ECEN 667

Power System Stability

Lecture 23: Small Signal Stability,
Oscillations

Prof. Tom Overbye

Dept. of Electrical and Computer Engineering
Texas A&M University
[email protected]
 Announcements
• Read Chapter 8
• Homework 5 is due Today
• Homework 6 is due on Tuesday December 3
• Final is at scheduled time here (December 9
from1pm to 3pm)

2
 Example: Bus 4 SMIB Dialog
• On the SMIB dialog, the General Information tab
shows information about the two bus equivalent

PowerWorld case B4_SMIB 3

 Example: Bus 4 SMIB Dialog

• On the SMIB dialog, the A Matrix tab shows the Asys

matrix for the SMIB generator

• In this example A21 is showing

4 , pu 1  PE ,4   1  1  
        1.2812 cos  23.94   
 4 2H 4   4   6    0.3  0.22  
 0.3753
4
 Example: Bus 4 with
GENROU Model
• The eigenvalues can be calculated for any set of
generator models
• The example can be extended by replacing the bus 4
generator classical machine with a GENROU model
– There are now six eigenvalues, with the dominate response
coming from the electro-mechanical mode with a frequency
of 1.84 Hz, and damping of 6.9%

PowerWorld case B4_SMIB_GENROU 5

 Relation to the Signal-Based
Approaches
• The results from the SMIB analysis can be compared
with the signal-based approach by applying a short self-
clearing fault to disturb the system
• This can be done easily in PowerWorld by running the
transient stability simulation, looking at the Results
from RAM page, right-clicking on the desired signal(s)
and selecting Modal Analysis Selected Column.
• The next slide shows the results for the Bus 4 Generator
rotor angle (which would not be directly observable)

6
 Bus 4 Generator Rotor Angle
• Notice that the main mode, at 1.84 Hz with 7%
damping closely matches

Not all modes will be easily observed in all signals

7
 Example: Bus 4 with
GENROU Model and Exciter
• Adding an relatively slow EXST1 exciter adds
additional states (with KA=200, TA=0.2)
– As the initial reactive power output of the generator is
decreased, the system becomes unstable

Case is saved as B4_GENROU_EXST1

8
 Example: Bus 4 with
GENROU Model and Exciter
• With Q4 = 25 Mvar the eigenvalues are

• And with Q4=0 Mvar the eigenvalues are

9
 Example: Bus 4 with
GENROU Model and Exciter
• Graph shows response following a short fault when Q4
is 0 Mvar
88

86

84

82

80

78

76

74

72

70

68

66

64

62

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Ro to r An g le_Gen Bus 4 #1
b
c
d
e
f
g

• This motivates trying to get additional insight into how

to increase system damping, which is the goal of
modal analysis
10
 Modal Analysis - Comments
• Modal analysis (analysis of small signal stability
through eigenvalue analysis) is at the core of SSA
software Goal is to
• In Modal Analysis one looks at: determine
– Eigenvalues how the various
– Eigenvectors (left or right) parameters
– Participation factors affect the
response of
– Mode shape
the system

• Power System Stabilizer (PSS) design in a multi-

machine context can be done using modal analysis
method (we’ll see another method later)
11
 Eigenvalues, Right Eigenvectors
• For an n by n matrix A the eigenvalues of A are the
roots of the characteristic equation:
det[ A   I]  A   I  0
• Assume l1…ln as distinct (no repeated eigenvalues).
• For each eigenvalue li there exists an eigenvector
such that:
Av i  i v i

• vi is called a right eigenvector

• If li is complex, then vi has complex entries
12
 Left Eigenvectors

• For each eigenvalue li there exists a left

eigenvector wi such that:

w ti A  w ti i
• Equivalently, the left eigenvector is the right
eigenvector of AT; that is,
At w i  i w i

13
 Eigenvector Properties
• The right and left eigenvectors are orthogonal i.e.
w v i  0 , w v j  0 (i  j )
t
i
t
i

• We can normalize the eigenvectors so that:

w v i  1 , w v j  0 (i  j )
t
i
t
i

14
 Eigenvector Example

1 4  1  4
A  , A  I  0  0
3 2  3 2
3 (3) 2
 4(10) 3 49
  3  10  0  1,2 
2
  5,  2
2 2

1  5
Right Eigenvectors
 v11  v11  4v21  5v11
Av1  5 v1  v1     choose v 21  1  v11  1
v21  3v11  2v21  5v21
1
Similarly , v1   
1
4
2  2  v 2   
 3
15
 Eigenvector Example

• Left eigenvectors
1 4 
1  5 w A  w 5  [ w11 w21 ] 
t
1
t
1   5[ w11 w21 ]
3 2 
w11  3w21  5w11
 Let w21  4, then w11  3
4 w11  2 w21  5w21
3 1
w1    2  2  w 2   
4  1
1 4 3 1
v1    v 2    w1    w 2   
1  3 4  1
Verify w1t v1  7 , w t2 v 2  7 , w t2 v1  0 , w1t v 2  0
We would like to make w v  1.
t
i i

This can be done in many ways. 16

 Eigenvector Example

1 3 1 
Let W   
7  4 1
Then WT V  I
1 3 4  1 4  1 0 
Verify      
7 1 1 1 3  0 1 

• It can be verified that WT=V-1 .

• The left and right eigenvectors are used in
computing the participation factor matrix.

17
 Modal Matrices
• The deviation away from an equilibrium point can be
defined as
Δx  AΔx
• If the initial deviation corresponds to a right
eigenvector, then the subsequent response is along this
eigenvector since Av   v
i i i

18
 Modal Matrices
• From this equation ( Δx  AΔx ) it is difficult to
determine how parameters in A affect a particular x
because of the variable coupling
• To decouple the problem first define the matrices
of the right and left eigenvectors (the modal
matrices)
V  [ v1 , v 2 ..... v n ] & W  [ w1 , w 2 ,..... w n ]
AV  VΛ when Λ  Diag (i )

19
 Modal Matrices
• It follows that
V 1AV  Λ
• To decouple the variables define z so
x  Vz  x  Vz  AΔx  AVz
• Then
z  V 1AVz  WAVz  Λz
• Since  is diagonal, the equations are now uncoupled
with
z i  i zi

• So x(t )  Vz (t )
20
 Example
• Assume the previous system with

1 4 
A 
 3 2 
1 4 
V 
1 3 
5 0 
V AV  
-1

 0 2 

21
 Modal Matrices
• Thus the response can be written in terms of the
individual eigenvalues and right eigenvectors as
n
x(t )   v i zi (0 )e i t Note, we are
i 1 requiring that
• Furthermore with the eigenvalues
be distinct!
Δx= VZ  z  V 1x  W T x
• So z(t) can be written as using the left eigenvectors
as
 x1 (t ) 
z (t )  W t x(t )  [ w 1 w 2 ....w n ]t   
 xn (t ) 

22
 Modal Matrices
• We can then write the response x(t) in terms of the
modes of the system
zi (t )  wi t x(t )
zi (0 )  wit x (0 )  ci
n
so x(t )   v i ci e i t
i 1

Expanding xi (t )  vi1c1e 1t  vi 2 c2 e 2t  ... vin cn e nt

• So ci is a scalar that represents the magnitude of

excitation of the ith mode from the initial conditions

23
 Numerical example

 x 1  0 1   x1  1

 x   8 2   x  , x(0)   4 
 2   2  
Eigenvalues are  1  4 ,  2  2
1 1
Eigenvectors are v1    , v 2   
 4  2
 1 1
Modal matrix V   
 4 2 
 0.2425 0.4472 
Normalize so V   
 0. 9701 0. 8944 
24
 Numerical example (contd)

Left eigenvector matrix is:

1.3745 0.6872 
1
W V 
T

1.4908 0.3727 
z = W T AVz
 z 1   4 0   z1 
 z    0 2   z 
 2   2

25
 Numerical example (contd)

z 1  4z1 , z (0 )  V 1x(0 )
 z1 (0 )  4.123 
z 2  2z2 ,    
z (0
 2   ) 0 
4t 4.123 
z1 (t )  z1 (0 )e ; z2 (t )  z2 (0 )e , C  W x(0 )  
2t T

 0 
x = Vz Because of the initial
 x1 (t )   1 1   z1 (t )  condition, the 2nd mode does
 x (t )    4 2   z (t )  not get excited
 2    2 
 0.2425  0.4472  2
z2 (t )   ci v i zi (0 )e
 it
 c1   z1 (t )  c2  
 0.9701 0.8944  i 1
26
26
 Mode Shape, Sensitivity and
Participation Factors
• So we have
x(t )  Vz (t ), z (t )  W t x(t )
• x(t) are the original state variables, z(t) are the
transformed variables so that each variable is
associated with only one mode.
• From the first equation the right eigenvector gives the
“mode shape” i.e. relative activity of state variables
when a particular mode is excited.
• For example the degree of activity of state variable xk
in vi mode is given by the element Vki of the the
right eigenvector matrix V

27
 Mode Shape, Sensitivity and
Participation Factors
• The magnitude of elements of vi give the extent of
activities of n state variables in the ith mode and
angles of elements (if complex) give phase
displacements of the state variables with regard to
the mode.
• The left eigenvector wi identifies which
combination of original state variables display only
the ith mode.

28
 Eigenvalue Parameter Sensitivity
• To derive the sensitivity of the eigenvalues to the
parameters recall Avi = livi; take the partial derivative
with respect to Akj by using the chain rule
A v i i vi
vi  A  v i  i
A kj Akj Akj Akj
Multiply by w ti

A v i t i v i
w t
i vi  w i A
t
 wi vi  w i i
t

A kj Akj Akj Akj

A v i t i
w t
i v i  w i [ A  i I ]
t
 wi vi
Akj Akj Akj
29
 Eigenvalue Parameter Sensitivity
• This is simplified by noting that w ti ( A  i I)  0
by the definition of wi being a left eigenvector
• Therefore
A i
w t
i vi 
Akj Akj
A
• Since all elements of Akj
are zero, except the kth
row, jth column is 1
• Thus i
 WkiV ji
Akj

30
 Sensitivity Example
• In the previous example we had
1 4  1 4  1 3 1 
A  , 1,2  5, 2, V    , W  
 3 2  1 3  7  4 1
• Then the sensitivity of l1 and l2 to changes in A are
i 1 1  3 3  2 1  4 3
 WkiV ji     ,  
Akj A 7  4 4  A 7  4 3 

• For example with 1 4 

ˆ
A , ˆ  5.61, 1.61,

 1,2
3 3 
31
 Eigenvalue Parameter Sensitivity
• This is simplified by noting that w ti ( A  i I)  0
by the definition of wi being a left eigenvector
• Therefore
A i
w t
i vi 
Akj Akj
A
• Since all elements of Akj
are zero, except the kth
row, jth column is 1
• Thus i
 WkiV ji
Akj

32
 Participation Factors

• The participation factors, Pki, are used to determine how

much the kth state variable participates in the ith mode
Pki  VkiWki

• The sum of the participation factors for any mode or any

variable sum to 1
• The participation factors are quite useful in relating the
eigenvalues to portions of a model

33
 Participation Factors

• For the previous example with Pki = VkiWik and

1 4  1 4  1 3 1 
A  , V  , W  
 3 2  1 3  7  4 1
• We get
1  3 4
P 
7  4 3

34
 PowerWorld SMIB
Participation Factors
• The magnitudes of the participation factors are
shown on the PowerWorld SMIB dialog
• The below values are shown for the four bus
example with Q4 = 0

Case is saved as B4_GENROU_Sat_SMIB_QZero

35
 Oscillations
• An oscillation is just a repetitive motion that can be
either undamped, positively damped (decaying with
time) or negatively damped (growing with time)
• If the oscillation can be written as a sinusoid then
e t  a cos  t   b sin  t    e t C cos  t   
 b 
where C  A  B and   tan  
2 2

 a 
• And the damping ratio is defined as (see Kundur 12.46)
 The percent damping is just the damping

 2 2 ratio multiplied by 100; goal is
sufficiently positive damping
36
 Power System Oscillations
• Power systems can experience a wide range of
oscillations, ranging from highly damped and high
frequency switching transients to sustained low
frequency (< 2 Hz) inter-area oscillations affecting an
entire interconnect
• Types of oscillations include
– Transients: Usually high frequency and highly damped
– Local plant: Usually from 1 to 5 Hz
– Inter-area oscillations: From 0.15 to 1 Hz
– Slower dynamics: Such as AGC, less than 0.15 Hz
– Subsynchronous resonance: 10 to 50 Hz (less than
synchronous)
37
 Example Oscillations
• The below graph shows an oscillation that was
observed during a 1996 WECC Blackout
4600 Observed COI Power (Dittmer Control Center)

4400

4200

4000

4600 Simulated COI Power (initial WSCC base case)

4400

4200

4000

0 10 20 30 40 50 60 70 80 90

Time in Seconds
38
 Example Oscillations
• The below graph shows oscillations on the
Michigan/Ontario Interface on 8/14/03

39
Fictitious System Oscillation

Movie shows
an example
of sustained
oscillations in
an equivalent
system

40
Forced Oscillations in WECC (from [1])
• Summer 2013 24 hour data: 0.37 Hz oscillations
observed for several hours. Confirmed to be forced
oscillations at a hydro plant from vortex effect.
• 2014 data: Another 0.5 Hz oscillation also
observed. Source points to hydro unit as well. And
0.7 Hz. And 1.12 Hz. And 2 Hz.
• Resonance is possible when a system mode is
poorly damped and close. Resonance can be
observed in model simulations

1. M. Venkatasubramanian, “Oscillation Monitoring System”, June 2015

http://www.energy.gov/sites/prod/files/2015/07/f24/3.%20Mani%20Oscillation%20Monitoring.pdf 41
 Inter-Area Modes in the WECC
• The dominant inter-area modes in the WECC have
been well studied
• A good reference paper is D. Trudnowski,
“Properties of the Dominant Inter-Area Modes in
the WECC Interconnect,” 2012 Below figure from
– Four well known modes are paper shows NS Mode A
NS Mode A (0.25 Hz), On May 29, 2012
NS Mode B (or Alberta Mode),
(0.4 Hz), BC Mode (0.6 Hz),
Montana Mode (0.8 Hz)

42