Copyright P.

Kundur
This material should not be used without the author's consent

1539pk

Outline

Description of Small Signal Stability Problems

local problems
global problems

Methods of analysis
time-domain analysis and its limitations
modal analysis using linearized model

Characteristics of local plant mode oscillations

Characteristics of interarea oscillations

Enhancement of Small Signal Stability

August 10, 1996 Disturbance of North American

Western Interconnected system

SSS - 1

1539pk

Small-Signal (or Small Disturbance) Stability is the

ability of a power system to maintain synchronism
when subjected to small disturbances
such disturbances occur continually on the system
due to small variations in loads and generation
disturbance considered sufficiently small if
linearization of system equations is permissible for
analysis

Corresponds to Liapunov's first method of stability

analysis

Small-signal analysis using powerful linear analysis

techniques provides valuable information about the
inherent dynamic characteristics of the power system
and assists in its robust design

SSS - 2

1539pk

Instability that may result can be of two forms:

aperiodic increase in rotor angle due to lack of
sufficient synchronizing torque
rotor oscillations of increasing amplitude due to lack
of sufficient damping torque

In today's practical systems, small signal stability is

usually one of insufficient damping of system
oscillations.
local problems or global problems

Local problems involve a small part of the system.

They may be associated with
rotor angle modes
local plant modes
inter-machine modes

control modes
torsional modes

Global problems have widespread effects. They are

associated with
interarea oscillations

SSS - 3

1539pk

Associated with either local plant mode oscillations or

inter-machine oscillations
frequency of oscillation in the range of 0.7 to 2.0 Hz

Local plant mode oscillations

oscillation of a single generator or plant against rest
of the power system

Inter-machine or inter-plant mode oscillations

oscillation between the rotors of a few generators
close to each other

Stability of the local plant mode oscillations is

determined by
strength of transmission as seen by plant
excitation control
plant output and voltage

Instability may also be associated with a non-oscillatory

mode
encountered with manual excitation control

SSS - 4

1539pk

Large interconnected systems usually have two

distinct forms of interarea oscillations:

A very low frequency mode involving all the generators

in the system
system is essentially split into two parts
generators in one part swing against generators in
the other part
frequency in the order of 0.1 to 0.3 Hz

Higher frequency modes involving sub-group of

generators swinging against each other
frequency typically in the range of 0.4 to 0.7 Hz

SSS - 5

1539pk

The 0.28 Hz mode

in WECC

The 0.49 Hz mode

in Eastern US/
Canada
interconnection

SSS - 6

1539pk

1539pk

The behaviour of a dynamic system can be described by

a set of first order differential equations in the
state-space form

x is an n-dimensional state vector

f is an n-dimensional nonlinear function
u is a r-dimensional input vector

The outputs of the system are nonlinear functions of the

state and input vectors

y is an m-dimensional output vector

g is an m-dimensional nonlinear function

At steady state, the system is at an equilibrium point x0

satisfying

SSS - 8

1539pk

Nonlinear dynamic behaviour:

Output variables:

At equilibrium point (x0,u0):

Small perturbation about equilibrium point:

New state equation:

Since perturbations are small:

f(x,u) can be expressed in terms of Taylor's series
expansion
terms involving second and higher order powers of x
and u may be neglected

SSS - 9

1539pk

A, B, C, D are the Jacobians of the system. A is also

referred to as the state matrix or the plant matrix.

SSS - 10

1539pk

Stability is concerned with determination of conditions

of an equilibrium point
what will happen if the system is perturbed at an
equilibrium condition

Stability of a linear system is independent of the input

Stability of a nonlinear system depends on

the type and magnitude of input
the initial state

In control system theory, it is common practice to

classify stability of nonlinear systems into the
following categories, depending on the region of state
space in which the state vector ranges:
local stability or stability in the small
finite stability
global stability or stability in the large

SSS - 11

1539pk

Local stability
The system is said to be locally stable about an
equilibrium point, if when subjected to a small
perturbation, it remains within a small region
surrounding the equilibrium point
If, as time increases, the system returns to the
original state, it is said to be asymptotically stable in
the small

Finite stability
If the state of a system remains within a finite region
R, the system is said to be stable within R
If, further, the state returns to the original equilibrium
point from any point within R, it is said to be
asymptotically stable within the finite region R

Global stability
The system is said to be globally stable if R includes
the entire finite space

SSS - 12

1539pk

The theoretical foundation for the analysis of stability

in the small is based on Liapunov's first method:

The stability in the small of a nonlinear system is given

by the roots of the characteristic equation of the system
of first approximation, i.e., by the eigenvalues of the
state matrix A

If the eigenvalues have negative real parts, then the

original system is asymptotically stable

When at least one of the eigenvalues has a positive real

part, the original system is unstable

SSS - 13

1539pk

Using nonlinear time domain simulations to analyze

small signal stability problems has the following
limitations:

Results can be deceptive

critical mode may not be sufficiently excited by the
chosen disturbance
poorly damped modes may not be dominant in the
observed response

It may be necessary to carry out simulations up to 20

seconds
computational burden is high
massive amount of data has to be analyzed

This approach does not give insight into the nature of

the problem
difficult to identify sources of the problem
mode shapes not clearly identified
corrective measures are not readily indicated

SSS - 14

1539pk

This is a method of spectral analysis

a method of fitting a series of damped sinusoids to a
given signal

For a signal record y(t), the Prony method fits a

function of the form

Ai

the amplitude of the ith mode in the signal

the frequency (Hz) of the ith mode
the damping ratio of the ith mode

the phase (rad) of the ith mode in the signal

Prony analysis has the advantage that it can be used

on any signal: measured or simulated

It can be used together with time-domain simulation to

estimate the frequency and damping of modes of
oscillations

First published in 1795

SSS - 15

1539pk

A linear model can be aggregated from a time domain

response

This gives useful information which can help verify

and complement the results of the linear modal
analysis

However, the basic limitations of the time domain

analysis approach still remain

The spectral analysis is subject to these drawbacks:

it can evaluate only those modes that exist in the
variables monitored
limited or no information about mode shapes
it only looks at the system through a small window
resolution of close modes presents challenges
for a nonlinear system, different windows may give
different results

Often, useful in extracting modal information from

measured time domain response

SSS - 16

1539pk

Modal analysis using eigenvalue approach has proven

to be the most practical way to analyze small signal
stability problems

Advantages are:
individual modes of oscillations are clearly identified
relationships between modes and system variables/
parameters can be easily determined by computing
eigenvectors

Frequency response, poles, zeros, and residues can be

easily computed. Such information is useful in control
system design

SSS - 17

1539pk

Eigenvalues and eigenvectors

is an eigenvalue
is the right eigenvector associated with
is the left eigenvector associated with

Modal matrices

is the right eigenvector matrix

is the left eigenvector matrix

Relationships

I is the unit matrix

is a diagonal matrix: =diag[1... n]

SSS - 18

1539pk

Free motion of a linear dynamic system is described

by

In order to eliminate the cross coupling between the

state variables consider the state transformation

State space equations in z is a set of decoupled

differential equations

The above represents uncoupled first order (scalar)

differential equations

Time domain response

where

is the initial condition

SSS - 19

1539pk

Response in terms of the original state vector:

x(t) = z(t)

The time response of the state variable xi is given by

a linear combination of n dynamic modes

corresponding to the n eigenvalues of the state matrix
ci=ix(0) represents the magnitude of the excitation of
the ith mode due to the initial conditions
if the initial condition lies along the jth eigenvector, only
the jth mode will be excited (since ij=0 for all i j)

If the vector representing the initial condition is not an

eigenvector, it can be represented by a linear
combination of the n eigenvectors. The response of the
system will be the sum of n responses
if a component along an eigenvector of the initial
conditions is zero, the corresponding mode will not be
excited.

SSS - 20

1539pk

A real eigenvalue corresponds to a non-oscillatory

mode

A pair of complex eigenvalues = j correspond to

an oscillatory mode
frequency of the mode

damping ratio of the mode

positive real eigen value indicate

non-oscillatory mode, however, the
magnitude continues to rise towards
instability.

Eigenvalue plots and corresponding

time response
SSS - 21

1539pk

Determines rate of decay of the amplitude of

oscillation

For an oscillatory mode represented by a complex pair

of eigenvalues j, the damping ratio is given by:

The time constant of amplitude decay is

The amplitude decays to 1/e or 37% of the initial

amplitude in seconds or in
cycles of oscillation

As we are dealing with oscillatory modes having a

wide range of frequencies, damping ratio rather than
time constant is more appropriate for expressing the
degree of damping

A 5 sec time constant represents amplitude decay to

37% of initial value in 5 cycles for 1 Hz local plant
mode, and in one-half cycle for 0.1 Hz interarea mode

A damping ratio of 0.03 represents the same degree of

amplitude decay in 5 cycles for all modes

SSS - 22

1539pk

A state variable is related to individual modes by

ji is the ith element in the right eigenvector j

if ji = 0, the jth mode is unobservable in xi
if ji is large, the jth mode will show up strongly in xi
therefore, j determines the mode shape of the jth
mode

The mode shape indicates the relative activities of the

state variables when a particular mode is excited.

The magnitudes of the the eigenvector elements give the

extent of the activities; the angles of the elements give
phase displacements.

The relative phase angles of the elements in the right

eigenvector can be used to determine the direction of
the oscillation in the associated state variables.

SSS - 23

1539pk

Mode shape geographical plot

This plot shows the generator rotor angles observable in

the mode at 0.28 Hz and 1.46% damping. The ones with
red cross oscillate against those with blue circles.
This is a mode in which units in northern WECC swing
against those in the southern WECC.
SSS - 24

1539pk

A mode is related to individual state variables by

ji is the ith element in the left eigenvector j

if ji = 0, the jth mode cannot be controlled by xi
if ji is large, the jth mode is largely determined by xi

One problem in using directly the left eigenvector to

quantitatively determine the contribution of a state
variable to a mode is that the elements of the left
eigenvector are dependent on units and scaling

The solution is to weight the left eigenvector by the right

eigenvector to obtain a quantity independent of unit and
scaling

this is called the participation factor

participation factors measure the participation of state
variables in modes. For instance, the larger pji the
more the state variable xi participates in the jth mode

SSS - 25

1539pk

In frequency domain, a mode refers to a real eigenvalue

or a pair of conjugate complex eigenvalues
Modal analysis of the linearized system model is the
perfect tool to obtain characteristics of individual
modes

In time-domain, a mode is a component in a time

response that has a single frequency and damping,
together with other attributes of the sinusoid (amplitude
and phase angle).
Prony analysis is one way to decompose time-domain
signals to obtain individual modes, although it is often
difficult to obtain the complete modal characteristics

By its dynamic nature, a power system inherently

consists of many modes.
If a mode has poor damping and when it is excited by a
disturbance, it can be observed from the postdisturbance time-domain responses of variables that
have high observability in this mode

When the system is unstable

individual mode(s) unstable

SSS - 26

1539pk

With Mode #1
only (DC offset)

With Modes #1
and #2

Simulated or
measured signal

With Modes #1,

#2, and #3

Signal composed
from modes
identified

Eigenvalue for mode #2: -0.19+j4.52

Eigenvalue for mode #3: -0.26+j4.21

SSS - 27

1539pk

system mode

torsional mode

torsional mode

torsional mode

torsional mode

Fig. 15.3 Rotor natural frequencies and mode shapes of a 555

MVA, 3,600 RPM steam-turbine generator
SSS - 28

1539pk

SSS - 29

1539pk

For a linear dynamic system

Apply state transformation x = z,

If the ith row of matrix -1B is zero, the ith mode is said to
be uncontrollable

If the ith column of matrix C is zero, the ith mode is said

to be unobservable

These concepts are useful in control system design

First proposed by R.E. Kalman in 1960

SSS - 30

1539pk

For a single-input-single-output (SISO) system (assuming

D=0)

The transfer function is given by

S1, S2, ... are poles (eigenvalues)

z1. z2... are zeros
R1, R2, ... are residues, given by:

SSS - 31

1539pk

The key problem in modal analysis is to compute

eigenvalues of the linearized model of power systems

Depending on the system size or analysis objectives,

one of two computation options can be used:

a) Computation of all eigenvalues of the system using QR

method
This is possible for small to medium sized power
systems (up to a couple of thousand states)
All modes present in the model will be captured

b) Computation of partial eigenvalues in specific

frequency ranges using Arnoldi-type algorithm
Applicable to systems of any size
Ideal for analysis of interarea modes
Selective computation also helps focus on the
modes of interest and facilitates analysis

SSS - 32

1539pk

The following shows the comparison of typical

computation speed for time-domain simulations and
eigenvalue calculations:

On one 2.0 GHz Pentium 4 PC with 512 MB memory,

running Windows 2000

For a power system model with 34,381 buses, 3,870

generators, and 41,382 dynamic states
Computation

Time (min)

One 10-second time-domain simulation - TSAT

7.9

Mode scan in the frequency range of 0.2 and 0.8

Hz and damping range of 0% and 10% (24 modes
were computed) - SSAT

2.9

Computation time is scalable to the system size and

number of contingencies to be processed

SSS - 33

1539pk

1539pk

Local mode oscillation problems most commonly

encountered
dates back to the 1950s and 1960s
associated with units of a plant swinging against rest of
system

Characteristics well understood

analysis using block diagram approach (K-constants)
gives physical insight

Encountered by a plant with high output feeding into weak

transmission network (K5 negative)
more pronounced with high response exciters/AVR

Adequate damping readily achieved using Power System

Stabilizers (PSS)
excitation control

SSS - 35

1539pk

First published by Heffron and Phillips to analyze a

single machine (or a plant) connected to a large
system (represented by an infinite bus) through a
transmission network

System is represented by a block diagram as shown in

Fig. SSS-1 with
K constants as parameters
, , fd, Efd as the state variables

This approach is limited to analysis of single machineinfinite bus systems with generator damper windings
neglected
here we use it to complement the results of
eigenvalue analysis
provides physical insight into the effects of
excitation control

SSS - 36

1539pk

= ROTOR ANGLE (rads)

Gex

= EXCITER TRANSFER FUNCTION

= ROTOR SPEED (p.u.)

GPSS

= PSS TRANSFER FUNCTION

fd

= FIELD FLUX LINKAGE

= INERTIA CONSTANT (2H)

Efd

= FIELD VOLTAGE

Ef

= TERMINAL VOLTAGE

Fig. SSS-1 Block diagram of a synchronous machine with

excitation control

SSS - 37

1539pk

The following expressions form the basis for the block

diagram

Rotor acceleration

Electrical torque

Field flux linkage

Terminal voltage

SSS - 38

1539pk

Consider the following simple system

Initial condition
P = 0.9,

Q = 0.3,

Et = 1.0

Small signal and transient stability are studies with

classical model
constant field voltage
static exciter with AVR
static exciter with AVR and PSS

SSS - 39

1539pk

Generator output: P = 0.9, Q = 0.3

K constants:

K1 = 0.758, KD = 0

State variables:

Eigenvalues

Eigenvectors

Participation matrix

zero damping

SSS - 40

1539pk

Block diagram representation

KS=K1=0.758

KD=0

(a)
Constant flux
Positive synchronizing torque coefficient
Negligible damping

SSS - 41

1539pk

K-constants:
K1=0.764, K2=0.865, K3=0.323, T3=2.36, K4=1.42

State variables:
, , fd

Eigenvalues:

positive damping - stable system

1,2 = -0.11 j 6.41

3 = -0.20 + j 0

( = 0.017,

fn = 1.02 Hz)

At 1.02 Hz
KS=K1+KS(fd)=0.7643-0.00172=0.7626 pu torque/rad
KD=KD(fd)=1.53 pu torque/pu speed change

Participation matrix

SSS - 42

1539pk

Generator Output:
P = 1.387,
K1 = 0.457,

Q = 0.462,
K2 = 0.760,

Et = 1.0,
K3 = 0.303,

K4 = 1.57

Eigenvalues:
1,2 = -0.23 j4.95 ( = 0.046, fn = 0.8 Hz)
3 = +0.006 + j 0 (aperiodic instability)

The steady-state synchronizing torque:

Ks = K1 - K2K3K4
= -0.0014

SSS - 43

1539pk

SSS - 44

1539pk

Field flux variations are caused only by feedback of

through K4
this represents the demagnetizing effect of armature
reaction

Constants K2 and K3 are always positive; K4 is usually

positive

Due to the phase lag introduced by field circuit time

constant (T3), the effect of fd due to armature reaction
is to introduce
negative synchronizing torque at low frequencies
positive damping torque and a small negative
synchronizing torque component at typical oscillating
frequencies of 1 Hz

System stability depends on the net synchronizing

torque
reaches steady state stability limit when
K 1 = K 2K 3K 4

SSS - 45

1539pk

Case 1: Generator output

K-constants:
K5 = -0.146,

P=0.9, Q=0.3

K6 = 0.417
oscillations growing

Eigenvalues:
1,2 = +0.5j 7.23

(= -0.07, fn = 1.15 Hz)

3 = -20.2 j 0
4 = -31.2 j 0

oscillatory instability

At 1.15 Hz:
KS(AVR+AR) = 0.2115 pu torque/rad
KD(AVR+AR) = -7.06 pu torque/pu speed change

Case 2: Generator output

K constants:
K5 = 0.025,

P = 0.3, Q = 0.1

K6 = 0.54

Eigenvalues:
1,2 = -0.04j 6.15

( =0.007, fn = 0.98 Hz)

SSS - 46

1539pk

negative K5 introduces negative damping

SSS - 47

1539pk

Effect of AVR depends on the values of K5 and K6

K6 is always positive
K5 can be positive or negative

K5 positive
for low Xe and low P0
effect of AVR is to introduce
negative synchronizing torque
positive damping torque

K5 negative
for high Xe and high P0
effect of AVR is to introduce
positive synchronizing torque
negative damping torque

above effect is more pronounced with higher exciter

response

Situations with K5 negative is commonly encountered

this could lead to oscillatory instability
an effective solution is to use PSS

SSS - 48

1539pk

Uses auxiliary stabilizing signal to control excitation

most logical signal is

If transfer functions of exciter and generator were pure

gains
direct feedback of would result in damping
torque

In practice, the generator and possibly exciter exhibit

frequency dependent gain and phase characteristics
phase compensation results in a pure damping
torque component

Exact phase compensation results in a pure damping

torque component
over compensation introduces negative
synchronizing torque component
under compensation introduces positive
synchronizing torque component

SSS - 49

1539pk

PSS introduce damping torque (proportional to delta 'w' rotor speed devation)
PSS introduce phase lead circuit to compensate for phase lag introduced by exciter.

Block Diagram of a synchronous machine with excitation

control

SSS - 50

1539pk

Fig. 12.14 Thyristor excitation system with AVR and PSS

SSS - 51

1539pk

Generator Output:
P = 0.9,

Q = 0.3

Eigenvalues:
= -1.0 j6.6

( = 0.15, fn = 1.06 Hz)

= -19.8 j12.8
= -39.1 j0
= -0.74 j0

At 1.06 Hz:
KS

= K1+KS(AVR+AR) + KS(PSS)
= 0.7643 + 0.21 - 0.14
= 0.8293 pu torque/rad

KD

= KD(AVR+AR) + KD(PSS)
= -8.69 + 22.77
= 14.08 pu torque/pu speed change

represent rotor angle mode

represent "exciter" mode

SSS - 52

1539pk

Fig. SMIB-1 Rotor Angle

SSS - 53

1539pk

Small signal stability problem is usually one of

insufficient damping of system oscillations

Power system stabilizers (PSS) are the most cost

effective means of solving SSS problems

The purpose is to add damping to the generator rotor

oscillations

This is achieved by modulating the generator

excitation so as to develop a component of electrical
torque in phase with rotor speed deviations

Common input signals include: shaft speed, integral of

power and generator terminal frequency

SSS - 54

1539pk

a) Application to Hydraulic Units:

Used successfully since mid-1960s

Requires minimization of noise

noise below 5 Hz level must be <0.02%
shaft runout (lateral movement) produces most
significant noise components
low frequency noise cannot be removed by
conventional filters; elimination must be intrinsic in
method of signal measurement

Speed outputs are summed from several locations on

shaft

Stabilizer disconnected at gate positions below 70%

prevent effects of turbine vibrations at partial gate
opening

b) Application to Thermal Units:

Stabilizer may cause instability of torsional

oscillations

Speed should be sensed at nodes of torsional modes;

requires torsional filters

SSS - 55

1539pk

The disadvantages of the Delta-Omega stabilizer are:

torsional filer is needed. This can introduce phase
lag at lower frequencies and destabilize exciter
mode
it imposes maximum limit on stabilizer gain
custom design is required for each unit to deal with
torsional modes

Delta-P-Omega stabilizer was developed to overcome

these problems

SSS - 56

1539pk

The frequency signal is obtained in one of two ways

terminal frequency signal is used directly as the
stabilizer input signal
Vt and It are used to derive the frequency of a
voltage behind a simulated machine reactance so as
to approximate the machine rotor speed

On steam-turbine units torsional modes must be

filtered

Gain may be adjusted to obtain the best possible

performance under weak ac transmission system
conditions

Better performance for damping interarea modes than

speed-based stabilizers

Disadvantages
spike may occur in EFD during a rapid transient
(terminal frequency signal will see a sudden phase
shift)
frequency signal often contains power system noise
torsional filtering is required

SSS - 57

1539pk

commonly used PSS

Signal proportional to rotor speed deviation can be

derived from the accelerating power

It is important to derive eq which does not contain

torsional modes

Torsional components are inherently attenuated in the

integral of Pe signal

The problem is to measure the integral of Pm free of

torsional modes

Neglecting Pm is unsatisfactory if mechanical power

changes

Delta-P-Omega uses:

SSS - 58

1539pk

Pm changes are slow; so, simple low pass filter can be

used to remove torsionals

Advantages over the Delta-Omega stabilizer

Pe signal has a high degree of torsional attenuation
generally there is no need for a torsional filter in the

main stabilizing path

higher stabilizer gain is possible which results in

better damping of system oscillations

Standard design for all units (end-of-shaft speed

sensing arrangement with a simple torsional filter)

SSS - 59

1539pk

1539pk

Used to occur in isolated situations

since mid-1980s has become more commonplace
increasingly being identified in planning and
operating studies

In the early 1990s major efforts undertaken to

investigate interarea oscillation problems:
Canadian Electric Association (CEA) Research
Project 294 T622 Report, 1993
IEEE System Oscillations Working Report 95TP101,
1994
CIGRE Technical Brochure on Analysis and Control
of Power System Oscillations" prepared by
TF38.01.07, 1996

We first illustrate the nature of interarea oscillations by

considering a simple two area system
see paper #2 in Appendix

SSS - 61

1539pk

Example 12.6: A Simple Two Area System

Two similar areas connected by a weak tie. Each area

consists of two 900 MVA thermal units, loaded to 700
MW

With all 4 units on manual excitation control,

determine:
all eigenvalues of the system state matrix
for each mode, state variables with high
participation
frequencies, damping ratios, and mode shapes of
rotor angle modes

SSS - 62

1539pk

Example 12.6 (cont'd)

Table E12.3 System modes with manual excitation control

zero
eigen
values

G3
G1

G2

G4

G4

G2
G1

G3

(a) Inter-area mode

(b) Area 1 local mode

(c) Area 2 local mode

f=0.545 Hz, =0.032

f=1.087 Hz, =0.072

f=1.117 Hz, =0.072

Fig. E12.10 Mode shapes of rotor angle modes with manual

excitation control
SSS - 63

1539pk

Example 12.6 (cont'd)

Determine the frequencies and damping ratios of rotor

angle modes with different types of control

Type of excitation
control

Eigenvalue/(frequency in Hz, damping ratio)

Inter-area mode

Area 1 local mode Area 2 local mode

(i) DC exciter

-0.018 j 3.27
(f = 0.52, = 0.005)

-0.485 j 6.81
(f = 1.08, = 0.07)

-0.500 j 7.00
(f = 1.11, = 0.07)

(ii) Thyristor with

high gain

+0.013 j 3.84
(f = 0.61, = -0.008)

-0.490 j 7.15
(f = 1.14, = 0.07)

-0.496 j 7.35
(f =1.17, = 0.07)

(iii) Thyristor with

TGR

+0.123 j 3.46
(f = 0.55, = -0.036)

-0.450 j 6.86
(f = 1.09, = 0.06)

-0.462 j 7.05
(f =1.12, = 0.06)

(iv) Thyristor with

PSS

-0.501 j 3.77
(f = 0.60, = 0.13)

-1.826 j 8.05
(f =1.28, = 0.22)

-1.895 j 8.35
(f =1.33, = 0.22)

Local inter-machine modes have same degree of damping with

DC and thyristor exciters (with and without TGR)

Inter-area mode has

a small positive damping with DC exciter
a small negative damping with a high gain thyristor
exciter
a large negative damping with thyristor exciter with TGR

PSS results in a significant damping of all the rotor angle

modes

SSS - 64

1539pk

Formulation of system state equations uses absolute

change in machine rotor angle and speed
will result in one or two zero eigenvalues

One of the zero eigenvalues associated with lack of

uniqueness of absolute rotor angle
angles of all machines may be changed by same
value without affecting stability
absent if "infinite buses" included

Second zero eigenvalue exists if all generator toques

are independent of speed deviation
no governors and KD = 0

Zero eigenvalues may not be computed exactly due to

limited computational accuracy

SSS - 65

1539pk

Note:

1. These results are from Paper #2 in Appendix

2. Loads assumed to have constant Z characteristics

SSS - 66

1539pk

Case 2B: Generators Real Power

SSS - 67

1539pk

Case 2B: Generators Real Power and Loads

SSS - 68

1539pk

 Oscillations between two groups of generators

Two distinct forms:
a) A very low frequency mode involving all generators
entire system split into two parts, with

generators in one part swinging against

generators in the other part
frequency in the range: 0.1 to 0.3 Hz

b) Higher frequency modes involving a subgroup of

generators swinging against another subgroup
frequency in the range: 0.4 to 0.7 Hz

SSS - 69

1539pk

Interarea Modes in the WSCC system

SSS - 70

1539pk

Many incidents of poorly damped or unstable oscillations

have been reported worldwide:

Michigan-Ontario-Quebec: 0.25 Hz in 1959

Saskatchewan-Manitoba-Ontario West: 0.35-0.45 Hz in the

1960s

WSCC (WECC): Between 0.1 - 0.33 Hz in 1967-1980; 0.28 Hz

August 1996 and August 2000

MAPP: Over 70 unstable oscillations (0.12-0.25 Hz) in 1971 and

1972

NPCC: 0.24-0.4 Hz in 1985

Nordel: 0.5 Hz in late 1960; 0.33 Hz in 1980

Italy-Yugoslavia-Austria: 0.17-0.22 Hz in 1971-1974

Australia: 0.6 Hx in 1975; 0.2 Hz in 1982 and 1983

Scotland-England: 0.5 Hz in 1978 and 1979

Taiwan: 1.0-1.1 Hz in 1984

Ghana-Ivory Coast: 0.6-0.7 Hz in 1985

Southern Brazil: 0.5, 0.8, 1.2 Hz in 1985-1987

Scandinavia: 0.5 Hz in January 1997

SSS - 71

1539pk

First observed in 1985 in planning and operating

studies

Confirmed by on-line measurements

Involved the entire NPCC and eastern interconnected

system

Frequency varied between 0.25 and 0.4 Hz depending

on operating conditions and load levels

Mode shape for one condition shown in Figure

Based on extensive investigations, following remedial

measures taken by OH:
returned PSS on all major units (static exciters)
retrofitted Pickering NGS with PSS (AC exciters);
most effective since close to load centre
PSS for new units at Darlington NGS designed to
damp LFIO
monitors installed throughout Ontario

SSS - 72

1539pk

SSS - 73

1539pk

Type of Excitation Control

Interarea Mode
Frequency

Damping Ratio

a)

Thyristor Exciter with high

transient gain

0.192 Hz

0.009

b)

Thyrister Exciter with

transient gain reduction

0.187 Hz

-0.057

d)

Thyristor Exciter with high

transient gain and PSS

0.179 Hz

0.122

Note: Based on results presented in Paper #1 in Appendix

SSS - 74

1539pk

Conclusions regarding fundamental nature of LFIO:

Characteristics (mode shape, damping) of LFIO are a

complex function of:
network configuration/strength
load characteristics
types of excitation systems and their locations

Load characteristics, in particular, have a major effect

more pronounced with slow exciters

In a stressed system, motor or constant power load at

receiving end has adverse effect on damping
sending end has slightly beneficial effect

A mode of oscillation in one part of system can excite

units in a remote part due to mode coupling

Analysis requires detailed and same level of

representation throughout the system

SSS - 75

1539pk

Damping of LFIO with PSS

The controllability of LFIO with PSS is a function of:

location of units with PSS
characteristics and locations of loads
types of exciters on other units

Damping of LFIO wit PSS achieved primarily by

modulating loads

Identification of units on which PSS most effective:

a high participation factor is a necessary but not
sufficient condition
initial screening by participation factors
residues and frequency responses can supplement
screening

SSS - 76

1539pk

1. Excitation Control: Power System Stabilizers

2. Supplementary Control of HVDC Links, SVCs,

and other FACTS devices

SSS - 77

1539pk

An important index related to small-signal stability

The nature of system response largely depends on
the damping ratios of individual modes

The minimum acceptable level of damping below

which the power system cannot be operated
satisfactorily is not clearly established
Situations with damping ratios of less than 0.02 for
local plant mode and interarea mode oscillations
must be accepted with caution

In addition to the absolute value of damping ratio, what

is important is its sensitivity to variations in operating
conditions and system parameters
A low damping ratio but less sensitive to operating
conditions and system parameters is often
acceptable

In the design of power system stabilizer and other

forms of controllers for damping power system
oscillations, a good design target is to have a damping
ratio of at least 0.1 for the mode(s) of interest

SSS - 78

1539pk

Design objectives:

Maximize the damping of the local plant modes as well

as interarea mode oscillations without compromising
the stability of other modes

Enhance system transient stability

Not adversely affect system performance during major

system upsets which cause large frequency
excursions

Minimize the consequences of excitation system

malfunction due to component failures

SSS - 79

1539pk

Fig. 17.5 Block diagram of thyristor excitation system

with PSS

SSS - 80

1539pk

Exciter gain
high value of KA for transient stability (200)
transient gain reduction (TGR) is required only if
voltage regulator time constant is large or exciter
has significant time delays
TA about 1 second
TB about 10 seconds
TGR not required for Thyristor exciters

Phase lead compensation

compensate for lag between exciter input and
resulting electrical torque
design should provide damping over wide range of
frequency to cover local and interarea modes

SSS - 81

1539pk

Phase lead compensation (con't)

compute the frequency response between the
exciter input and the generator electrical torque with
the generator speed and rotor angle remaining
constant (assuming large inertia for the machine)
frequency response of any machine is sensitive to
the Thevenin equivalent system impedance at its
terminals but relatively independent of the dynamics
of other machines (assuming other machines are
infinite buses)
resulting phase characteristic has a relatively simple
form free from the effects of natural frequencies of
the external machines
select characteristic suitable for different system
conditions
Do not overcompensate !

SSS - 82

1539pk

High pass filter which prevents steady change in

speed from modifying the field voltage

Washout time constant TW should be high enough to

allow signals associated with oscillations in rotor
speed to pass unchanged

From the viewpoint of washout function, can be in the

range of 1 to 10 seconds
pass stabilizing signals at the frequencies of interest
not so long that it leads to undesirable generator
voltage excursions as a result of stabilizer action during
system islanding conditions

For local plant modes, TW of 1.5 s or higher

satisfactory

TW of less than 5 s results in significant phase lead at

low frequencies associated with interarea oscillations
this can reduce the synchronizing torque component

For systems with dominant interarea oscillations

either set TW to about 10 s, or
use one of the phase compensation blocks to provide
phase lag at low frequencies

SSS - 83

1539pk

The stabilizer gain, KSTAB, has an important effect on

damping or rotor oscillations

It is necessary to examine the effect of KSTAB for a

wide range of values

Damping increases with an increase in KSTAB up to a

point

Gain is set to provide maximum damping with the

following considerations:
Delta-Omega stabilizer: due to the effect of the
torsional filter, the stability of the "exciter mode"
becomes an overriding consideration
Delta-P-Omega stabilizer: exciter mode stability is
not a problem, and a considerably higher value of
gain is acceptable limited only by amplification of
signal noise considerations

SSS - 84

1539pk

Positive output limit is set at a relatively large value

(0.1 to 0.2 pu)

This allows a high level of contribution from PSS

during large swings

Et limiter needed

High gain limiter can cause torsional mode instability

(Et has small components of torsionals) ... choose TC
and TD to provide high attenuation at torsional
frequencies, in addition to ensuring adequate degree
of limiter loop stability

Negative limit of -0.05 to -0.1 pu allows sufficient

control range and satisfactory transient response

SSS - 85

1539pk

Overall performance should be evaluated for the

stabilizer parameter setting
small signal stability program can be used to
examine performance over range of conditions
there should be no adverse interactions with the
controls of other nearby generating units and
devices, such as HVDC converters and SVCs
transient stability and long-term stability simulations
should also be used to verify the performance
coordination with other protections and controls,
such as Volts/Hz limiters and overexcitation/
underexcitation protection

SSS - 86

1539pk

Actual response of unit with PSS measured and used

to verify analytical results
step change to AVR reference
disturbance external to plant, e.g., line switching

If there are discrepancies between measured and

computed responses
models modified and revised PSS settings
determined

During commissioning, PSS gain increased slowly up

to twice the chosen setting
exciter mode stability margin
input signal noise amplification

On-line tuning of PSS impractical !

SSS - 87

1539pk

Allow setting of PSS parameters over sufficiently wide

range

Ensure high degree of functional reliability and

flexibility for maintenance
component redundancy
duplicate PSS, AVR

Built-in facilities for dynamic tests

routine testing to avoid undetected failures

Importance of good design features often overlooked

many instances of operators disconnecting PSS

SSS - 88

1539pk

Two thermal 488 MVA units equipped with thyristor

excitation systems

Units exhibit two dominant rotor oscillation modes: an

interarea mode of about 0.5 Hz and a local intermachine mode of about 2.0 Hz

Objective of excitation control design is to enhance

the transient as well as small signal stability of the
power system

Examine the performance using slow rotating exciters

also

SSS - 89

1539pk

Thyristor excitation system:

Power system stabilizer

High exciter gain of 212 (with no TGR) is used to

ensure good transient stability performance

SSS - 90

1539pk

Slow rotating excitation system:

Self-excited dc exciter

1. Type DC1A exciter model

SSS - 91

1539pk

Slow signal stability performance:

Local inter-machine mode

Inter-area mode

Type of exciter
Frequency

Frequency

(a) Thyristor (no PSS)

1.823 Hz

0.049

0.550 Hz

0.006

(b) Rotating exciter

1.793 Hz

0.075

0.498 Hz

0.046

Selection of PSS parameters

Determination of phase-lead compensation

G1 and G2 are represented with large inertia - all

others as infinite buses

SSS - 92

1539pk

Parameters for compensation shown

T1 = 0.06 T2 = 0.02 T3 = 1.5

KSTAB

T4 = 4.0

Local inter-machine mode

TW = 7.5
Inter-area mode

Frequency

Frequency

1.823 Hz

0.049

0.550 Hz

0.006

20

2.079 Hz

0.156

0.547 Hz

0.087

30

2.218 Hz

0.197

0.548 Hz

0.124

40

2.366 Hz

0.227

0.533 Hz

0.156

Other parameters
Vsmax = 0.2
TC = 0.025

Vsmin = -0.05
TD = 1.212

VLS = 1.15
TRL = 0.01

SSS - 93

KL = 17

1539pk

Fig. 17.13 Response of unit G1 to a five-cycle threephase fault; peak load conditions
SSS - 94

1539pk

For damping inter-area oscillations, best locations for

PSS may not be obvious in large systems

PSS adds damping to an inter-area mode largely by

modulating system loads

PSS with regard to a local mode is only slightly

affected by the load characteristics

Understanding these mechanisms is essential

Participation factors corresponding to speed

deviations of generating units are very useful for initial
screening

Rigorous evaluation using residues and frequency

responses should be carried out to determine
appropriate locations for the stabilizers

SSS - 95

1539pk

Main tasks in controller design:

Selection of device and its location

option may not always exist
based on participation factors, residues, and
frequency response

Selection of feedback signal:

modes of concern must be observable in signal(s)
based on frequency response between device input
and potential signals

Controller design procedure

a variety of linear techniques available
varying degrees of procedure automation,
complexity, robustness, and applicability to large
system

Test overall performance

broad range of conditions and contingencies
small-signal and transient stability

SSS - 96

1539pk

The physical mechanism by which non-generator

devices contribute to damping of inter-area
oscillations in highly meshed networks is not obvious

Control design techniques based on physical

principles, such as in PSS design, cannot be readily
applied

Control design methods based on linear control theory

were investigated

SSS - 97

1539pk

Phase and gain margin

based on Nyquist criteria
controller is designed to improve phase and gain
margins of the closed loop system

Pole placement
controller is designed so that the closed loop
system has a pole (eigenvalue) at a specific location

H-infinity
a computer aided control design technique
minimize H-infinity norm of the system transfer
function from the disturbance to the output over the
set of all stabilizing controllers
produces a controller that is robust in some sense
reduced order system model required

SSS - 98

1539pk

Design of controller for an embedded HVDC link (with

parallel ac paths)

System based on the WSCC system with two HVDC

links:
Pacific Intertie, and
Intermountain

Controller designed for Pacific DC Intertie to improve

damping of 0.3 Hz north-south oscillations

All three methods of control design gave satisfactory

results

SSS - 99

1539pk

All ccts I/S

Grizzly-Malin O/S

Mode
No.

Freq. (Hz)

Damp Ratio

Freq. (Hz)

Damp Ratio

1.

0.298

0.079

0.284

0.077

2.

0.446

0.059

0.442

0.058

3.

0.607

0.045

0.606

0.046

4.

0.735

0.011

0.724

0.006

5.

0.747

0.073

0.746

0.072

6.

0.779

0.048

0.781

0.053

7.

0.804

0.045

0.796

0.042

8.

0.862

0.015

0.861

0.015

Note: Active component of loads assumed to have constant

current characteristics

SSS - 100

1539pk

Mode to be damped, must be

observable in the signal
controllable by the device

Mode 1 (north-south mode) is controllable by the DC Intertie

Mode 4 (Arizona-California) is not controllable by the DC
Intertie but may be stabilized by PSS at Helms GS in California
PSS location identified by participation factor

Several signals considered for modulating the DC Intertie

difference between angles of rectifier and inverter ac bus
voltages selected for modulating the rectifier controls

SSS - 101

1539pk

Frequency Response for Angle Difference in the Post

Disturbance System
SSS - 102

1539pk

For application H-infinity robust controller design

technique

Objective:
Reduce a large order system model with up to 20,000
states to a transfer function model of order less than
15 which captures the essential characteristics of the
system

Approaches
compute poles and zeros in a dynamically reduced
model and eliminate close poles/zeros
compute important poles and zeros in large system
and supplement the transfer function using system
identification techniques

SSS - 103

1539pk

Full system: 3866 states

Reduced transfer function: 12th order

SSS - 104

1539pk

Frequency Response Between the Rectifier Current Reference and

Difference Between Rectifier and Inverter AC Voltage Phase Angles
SSS - 105

1539pk

Controller
Design

None

Pole
Modification

Phase/Gain
Margin

H-infinity

Freq. (Hz)

0.284

0.294

0.316

0.311

Damp Ratio

0.077

0.169

0.173

0.176

PSS

Out-of-Service

In-Service

Freq. (Hz)

0.724

0.730

Damp Ratio

0.006

0.016

SSS - 106

1539pk

Method

Allowable
System
Model

Effect on
Other Modes

Achievement of
Robustness

Handle on
Damping

Phase and Gain

Margins

Large

Can be
considered

Compromise in
design

Good

Pole Placement

Large

Difficult to
predict

Compromise in
design

Very good

H-infinity

Reduced
(<15 states)

Considered
in design

Automatically
considered

Weak

SSS - 107

1539pk

General:

Each of the three methods has its advantages and

disadvantages

With care and judgment, any method can be applied

successfully

H-infinity:

Robustness is built into the method

Selection of the weighting functions is critical

Method needs reformulation to be more directly

applicable to power systems

SSS - 108

1539pk

SSS - 109

1539pk

High ambient temperatures in Northwest; high power

transfer from Canada to California

Prior to main outage, three 500 kV line sections from

lower Columbia River to load centres in Oregon were
out of service due to tree faults

Line outages caused voltage reduction in lower

Columbia River area from around 540 kV to 510 kV

Main outage at 15:47:36, loss of Ross-Lexington 230 kV

line due to tree flash over

Growing 0.234 Hz oscillations caused:

tripping of lines resulting in formation of four islands
loss of 30,500 MW load

SSS - 110

1539pk

4
4
1
2

Event 5:
Event 4:
Event 3:
Event 2:
Event 1:

15:47:36-15:48:09
8 McNary Units trip
15:47:36 Ross-Lexington 230 kV - flashed to tree (+ 115 kv cct loss)
15:42:03 Keeler-Alliston 500 kV - LG - flashed to tree
14:52:37 John Day - Marion 500 kV LG - flashed to tree
14:06:39 Big Eddy - Ostrander 500kV LG fault - flashed to tree

SSS - 111

1539pk

Recordings showed undamped oscillations throughout

WSCC with a frequency of about 0.23 Hz

SSS - 112

1539pk

As a result of the
undamped
oscillations, the
system split into
four large islands

Over 7.5 million

customers experienced
outages ranging from a
few minutes to nine
hours! Total load loss
30,500 MW

SSS - 113

1539pk

MEASURED
RESPONSE

SIMULATED
RESPONSE

SSS - 114

1539pk

SSS - 115

1539pk

Mode shape and participation

factors were computed for the
critical mode
Participations of generator
speed terms, controllability
and observability used to find
best locations for PSS tuning
or additions

SSS - 116

1539pk

San Onofre
(Addition) Palo Verde
(Tune existing)

SSS - 117

1539pk

With existing controls

Eigenvalue = 0.0597 + j 1.771
Frequency = 0.2818 Hz
Damping ratio = -0.0337

With PSS modifications

Eigenvalue = -0.0717 + j 1.673
Frequency = 0.2664
Damping ratio = 0.0429

SSS - 118

1539pk

HVDC shown (as expected) to have low participation in

mode

Often however, HVDC can be modulated to improve

damping provided adequate input signal is found and
proper compensator is designed

Frequency responses were examined for several

potential input signals

Frequency response magnitude identifies local bus

frequency as having good observability/
controllability of mode of interest

Frequency response phase used to design

compensator which provides proper modulation signal
to HVDC controls

Time-domain and eigenvalue analysis used to verify

modulation performance

SSS - 119

1539pk

Without HVDC Modulation

Eigenvalue = 0.0597 + j
1.771
Simulation event

Frequency = 0.2818 Hz
Damping ratio = -0.0337

With HVDC Modulations

Eigenvalue = -0.108 + j 1.797
Frequency = 0.2859 Hz
Damping ratio = 0.0602

SSS - 120

1539pk

Copies of Papers:
1. Application of Power System Stabilizers for
Enhancement of Overall System Stability
2. A Fundamental Study of Inter-Area Oscillations in
Power Systems
3. Analytical Investigation of Factors Influencing Power
System Stabilizers Performance
4. Effective Use of Power system Stabilizers for
Enhancement of Power system Reliability

SSS - 121

1539pk

SSS - 122

1539pk

SSS - 123

1539pk

SSS - 124

1539pk

SSS - 125

1539pk

SSS - 126

1539pk

SSS - 127

1539pk

SSS - 128

1539pk

SSS - 129

1539pk

SSS - 130

1539pk

SSS - 131

1539pk

SSS - 132

1539pk

SSS - 133

1539pk

SSS - 134

1539pk

SSS - 135

1539pk

SSS - 136

1539pk

SSS - 137

1539pk

SSS - 138

1539pk

SSS - 139

1539pk

SSS - 140

1539pk

SSS - 141

1539pk

SSS - 142

1539pk

SSS - 143

1539pk

SSS - 144

1539pk

SSS - 145

1539pk

SSS - 146

1539pk

SSS - 147

1539pk

SSS - 148

1539pk

SSS - 149

1539pk

SSS - 150

1539pk

SSS - 151

1539pk

SSS - 152

1539pk

SSS - 153

1539pk

SSS - 154

1539pk

SSS - 155

1539pk

SSS - 156

1539pk

SSS - 157

1539pk

SSS - 158

1539pk

SSS - 159

1539pk