd (s) = − (K5 ∆δ(s) + K6 ∆Eq0 (s)) + ∆Vref (s) (82)
1 + sTA 1 + sTA
∆ω(s) = s∆δ(s) (83)
ωbase ωbase ωbase
s 2 ∆δ(s) = ∆Tm (s) − (K1 ∆δ(s) + K2 ∆Eq0 (s)) − Ds∆δ(s)
2H 2H 2H
(84)

Siva (IIT P) EE549 56 / 97

Simplifying (80) gives

K4 K3 K3
∆Eq0 (s) = − 0 ∆δ(s) + 0 ∆Efd (s) (81)
1 + sK3 Td0 1 + sK3 Td0
KA KA
∆Efd (s) = − (K5 ∆δ(s) + K6 ∆Eq0 (s)) + ∆Vref (s) (82)
1 + sTA 1 + sTA
∆ω(s) = s∆δ(s) (83)
ωbase ωbase ωbase
s 2 ∆δ(s) = ∆Tm (s) − (K1 ∆δ(s) + K2 ∆Eq0 (s)) − Ds∆δ(s)
2H 2H 2H
(84)

Assuming ∆Vref (s) = 0 and substituting (82) in (81), we get

(K4 (1 + sTA ) + KA K5 )K3

∆Eq0 (s) = − 0 )(1 + sT ) + K K K ∆δ(s) (85)
(1 + sK3 Td0 A A 6 3

Siva (IIT P) EE549 56 / 97

The linearized electrical torque ∆Te expression is

∆Te = ∆Eq0 Iq + Eq0 ∆Iq + (Xq − Xd0 )Iq ∆Id + (Xq − Xd0 )Id ∆Iq
(86)
 
 ∆Id
= Iq ∆Eq0 + (Xq − Xd0 )Iq Eq0 + (Xq − Xd0 )Id

∆Iq

On substituting (69) in (86), we get

∆Te = K1 ∆δ + K2 ∆Eq0 (87)

By taking the Laplace transform of (87) and substituting (85) in it,

∆Te (s) = H(s)∆δ(s) (88)

where
(K4 (1 + sTA ) + KA K5 )K3 K2
H(s) = K1 − 0 )(1 + sT ) + K K K
(1 + sK3 Td0 A A 6 3

Siva (IIT P) EE549 57 / 97

On substituting (88) in (84) and assuming ∆Tm (s) = 0, we get
 ωbase ωbase 
s2 + Ds + H(s) ∆δ(s) = 0 (89)
2H 2H

The small signal oscillations are typically in the range of 0.1 Hz to 3

Hz.
In this region of frequency interest, the transfer function H(s) can be
split into two components.
(K4 (1 + ωTA ) + KA K5 )K3 K2
H(s = ω) = K1 − 0 )(1 + ωT ) + K K K
(1 + ωK3 Td0 A A 6 3
(K4 (1 + ωTA ) + KA K5 )K3 K2
= K1 − 0 ) + ω(K T 0 + T )
(1 + KA K6 K3 − ω 2 K3 TA Td0 3 d0 A
K2 ((K4 + KA K5 ) + ωK4 TA )
= K1 − 1 0 ) + ω(T 0 + TA )
( K3 + KA K6 − ω 2 TA Td0 d0 K3
(90)

Siva (IIT P) EE549 58 / 97

Neglecting the effect of constant K4
K2 KA K5
H(ω) = K1 − TA
( K13 + KA K6 − 2 0 )
ω TA Td0 0 +
+ ω(Td0 K3 )
 
K2 KA K5 1 0
+ KA K6 − ω 2 TA Td0
K3
= K1 −  2  2
1 0
+ KA K6 − ω 2 TA Td0 + ω(Td0 0 + TA )
K3 K3
| {z } (91)
Ks
 
K2 KA K5 Td0 0 + TA
K3
+ ω  2  2
1 0 0 + TA
K3 + KA K6 − ω 2 TA Td0 + ω(Td0 K3 )
| {z }
Kd

H(s) = Ks + sKd (92)

Siva (IIT P) EE549 59 / 97

The electrical torque is

∆Te (ω) = Ks ∆δ + ωKd ∆δ

(93)
= Ks ∆δ + Kd ∆ω

It has two components.

The synchronising torque Ks is in phase with the change in rotor
angle ∆δ.
The damping torque Kd is in quadrature to the rotor angle or in
phase with the change in speed ∆ω.
Substituting (79) in (76) gives
 ωbase ωbase 
s2 + (Kd + D)s + Ks ∆δ(s) = 0 (94)
2H 2H

Siva (IIT P) EE549 60 / 97

Equation (94) is a damped second order homogeneous system.
If the damping coefficient (D+Kd ) were zero, the system would become
undamped with natural frequency of oscillations being
r
ωbase
ωn = ± Ks (95)
2H
In the undamped system, if Ks is zero for a forced system that is ∆Tm 6=
0, the system will become aperiodically unstable that is the rotor angle
keeps on increasing and ultimately leading to loss of synchronism.
In case, (Kd + D) becomes negative due to system conditions, there
will be oscillatory instability that is the peak of the oscillations keep on
increasing leading to loss of synchronism.
If both Ks and Kd are positive for a given system, the system will settle
down to a steady state operating point with oscillations being damped
out after a disturbance.

Siva (IIT P) EE549 61 / 97

  
K2 KA K5 1 0
+ KA K6 − ω 2 TA Td0
K3
Ks = K1 −  2  2
1
+ K K − ω 2T T 0 + ω(T 0 + TA )
K3 A 6 A d0 d0 K3
  (96)
K2 KA K5 Td00 + TA
K3
Kd =  2  2
1
+ K K − ω 2T T 0 + ω(T 0 + TA )
K3 A 6 A d0 d0 K3

From the above equation, the following observations are made.

If K5 is negative, Ks will be positive and Kd will be negative. Hence,
negative damping and oscillatory instability.
This effect of negative damping torque due to negative K5 gets
pronounced if the exciter has a very high gain KA .
In a practical system, K5 can become negative in a heavily loaded
case and a high exciter gain KA can lead to stability issues.
The constant K5 becomes important while assessing the stability.
Siva (IIT P) EE549 62 / 97
Example 12.4 from P. Kundur : This example analyzes the small signal
stability of the example 12.3 including exciter dynamics.

The parameters of each of the four generators of the plant are in per unit
on its rating are as follows:

Xd = 1.81 Xq = 1.76 Xd0 = 0.3 Xl = 0.16

0
Rs = 0.003 Tdo = 8 s H = 3.5 D=0

Siva (IIT P) EE549 63 / 97

The exciter parameters are

KA = 200 TA = 0.02

If the plant output is

P = 0.9 Q = 0.3 Vt = 1.0∠36◦ V∞ = 0.995∠0◦

compute the following:

1 Eigenvalues of A.

Siva (IIT P) EE549 64 / 97

The constants K 1 to K 6 are

K 1 = 0.7478; K 2 = 1.0399; K 3 = 0.3862

K 4 = 1.5660; K 5 = −0.1432; K 6 = 0.4756

The system matrix A is
 
−0.3 −0.2 0 0.1
 0 0 1 0 
A=
 −56

−40.3 0 0 
−4756 1432.4 0 −50

The eigenvalues are

λ1 = −28.4666 λ2 = −22.8857 λ3 , λ4 = 0.5143 ± 7.2137

For KA = 10, the eigenvalues are

λ1 = −49.3901 λ2 = −0.9110; λ3 , λ4 = −0.0113 ± 6.3287

Siva (IIT P) EE549 65 / 97

Power System Stabilizer (PSS)

Power system stabilizer is a device used for overcoming the negative

damping effect of the high gain exciter.
It acts as a supplementary controller to the excitation system.
Inputs to the power system stabilizer can be change in frequency, speed,
power or a combination of these.
The output is a voltage signal introduced in the excitation system to
control the output of the exciter.
The basic idea of the power system stabilizer is to introduce a pure
damping term in (94) so as to counter the negative damping effect of
the exciter.

Siva (IIT P) EE549 66 / 97

 Let PSS have a transfer function G (s) and the change in the speed
be the input to the PSS.
The output of the PSS be ∆VPSS .
The output of PSS is added as a supplementary signal to the exciter
reference.
Therefore, the equation (74) becomes

d∆Efd
TA = −∆Efd + KA (∆Vref + ∆VPSS − ∆Vt ) (97)
dt
The transfer function between ∆VPSS and ∆Te is found to see the effect
of ∆VPSS on Ks and Kd .

Siva (IIT P) EE549 67 / 97

To get the transfer function, let us assume that ∆δ = 0 and ∆Vref = 0.
From (81) and (82),

K3
∆Eq0 (s) = 0 ∆Efd (s) (98)
1 + sK3 Td0
KA K6 KA
∆Efd (s) = − ∆Eq0 (s) + ∆VPSS (s) (99)
1 + sTA 1 + sTA

Substituting (99) in (98) gives

K3 KA
∆Eq0 (s) = 0 )(1 ∆VPSS (s) (100)
(1 + sK3 Td0 + sTA ) + K3 KA K6

Substituting (100) in the Laplace transform of (100) gives

K2 K3 KA
∆Te (s) = K2 ∆Eq0 (s) = 0 ∆VPSS (s)
(1 + sK3 Td0 )(1 + sTA ) + K3 KA K6
(101)

Siva (IIT P) EE549 68 / 97

Equation (101) can be approximated for the usual range of constants as
follows
K2 KA
∆Te (s) = 0
sTd0
∆VPSS (s) (102)
(1 + KA K6 )(1 + sTA )( K13 + KA K6 )

1
For a high gain exciter, K3 << KA K6 .

K2
K6
∆Te (s) = 0
sTd0
∆VPSS (s) (103)
(1 + KA K6 )(1 + sTA )

Since PSS has a transfer function G (s) with input ∆ω, Equation (103)
can be written as
K2
K6
∆Te (s) = 0
sTd0
G (s)∆ω (104)
(1 + KA K6 )(1 + sTA )

Siva (IIT P) EE549 69 / 97

If PSS is to contribute pure damping throughout the frequency of interest,
the transfer function between ∆Te and ∆ω should be
0
sTd0
G (s) = KPSS (1 + )(1 + sTA )
KA K6
where KPSS is the gain of PSS.
K2 K2
∆Te = KPSS ∆ω = KPSS s∆δ (105)
K6 K6
With the help of (105), the equation (94) can be approximated as
 
2 ωbase K2 ωbase
s + (Kd + D + KPSS )s + Ks ) ∆δ(s) = 0 (106)
2H K6 2H

Hence, PSS improves the damping of system by providing a positive

damping torque.

Siva (IIT P) EE549 70 / 97

 Since G (s) is a pure lead function, it is not physically realizable.
Hence, we have a compromise resulting in what is called a lead-lag
type transfer function such that it provides enough phase lead over
the expected range of frequencies.
1 + sT1 n
 
sTw
G (s) = KPSS (107)
1 + sTw 1 + sT2
where

KPSS is the gain.

Tw is the washout filter time constant.
T1 , T2 are the lead-lag network time constants.
n is the number of lead-lag network blocks.

Siva (IIT P) EE549 71 / 97

 Figure: Block diagram of PSS

The procedure for finding the parameters KPSS , T1 and T2 is as follows:

Let
K2 K3 KA
GEX (s) = 0 (108)
(1 + sK3 Td0 )(1 + sTA ) + K3 KA K6

Siva (IIT P) EE549 72 / 97

1 Find the natural frequency of oscillations of the system without damp-
ing. The flux linkage can be considered constant so that ∆Eq0 is zero.

∆Te = K1 ∆δ

Also the damping coefficient D is assumed to be zero. With these

assumptions, the equation (94) will become
 ωbase 
s2 + K1 ) ∆δ(s) = 0 (109)
2H
r
ωbase
ωn = ± K1 (110)
2H
2 At the natural frequency, find the time constant of lead-lag network
such that the lag due to GEX (s) is compensated.

∠G (ωn ) + ∠GEX (ωn ) = 0 (111)

Siva (IIT P) EE549 73 / 97

3 For a high value of washout time constant, the washout filter will not
contribute any angle to the transfer function of PSS hence for n = 1.

∠(1 + ωn T1 ) − ∠(1 + ωn T2 ) + ∠GEX (ωn ) = 0 (112)

One of the lead-lag time constants can be arbitrarily chosen. The other
time constant can be derived from (112).
4 To compute KPSS , we can compute KPSS∗ i.e., the gain at which the
system becomes unstable using the root locus, and then have KPSS =
1 ∗
3 KPSS .
5 The PSS should be activated only when low-frequency oscillations de-
velop and should be automatically terminated when the system oscil-
lation ceases. It should not interfere with the regular function of the
excitation system during steady state operation of the system frequency.
The washout stage has the transfer function

sTW
Gw (s) = (113)
1 + sTW
Siva (IIT P) EE549 74 / 97
 Washout filter acts like a high-pass filter allowing those signals which
are above the cut-off frequency. The washout filter time constant is
chosen so that the signal with lowest frequency of interest can be passed
through. For washout filter time constant of Tw = 10 s, signals of
frequency 0.1 Hz and above can be passed. The stead state operation
is considered as DC signals and hence blocked by the washout filter.
Let the output of the washout filter be ∆VWF .
Let the output of the lead-lag block be ∆VPSS .
The linearized dynamic equations of the PSS are
1
∆V̇WF = − ∆VWF + KPSS ∆ω̇ (114)
TW
1 1 T1
∆V̇PSS = − ∆VPSS + ∆VWF + ∆V̇WF (115)
T2 T2 T2

Siva (IIT P) EE549 75 / 97

By substituting ∆ω̇ from (79) in (114) and by substituting the resultant
equation in (115), we get
∆Eq0
 
 ∆δ 
h ω ωbase ωbase i 
base K K
∆V̇WF = − 2H − 2H K1 KPSS − 2H DKPSS 0 − T1 0  ∆ω 
2 PSS
 ∆Efd 
 
W

∆VWF
 (116)
∆VPSS
i 
ωbase ∆Vref
h
+ 0 2H
KPSS
∆Tm

∆Eq0
 
 ∆δ 
h T ω T ωbase T ωbase T
i 
− T1  ∆ω 
 
1 − T 1 T1
base K K
∆V̇PSS = − T1 2H − T1 K1 KPSS − T1 DKPSS 0
 
2 PSS 2H 2H T2
2 2 2 W 2 2  ∆Efd 
∆VWF
 
∆VPSS
i 
ωbase T1 ∆Vref
h
+ 0 2H T2
KPSS
∆Tm
(117)

Siva (IIT P) EE549 76 / 97

The state space model including the PSS dynamics is
K
− 10 − 04 1
 
0 0 0
0  K3 T T T0
∆E˙q

 d0 d0 d0 
 ∆δ̇  
 0 0 1 0 0 0 
ωbase ωbase ωbase 

 ∆ω̇  
  − 2H
K2 − 2H K1 − 2H D 0 0 0 

 ∆Ė  =  KA K6 A 5 K K
− T1
KA 
  
fd  − T − T 0 0 TA 

  A ωbase A A
 − ωbase K2 KPSS ωbase
 ∆V̇
− T1
  
WF
2H
− 2H K1 KPSS − 2H DKPSS 0 0 
∆V̇PSS  W 
T1 ωbase T ωbase T ωbase T
 
−T K2 KPSS − T1 K1 KPSS − T1 DKPSS 0 1 − T 1 T1 − T1
2 2H 2 2H 2 2H T2 W 2 2
0 0
 
∆Eq0
 
 0 0 
 ∆δ   ωbase 
 0 2H
 

 ∆ω  
  ∆Vref
 +  KA 0 
 ∆Tm
 ∆Efd  

T A
  ωbase 
∆VWF  0 KPSS 

 2H 
∆VPSS ωbase T1
0 2H T
KPSS
2
(118)

Siva (IIT P) EE549 77 / 97

Example 12.4 from P. Kundur : This example analyzes the small signal
stability of the example 12.3 including exciter dynamics with PSS.

The parameters of each of the four generators of the plant are in per unit
on its rating are as follows:

Xd = 1.81 Xq = 1.76 Xd0 = 0.3 Xl = 0.16

0
Rs = 0.003 Tdo = 8 s H = 3.5 D=0

Siva (IIT P) EE549 78 / 97

The exciter parameters are

KA = 200 TA = 0.02

The PSS parameters are

KPSS = 9.5; TW = 1.4; T1 = 0.154; T2 = 0.033

If the plant output is

P = 0.9 Q = 0.3 Vt = 1.0∠36◦ V∞ = 0.995∠0◦

compute the following:

1 Eigenvalues of A.

Siva (IIT P) EE549 79 / 97

For KA = 200, the eigenvalues are

λ1 = −46.4791

λ2 , λ3 = −15.8405 ± 14.6135

λ4 , λ5 = −1.22 ± +6.6794

λ6 = −0.741

Siva (IIT P) EE549 80 / 97

Small Signal Stability Analysis of Multi-Machine System

Let each generator be modelled by a sub-transient model along with a

power system stabilizer and stator algebraic equations.
The network is represented by real and reactive power balance equa-
tions.

Siva (IIT P) EE549 81 / 97

Differential equations (for i = 1, 2, · · · , ng generator):

0
dEqi0 (Xdi0 − Xdi00 )
Td0i = −Eqi0 − (Xdi − Xdi0 )[Idi − (−Eqi0
dt (Xdi0 − Xlsi )2 (119)
+ (Xdi0 − Xlsi )Idi + ψ1di ] + Efdi

00 dψ1di
Td0i = Eqi0 − (Xdi0 − Xlsi )Idi − ψ1di (120)
dt
0 dEdi0
Tq0i = −Edi0 + (Xqi − Xqi0 ){(Iqi
dt
(Xqi0 − Xqi00 ) 0 (121)
− 0 (E + (Xqi0 − Xlsi )Iqi + ψ2qi )}
(Xqi − Xlsi )2 di

00 dψ2qi
Tq0i = −Edi0 − (Xqi0 − Xlsi )Iqi − ψ2qi (122)
dt

Siva (IIT P) EE549 82 / 97

 dδi
= ωi − ωs (123)
dt
2Hi dωi
= Tmi − Tei − Di (ωi − ωbase ) (124)
ωbase dt

dEfdi
TEi = −(KEi + SEi (Efdi ))Efdi + VRi (125)
dt
dVR KA KF
TA = −VR + KA RF − Efd + KAi (Vrefi − Vti ) (126)
dt TF
dRFi KFi
TFi = −RFi + Efdi (127)
dt TFi

1 ωbase
V̇WF = − VWFi + KPSSi (Tmi − Tei − Di (ωi − ωbase )) (128)
TWi 2Hi

Siva (IIT P) EE549 83 / 97

  
1 1 1 T1i
V̇PSS = − VPSSi + − ∆VWFi
T2 T2i TWi T2i
(129)
T1i ωbase
+ KPSSi (Tmi − Tei − Di (ωi − ωbase ))
T2i 2Hi

 
dTMi KHPi TRHi KHPi TRHi
TRHi = −∆TMi + 1 − PCHi + PSVi (130)
dt TCHi TCHi
dPCHi
TCHi = −PCHi + PSVi (131)
dt

 
dPSVi 1 ωi max
TSVi =− − 1 + Prefi − PSVi , 0 ≤ PSVi ≤ PSVi
dt RDi ωbase
(132)

Siva (IIT P) EE549 84 / 97

Stator algebraic equations :

(Xqi00 − Xlsi ) (Xqi0 − Xqi00 )

Vi sin(δi − θi ) + Rsi Idi − Xqi00 Iqi − Edi0 + ψ2qi = 0
(Xqi0 − Xlsi ) (Xqi0 − Xlsi )
(133)
(Xdi00 − Xlsi ) 0 (Xdi0 − Xdi00 )
Vi cos(δi − θi ) + Rsi Iqi + Xdi00 Idi − E − ψ1di = 0
(Xdi0 − Xlsi ) qi (Xdi0 − Xlsi )
(134)

Network Equations :
For i = 1, 2, · · · , ng generator buses,
n
X
Vi sin(δi − θi )Idi + Vi cos(δi − θi )I qi − PDi (Vi ) − Vi Vj Yij cos(αij + θj − θi ) = 0
j=1
(135)
n
X
Vi cos(δi − θi )Idi − Vi sin(δi − θi )I qi − QDi (Vi ) + Vi Vj Yij sin(αij + θj − θi ) = 0
j=1
(136)
Siva (IIT P) EE549 85 / 97
For i = ng + 1, ng + 2, · · · , n load buses,
n
X
PDi (Vi ) + Vi Vj Yij cos(αij + θj − θi ) = 0 (137)
j=1
X n
QDi (Vi ) − Vi Vj Yij sin(αij + θj − θi ) = 0 (138)
j=1

The differential equations (119) to (132) can be represented as

Ẋi = fi (Xi , Idi , Iqi , θg , Vg , U) (139)

where

Xi =[Eqi0 ψ1di Edi0 ψ2qi δi ωi VRi Efdi RFi TMi PSVi PCHi VWFi VPSSi ]T
i = 1, 2, · · · , ng

 T
Ui = Vrefi Prefi i = 1, 2, · · · , ng

Siva (IIT P) EE549 86 / 97

 θg = [θ1 , · · · , θng ]T , Vg = [V1 , · · · , Vng ]T

θl = [θng +1 , · · · , θn ]T , Vl = [Vng +1 , · · · , Vn ]T
The stator equations (133) and (134) can be written as

gi (Xi , Idi , Iqi , θg , Vg ) = 0 i = 1, · · · , ng (140)

The network equations at the generator buses (135) and (136) can be
written as
hi (Xi , Idi , Iqi , θg , Vg , θl , Vl ) = 0 i = 1, · · · , ng (141)
The network equations at the load buses (137) and (138) can be written as

ki (θg , Vg , θl , Vl ) = 0 i = ng + 1, · · · , n (142)

Siva (IIT P) EE549 87 / 97

 The functions f , g , h, k are non linear.
Let the steady state operating point be [Xi0 , Idi0 , Iqi0 , Ui0 ].
The network voltages and angles at the steady state are
[θg 0 , Vg 0 , θl0 , Vl0 ]
Linearizing (139) around the initial operating conditions gives
  
    ∆θg  
∂fi h
∂fi ∂fi
i ∆I
di  ∂fi ∂fi
 
∆Vg  + ∂fi ∆Ui

∆Ẋi = ∆Xi + ∂I ∂Iqi + 0 0
∂Xi | {z } | ∆I  ∂θg ∂Vg   ∆θl  ∂Ui
di
 
qi
| {z }
B1i
{z } | {z } ∆Vl | {z }
A1i ∆Idqi B2i Wi
(143)

∆Ẋi =A1i(14×14) ∆Xi(14×1) + B1i(14×2) ∆Idqi (2×1)

 
∆θg
∆Vg  (144)
+ [B2i(14×2ng ) 0(14×2(n−ng )) ]  
 ∆θl  + Wi(14×2) ∆Ui(2×1)
∆Vl (2n×1)

Siva (IIT P) EE549 88 / 97

For ng generators,

∆Ẋ =A1i (14ng ×14ng ) ∆X(14ng ×1) + B1 (14ng ×2) ∆Idq (2ng ×1)
 
∆θg
∆Vg 
+ [B2 (14ng ×2ng ) 0(14ng ×2(n−ng )) ]  
 ∆θl  (145)
∆Vl (2n×1)
+ Wi (14ng ×2ng ) ∆Ui (2ng ×1)

Siva (IIT P) EE549 89 / 97

Similarly, by linearizing (140), we get
  
    ∆θg
∂gi h
∂g ∂g
i ∆I
di
 ∂g ∂g
 i i  ∆Vg 

∆Xi + ∂Idii ∂Iqii + 0 0  =0 (146)
∂Xi | {z } ∆Iqi  ∂θg ∂Vg   ∆θl 
| {z }
C2i
| {z } | {z } ∆Vl
C1i ∆Idqi D1i

 
∆θg
∆Vg 
C1i(2×14) ∆Xi(14×1) +C2i(2×2) ∆Idqi (2×1) +[D1i(2×2ng ) 0(2×2(n−ng )) ]  
 ∆θl  =0
∆Vl (2n×1)
(147)
For ng generators,

C1(2ng ×14ng ) ∆X(14ng ×1) + C2(2ng ×2ng ) ∆Idq (2ng ×1)

 
∆θg
∆Vg  (148)
+[D1(2ng ×2ng ) 0(2ng ×2(n−ng )) ]  
 ∆θl  =0
∆Vl (2n×1)

Siva (IIT P) EE549 90 / 97

Similarly, by linearizing (141), we get
  
  ∆θg
∂hi ∆Idi i 
 ∂h ∂h ∂h ∂h  
i  ∆Vg 
h
∂hi ∂hi  i i i
∆Xi + ∂I ∂I + =0 (149)
∂Xi ∆Iqi  ∂θg ∂Vg ∂θl ∂Vl   ∆θl 
di qi
 
| {z }
| {z }
C4i
| {z } | {z } | {z }
D
∆Vl
C3i ∆Idqi D2i 3i

 
∆θg
∆Vg 
C3i(2×14) ∆Xi(14×1) +C4i(2×2) ∆Idqi (2×1) +[D2i(2×2ng ) D3i(2×2(n−ng )) ]  
 ∆θl  =0
∆Vl (2n×1)
(150)
For ng generators,

C3(2ng ×14ng ) ∆X(14ng ×1) + C4(2ng ×2ng ) ∆Idq (2ng ×1)

 
∆θg
∆Vg  (151)
+[D2(2ng ×2ng ) D3(2ng ×2(n−ng )) ]  
 ∆θl  =0
∆Vl (2n×1)

Siva (IIT P) EE549 91 / 97

Similarly, by linearizing (142), we get
  
∆θg
 ∂k ∂k ∂k ∂k  
 i i i i  ∆Vg 
=0 (152)
 ∂θg ∂Vg ∂θl ∂Vl   ∆θl 
 
| {z } | {z }
D D5i ∆Vl
4i

 
∆θg
∆Vg 
[D4i (2×2ng ) D5i (2×2(n−ng )) ]  
 ∆θl  =0 (153)
∆Vl (2n×1)
For ng generators,
 
∆θg
∆Vg 
[D4 (2ng ×2ng ) D5 (2ng ×2(n−ng )) ]  
 ∆θl  =0 (154)
∆Vl (2n×1)

Siva (IIT P) EE549 92 / 97

Equations (145), (146), (151) and (154) can be written as
 
∆θg
∆Vg 
∆Ẋ = A1 ∆X + B1 ∆Idq + [B2 0] 
 ∆θl  + W∆U
 (155)
∆Vl
 
∆θg
∆Vg 
0 = C1 ∆X + C2 ∆Idq + [D1 0]  
 ∆θl  (156)
∆Vl
 
∆θg
∆Vg 
0 = C3 ∆X + C4 ∆Idq + [D2 D3 ]  
 ∆θl  (157)
∆Vl
 
∆θg
∆Vg 
0 = [D4 D5 ]  
 ∆θl  (158)
∆Vl

Siva (IIT P) EE549 93 / 97

From (158), the load bus voltages and angles can be determined as follows:
   
∆θl −1 ∆θg
= −D5 D4 (159)
∆Vl ∆Vg

Let
D6 = D2 − D3 D5 −1 D4 (160)
Substituting (160) and (159) in (157) and (156), and finding the generator
bus voltages, angles, direct axis current and quadrature axis current give
 
∆Idq  −1  
 ∆θg  = − C2 D1 C1  
∆X (161)
C4 D 6 C3
∆Vg

Siva (IIT P) EE549 94 / 97

By substituting (161) in (155),
 
  ∆Idq
∆Ẋ = A1 ∆X + B1 B2  ∆θg  + W∆U
∆Vg (162)

= Asys ∆X + W∆U

where
 C2 D1 −1 C1
   

Asys = A1 − B1 B2
C4 D6 C3

The small signal stability of the multi-machine system can now be

known by finding the eigenvalues of the state transition matrix Asys .
Since, the rotor angle of any synchronous machine should be defined
with respect to a reference, one state variable corresponding to the
rotor angle of a generator which is taken as reference becomes re-
dundant hence there will be one zero or a slightly positive eigenvalue
corresponding to the redundancy of the reference angle.
Siva (IIT P) EE549 95 / 97
Example 12.6 from P. Kundur :

Figure: Two Area System

Siva (IIT P) EE549 96 / 97

1 Perform Load flow.
2 Determine the machines’ steady state values (δ, Vd , Vq , Id , Iq , Eq0 ,
Ed0 , Efd , ψ1d and ψ2q ).
3 Calculate the system matrix as given in (162).
The size depends on the models.
For example, the sub-transient model of a synchronous machine
without turbine, speed governor dynamics and PSS has seven states.

∆Xi = [∆δi ∆ωi ∆Edi0 ∆Eqi0 ∆ψ1di ∆ψ2qi ∆Efdi ]

Hence, the number of states is 28. The size of A is 28 × 28.

If PSS is considered, the number of states will increase. Suppose there
is one lead-lag block, 2 additional states for each machine, hence the
size of A will be 36 × 36
If there are two lead-lag blocks in PSS, 3 additional states for each
machine,hence the size of A will be 40 × 40
4 Determine the eigenvalues of A.

Siva (IIT P) EE549 97 / 97