Dynamics and Control of

Electric Power Systems

Lecture 227-0528-00, ITET ETH

Goran Andersson
EEH - Power Systems Laboratory

ETH Z
urich
March 2007

ii

Contents
Preface

1 Introduction
1.1 Control Theory Basics - A Review
1.1.1 Simple Control Loop . . . .
1.1.2 State Space Formulation . .
1.2 Control of Electric Power Systems

.
.
.
.

1
2
2
5
5

2 Frequency Control in Electric Power Systems

2.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Model of Frequency Control . . . . . . . . . . . . . . . . . . .
2.2.1 Frequency Dependency of Loads . . . . . . . . . . . .

9
9
12
13

3 Primary Frequency Control

3.1 Introduction . . . . . . . .
3.2 Turbine Control . . . . . .
3.2.1 Turbine Models . .
3.2.2 Turbine Governors
3.3 Role of Speed Droop . . .

.
.
.
.
.

17
17
22
22
31
33

.
.
.
.

41
41
44
44
53

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

4 Load Frequency Control

4.1 Automatic Generation Control - Static
4.2 Dynamic Model . . . . . . . . . . . . .
4.2.1 One Node System . . . . . . .
4.2.2 Two Area System . . . . . . .

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

Model
. . . .
. . . .
. . . .

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

5 Model of the Synchronous Machine

59
5.1 Parks Transformation . . . . . . . . . . . . . . . . . . . . . . 59
5.2 The Inductances of the Synchronous Machine . . . . . . . . . 63
5.3 Voltage Equations for the Synchronous Machine . . . . . . . . 65
5.4 Synchronous, Transient, and Subtransient Inductances, and
Time Constants. . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.5 Simplified Models of the Synchronous Machine . . . . . . . . 73
iii

iv
6 The
6.1
6.2
6.3
6.4
6.5

Contents
Excitation System of the Synchronous
Construction of the Excitation System . . .
Compensation Equipment . . . . . . . . . .
DC Excitation Systems . . . . . . . . . . .
AC Excitation Systems . . . . . . . . . . . .
Static Excitation Systems . . . . . . . . . .

Machine
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

77
77
78
79
80
80

7 Damping in Power Systems

83
7.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.2 Causes of Damping . . . . . . . . . . . . . . . . . . . . . . . . 84
7.3 Methods to Increase Damping . . . . . . . . . . . . . . . . . . 85
8 Load Modelling
8.1 The Importance of the Loads for System
8.2 Load Models . . . . . . . . . . . . . . .
8.2.1 Static Load Models . . . . . . . .
8.2.2 Motor Loads . . . . . . . . . . .
8.2.3 Equivalent Dynamic Loads . . .

Stability
. . . . . .
. . . . . .
. . . . . .
. . . . . .

.
.
.
.
.

References

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

87
87
88
88
89
90
93

A Connection between per unit and SI Units for the Swing

Equation
95
B Influence of Rotor Oscillations on the Curve Shape

97

Preface
These lectures notes are intended to be used in the Systemdynamik und Leittechnik der elektrischen Energieversorgung (Vorlesungsnummer ETH Z
urich
227-0528-00) lectures given at ETH Z
urich in the eighth semester in Information Technology and Electrical Engineering.
The main topic covered is frequency control in power systems. The
needed models are derived and the primary and secondary frequency control are studied. A detailed model of the synchronous machine, based on
Parks transformation, is also included. The excitation and voltage control
of synchronous machines are briefly described. An overview of load models
is also given.
Z
urich in March 2007
G
oran Andersson

vi

Preface

1
Introduction
In this chapter a general introduction to power systems control is given.
Some basic results from control theory are reviewed, and an overview of the
use of different kinds of power plants in a system is given.

The main topics of these lectures will be

Power system dynamics
Power system control
Security and operational efficiency.
In order to study and discuss these issues the following tools are needed
Control theory (particularly for linear systems)
Modelling
Simulation
Communication technology.
The studied system comprises the subsystems Electricity Generation, Transmission, Distribution, and Consumption (Loads), and the associated control
system has a hierarchic structure. This means that the control system consists of a number of nested control loops that control or regulate different
quantities in the system. In general the control loops on lower system levels,
e.g. locally in a generator, are characterized by smaller time constants than
the control loops active on a higher system level. As an example, the Automatic Voltage Regulator (AVR), which regulates the voltage of the generator
terminals to the reference (set) value, responds typically in a time scale of a
second or less, while the Secondary Voltage Control, which determines the
reference values of the voltage controlling devices, among which the generators, operates in a time scale of tens of seconds or minutes. That means that
these two control loops are virtually de-coupled. This is also generally true
for other controls in the systems, resulting in a number of de-coupled control
loops operating in different time scales. A schematic diagram showing the
different time scales is shown in Figure 1.1.
1

1. Introduction
Tie Line Power and
Frequency Control
Turbine Control
Voltage Control
Protection

1/10

10

100

Time (s)

Figure 1.1. Schematic diagram of different time scales of power system controls.

The overall control system is very complex, but due to the de-coupling
it is in most cases possible to study the different control loops individually.
This facilitates the task, and with appropriate simplifications one can quite
often use classical standard control theory methods to analyse these controllers. For a more detailed analysis, one has usually to resort to computer
simulations.
A characteristic of a power system is that the load, i.e. the electric
power consumption, varies significantly over the day and over the year. This
consumption is normally uncontrolled. Furthermore, since substantial parts
of the system is exposed to external disturbances, the possibility that lines
etc. could be disconnected due to faults must be taken into account. The
task of the different control systems of the power system is to keep the power
system within acceptable operating limits such that security is maintained
and that the quality of supply, e.g. voltage magnitudes and frequency, is
within specified limits. In addition, the system should be operated in an
economically efficient way. This has resulted in a hierarchical control system
structure as shown in Figure 1.2.

1.1

Control Theory Basics - A Review

The de-coupled control loops described above can be analyzed by standard

methods from the control theory. Just to refresh some of these concepts,
and to explain the notation to be used, a very short review is given here.

1.1.1

Simple Control Loop

The control system in Figure 1.3 is considered. In this figure the block G(s)
represents the controlled plant and also possible controllers. From this figure

1.1. Control Theory Basics - A Review

System Control
Centre
State Estimation
Power Flow Control
Economical Dispatch
Security Assessment

NETWORK
Generation
(in power station)
Turbine Governor
Voltage Gontrol

Power Transmission
and Distribution
Tap Changer Control
(Direct and Quadrature)
Reactive power
compensation

Loads

In normal case
not controlled

HVDC, FACTS

Figure 1.2. The structure of the hierarchical control systems of a power system.

the following quantities are defined

r(t) = Reference (set) value (input)

e(t) = Control error
y(t) = Controlled quantity (output)
v(t) = Disturbance
Normally the controller is designed assuming that the disturbance is equal
to zero, but to verify the robustness of the controller realistic values of v
must be considered.
In principle two different problems are solved in control theory:
1. Regulating problem
2. Tracking problem
1
Here the quantities in the time domain are denoted by small letters, while the Laplace
transformed corresponding quantities are denoted by capital letters. In the following this
convention is not always adhered to, but it should be clear from the context if the quantity
is expressed in the time or the s domain.

1. Introduction

v
r

G(s)
_

H(s)

Figure 1.3. Simple control system with control signals.

In the regulating
problem, the reference value r is normally kept constant
P
and the task is to keep the output close to the reference value even if disturbances occur in the system. This is the most common problem in power
systems, where the voltage, frequency and other quantities should be kept
at the desired values irrespective of load variations, line switchings, etc.
In the tracking problem the task is to control the system so that the
output y follows the time variation of the input r as good as possible. This
is sometimes also called the servo problem.
The transfer function from the input, R, to the output, Y , is given by
(in Laplace transformed quantities)
F (s) =

C(s)
G(s)
Y (s)
=
=
R(s)
R(s)
1 + G(s)H(s)

(1.1)

In many applications one is not primarily interested in the detailed time

response of a quantity after a disturbance, but rather the value direct after
the disturbance or the stationary value when all transients have decayed.
Then the two following properties of the Laplace transform are important:
g(t 0+) = lim sG(s)

(1.2)

g(t ) = lim sG(s)

(1.3)

and
s0

where G is the Laplace transform of g. If the input is a step function,

Laplace transform = 1/s, and F (s) is the transfer function, the initial and
stationary response of the output would be
y(t 0+) = lim F (s)

(1.4)

y(t ) = lim F (s)

(1.5)

and
s0

1.2. Control of Electric Power Systems

1.1.2

State Space Formulation

A linear and time-invariant controlled system is defined by the equations

x = Ax + Bu
(1.6)

y = Cx + Du

The vector x = (x1 x2 . . . xn )T contains the states of the system, which

uniquely describe the system. The vector u has the inputs as components,
and the vector y contains the outputs as components. The matrix A, of
dimension n n, is the system matrix of the uncontrolled system. The
matrices B, C, and D depend on the design of the controller and the available
outputs. In most realistic cases D = 0, which means that there is zero
feedthrough, and the system is said to be strictly proper. The matrices A
and B define which states that are controllable, and the matrices A and
C which states that are the observable. A controller using the outputs as
feedback signals can be written as u = Ky = KCx, assuming D = 0,
where the matrix K defines the feedback control, the controlled system
becomes
x = (A BKC)x
(1.7)

1.2

Control of Electric Power Systems

The overall control task in an electric power system is to maintain the balance between the electric power produced by the generators and the power
consumed by the loads, including the network losses, at all time instants.
If this balance is not kept, this will lead to frequency deviations that if too
large will have serious impacts on the system operation. A complication is
that the electric power consumption varies both in the short and in the long
time scales. In the long time scale, over the year, the peak loads of a day are
in countries with cold and dark winters higher in the winter, so called winter
peak, while countries with very hot summers usually have their peak loads
in summer time, summer peak. Examples of the former are most European
countries, and of the latter Western and Southern USA. The consumption
vary also over the day as shown in Figure 1.4. Also in the short run the
load fluctuates around the slower variations shown in Figure 1.4, so called
spontaneous load variations.
In addition to keeping the above mentioned balance, the delivered electricity must conform to certain quality criteria. This means that the voltage
magnitude, frequency, and wave shape must be controlled within specified
limits.
If a change in the load occurs, this is in the first step compensated by
the kinetic energy stored in the rotating parts, rotor and turbines, of the

1. Introduction

120

100

100

weekdays

weekdays

60

Load (%)

Load (%)

80

holidays

40

80

holidays
holidays

60
40

20

20
0

0
1

10

13

Time

16

19

22

10

13

16

19

22

Time

Figure 1.4. Typical load variations over a day. Left: Commercial

load.; Right: Residential load.

generators resulting in a frequency change. If this frequency change is too

large, the power supplied from the generators must be changed, which is done
through the frequency control of the generators in operation. An unbalance
in the generated and consumed power could also occur as a consequence of
that a generating unit is tripped due to a fault. The task of the frequency
control is to keep the frequency deviations within acceptable limits during
these events.
To cope with the larger variations over the day and over the year generating units must be switched in and off according to needs. Plans regarding
which units should be on line during a day are done beforehand based on
load forecasts 2 . Such a plan is called unit commitment. When doing such
a plan, economic factors are essential, but also the time it takes to bring a
generator on-line from a state of standstill. For hydro units and gas turbines this time is typically of the order of some minutes, while for thermal
power plants, conventional or nuclear, it usually takes several hours to get
the unit operational. This has an impact on the unit commitment and on
the planning of reserves in the system3 .
Depending on how fast power plants can be dispatched, they are classified as peak load, intermediate load, or base load power plants. This
classification is based on the time it takes to activate the plants and on the
2
With the methods available today one can make a load forecast a day ahead which
normally has an error that is less than a few percent.
3
In a system where only one company is responsible for the power generation, the unit
commitment was made in such a way that the generating costs were minimized. If several
power producers are competing on the market, liberalized electricity market, the situation
is more complex. The competing companies are then bidding into different markets, pool,
bi-lateral, etc, and a simple cost minimizing strategy could not be applied. But also in
these cases a unit commitment must be made, but after other principles.

1.2. Control of Electric Power Systems

fuel costs and is usually done as below4 . The classification is not unique and
might vary slightly from system to system.
Peak load units, operational time 10002000 h/a
Hydro power plants with storage
Pumped storage hydro power plants
Gas turbine power plants
Intermediate load units, operational time 30004000 h/a
Fossil fuel thermal power plants
Bio mass thermal power plants
Base load units, operational time 50006000 h/a
Run of river hydro power plants
Nuclear power plants
In Figure 1.5 the use of different power plants is shown in a load duration
curve representing one years operation.
The overall goal of the unit commitment and the economic dispatch is
the
Minimization of costs over the year
Minimization of fuel costs and start/stop costs

The fuel costs should here be interpreted more as the value of the fuel. For a hydro
power plant the fuel has of course no cost per se. But if the hydro plant has a storage
with limited capacity, it is obvious that the power plant should be used during high load
conditions when generating capacity is scarce. This means that the water value is high,
which can be interpreted as a high fuel cost.

1. Introduction

System Load

Gas turbines
Hydro power reserves
Thermal power (fossil fuel)
Controllable hydro power

Nuclear power
Run of river hydro power

1 year

Time

Figure 1.5. Duration curve showing the use of different kinds of power plants.

2
Frequency Control in Electric Power
Systems
In this chapter, a model that can be used for studying frequency variations
following a disturbance, like generator or load tripping, in an electric power
system is developed. Models for the frequency dependency of loads are discussed.

2.1

System Model

It was discussed in chapter 1 that the systems frequency will deviate from
the desired (or nominal) frequency if the balance between power generation
and load consumption (including losses) is not maintained. Since the swing
equations constitute a general model of the movement of the generators
rotors, these equations can be used for studying the variation of the system
frequency.
After a disturbance, like loss of production, in the system, the frequency
in different parts of the system will vary according to Figure 2.1. The
frequencies of the different machines can be viewed as comparatively small
variations over an average frequency in the system. This average frequency,
called the system frequency in the following, is the frequency that can be
defined for the socalled centre of inertia (COI) of the system.
We want to derive a model that is valid for reasonable frequency deviations, the exact version of swing equation will be used, i.e.
i =

0
(Tmi (p.u.) Tei (p.u.)) ,
2Hi

(2.1)

with the usual notation. The indices m and e denote mechanical (turbine)
and electrical quantities respectively. i is used here to denote the deviation in rotor angular frequency as compared with the nominal one 0 .
(For rotor oscillations the frequency of i is often of interest, while the
amplitude of i is the main concern in frequency control.) By extending
the right hand side with the actual angular frequency and expressing the
9

10

2. Frequency Control in Electric Power Systems

50

49.8

f(Hz)

49.6

49.4

49.2

49

48.8

t(s)

Figure 2.1. The frequency in different locations in an electric power

system after a disturbance. The thicker solid curve indicates the average system frequency. Other curves depict the frequency of individual
generators.

power in SIunits, like MW instead of p.u.,

02
(Pmi Pei )
i

2Hi Si i =

(2.2)

is obtained. Pmi and Pei are now expressed in the same SIunit as Si ,
cf. Appendix A. Further, for every synchronous machine i
i = 0 + i ,

(2.3)

and for the frequency in the centre of inertia (COI)

= 0 + .
The system frequency deviation (or ) is now defined as
X
Hi Si i = H

(2.4)

(2.5)

with
H=

Hi Si .

(2.6)

Adding eq. (2.2) for all generators in the system leads to

X Pmi Pei
2H = 02
.
i
i

(2.7)

11

2.1. System Model

A very simple and useful model can be derived if some more assumptions
are made.
The overall goal of our analysis is to derive an expression
P that gives
P the
variation of after disturbance of the balance between
Pmi and
Pei .
Therefore, we define
Pm =

Pmi = Pm0 + Pm ,

(2.8)

where Pm denotes a possible disturbance, like loss of a generator (turbine).

The total generated power is consumed by the loads and the transmission
system losses, i.e.
X
Pe =
Pei = Pload + Ploss ,
(2.9)
i

which can, in the same way as in eq. (2.8), be written as

Pe = Pe0 + Pload + Ploss

(2.10)

Pe0 = Pload0 + Ploss0 .

(2.11)

with
If the system is in equilibrium prior to the disturbance,
Pm0 = Pe0

(2.12)

Pm0 = Pload0 + Ploss0

(2.13)

and
are valid.
The following assumptions are now made:
The transmission losses after and before the disturbance are equal,
i.e. Ploss = 0.
All generator angular frequencies i in the right hand side of eq. (2.7)
are set equal to the system average angular frequency .
If neither the disturbance nor the oscillations in the transmission system are
too large, these approximations are reasonable. Using these assumptions
and eqs. (2.8) (2.13), eq. (2.7) can now be written as
=

02
(Pm Pload ) .
2H

Eq. (2.14) can be represented by the block diagram in Figure 2.2.

(2.14)

12

2. Frequency Control in Electric Power Systems

P m

*/*

b
0

P load

a/b

0
-------2H

1
--s

Figure 2.2. Block diagram of eq. (2.14).

2.2

Model of Frequency Control

Eq. (2.14) describes the variation of the systems average frequency when
the balance between generated and demanded power is no longer preserved,
i.e. when Pm 6= Pload . In the most common case Pm Pload is
negative after a disturbance, like the tripping of a generator station. It is
also possible that the frequency rises during a disturbance, for example when
an area that contains much generation capacity is isolated. Since too large
frequency deviations in a system are not acceptable, automatic frequency
control, which has the goal of keeping the frequency during disturbances at
an acceptable level, is used. The spontaneous load variations in an electric
power system result in a minutetominute variation of up to 2%. This
alone requires that some form of frequency control must be used in most
systems.
There are at least two reasons against allowing the frequency to deviate
too much from its nominal value. A nonnominal frequency in the system
results in a lower quality of the delivered electrical energy. Many of the
devices that are connected to the system work best at nominal frequency.
Further, too low frequencies (lower than 47 48 Hz) lead to damaging vibrations in steam turbines, which in the worst case have to be disconnected.
This constitutes an even worse stress on the system and can endanger the
systems security. (Hydro power plants are more robust and can normally
cope with frequencies down to 45 Hz.)
The frequency after a disturbance thus depends on the parameters contained in eq. 2.14). In addition to the constant of inertia, H, the trajectory
is determined by
the frequency dependency of the loads,
the control of the hydro turbines,

13

2.2. Model of Frequency Control

Load

P 4

Shedding

Turbine Control of

P 3

Thermal Power

Turbine Control of

P 2

Hydro Power

Frequency

P 1

Dependency of Loads

System

P tot

Model

Disturbance

Figure 2.3. Block diagram of the frequency control in a power system.

the control of the steam turbines, and

load shedding.
This is summarized in the block diagram in Figure 2.3. Under frequency
load shedding is a form of system protection.
In many systems, a rotating scheme for how load should be shed, if that
is necessary, is devised. Such a scheme is often called rotating load shedding.
The frequency dependency of loads will be discussed in the next section,
while the frequency control performed by generating units will be discussed
in subsequent two chapters.
To give the reader a pre-view of how a typical frequency response would
be in systems with different frequency controls, Figure 2.4 is included. We
will revert to this figure when the models in the next chapter has been
developed.

2.2.1

Frequency Dependency of Loads

In an electric power system, the power demand of the load varies with the
frequency and the voltage. These variations are, as a rule, highly complicated and change during the day and over the year. For compound loads,
such as a transformer station in the highvoltage net, the compound structure of the separate loads is difficult to represent in a simple way. Generally,

14

2. Frequency Control in Electric Power Systems

more or less simplified models have to be used. For the phenomenon studied
here, the main influence is the frequency dependency of the loads, which,
for small frequency variations, can be written as
Pload = Pload0 (1 + Dl ) = Pload0 + D .

(2.15)

The parameter Dl changes of course, depending on the load type and composition, but typical values are in the interval 02% per % frequency variation.
Since Dl is positive, the frequency dependency of the loads leads to natural
stabilization of the frequency in the system. If that is the only stabilizing
control in the system, an unbalance between generation and load power of
P leads to a value of = P/D for the remaining frequency deviation
in the system.

15

2.2. Model of Frequency Control

Pm - Pload = P

0
No Frequency Control
P
D

P
R+D

Thermal Power without

Intermediate Re-Heater

P
R+D

Thermal Power with

Intermediate Re-Heater

P
R+D

Hydro Power

10

20

30

40

t(s)

Figure 2.4. Frequency variation after a step in Pm Pload . That

step can, for example, be caused by loss of production in the system.
D = frequency dependency of loads (see section 2.2.1). R = frequency
control droop (see chapter 3)

16

2. Frequency Control in Electric Power Systems

3
Primary Frequency Control
In this chapter the part of the frequency control that is performed on the
power plant level will be described. This is called the primary frequency
control. The static, i.e. steady state, characteristics of this is described by
the speed (or frequency) droop, which determines the permanent frequency
deviation after the primary frequency control has acted. Turbine models are
also derived.

3.1

Introduction

A schematic block diagram of the power and frequency control of a power

plant is given in Figure 3.1. The control loop consists here of the turbine,
generator, and the network.
The primary control indicates here the control actions that are done locally based on the set values for frequency (normally the nominal frequency)
and the power. The actual values of these could be measured locally, and
deviations from the set values will results in an signal that will influence the
valves, gates, servos, etc, so that the desired active power output is delivered
from the generating unit. In the primary frequency control the control task
of priority is to bring the frequency back to (short term) acceptable values,
and this control task is shared by all generators participating in the primary
frequency control irrespective of what caused the control action.
In the secondary frequency control, also called load frequency control, the
set values of the generator power is adjusted to compensate for undesired
fluctuations introduced by the primary frequency control. These undesired
fluctuations could be too large a frequency deviation during steady state
or power flows on tie lines outside the scheduled values. In this control
loop the cause of the control error is considered when the control action is
determined. This control could be done automatically, and is then often
called Automatic Generation Control, or it could be done manually. This is
further discussed in chapter 4.
The tertiary control loop in Figure 3.1 concerns the unit commitment
and economic dispatch problems and is done off line based on economic
optimization. This will not be elaborated on in these lectures.
17

18

3. Primary Frequency Control

Network

P, f

R
Primary control

Power Control

Secondary control

Energy Control
Tertiary control

Figure 3.1. Schematic block diagram of the power and frequency control of a power plant. R = Controller. T = Turbine. G = Generator.

For a thermal unit the block diagram of the primary control is shown in
Figure 3.2. The controller controls by use of a servomotor the valve through
which the high pressure and high temperature steam flows from the boiler to
the turbines. In the high pressure turbine part of the energy of the steam is
converted into mechanical energy. Often the steam is then reheated before
it is injected into a medium pressure or low pressure turbine, where more energy is extracted from the steam. (More about different steam turbines and
their modelling can be found in subsection 3.2.1.) These turbine-generator
systems can be very large. In a big thermal unit of rating 1000 MW, the
total length of the turbine-generator shaft could be more than 50 m.
We will now make some simplified studies concerning the dynamics of
the frequency response of the system in Figure 3.2. If the total inertia
constant of the turbine-generator system is H, the mechanical power from
the turbine(s) is Pm , the electrical power out from the generator is Pe , then
the equation of motion will be
2H d
= Pm Pe
0 dt

(3.1)

where both Pm and Pe (normally) are positive for a generator. If the

frequency controller is implemented as a proportional controller, i.e. as
Pm = K, the block diagram of Figure 3.3 is obtained. This model
can now be used to study the frequency response. The relationship between

19

3.1. Introduction
Reheater

Boiler

Valve

HP

P, f
Measured values

LP

Controller

Servomotor

Figure 3.2. Block diagram of the primary control of a thermal unit.

HP = High Pressure Turbine. LP = Low Pressure turbine.
Pe


Pm

Pm0

1
s 2 H / Z0




'Pm

'Z

Zref

Figure 3.3. Block diagram of simple system with proportional freZ 

quency controller.

Z
power and torque
is given by:

T =

(3.2)

If only small frequency deviations from the nominal frequency are studied,
in eq. (3.2) can be put equal to 0 and a linear relation between power
and torque can be assumed. This motivates the use of eq. (3.1), which is a
simplification of the more exact eqs. (2.1) and (2.2).
Two cases will be studied:
1. Change of the reference value of the frequency by a step.
2. Change of the active power load Pe with a step.

20

3. Primary Frequency Control

1.5

Delta omega

0.5

10

15

20

25

Time (s)

Figure 3.4. Step response in the frequency after a step change, ref ,
in the reference value of the frequency. (In per unit of ref ) ( = 5
s)

It is straightforward to derive the following relationship for case 1

1
1
=
ref
s 1 + s

(3.3)

and the following for the second case

1 1/K

=
Pe
s 1 + s

(3.4)

The time constant is given by

=

2H
0 K

(3.5)

These step responses are shown in Figures 3.4 and 3.5.

It is common to express the gain K of the controller in per unit:
K=

Pm /PN
/0

(3.6)

where the value of K often is in the range

0.03 < 1/K < 0.1

(3.7)

This means that a change in the frequency of 3 to 10 % will result in a

change in the delivered electrical torque of 100 %. This is called the static
speed droop of the controller, see Figure 3.6. For small deviations of the
frequency, the per unit values of torque and power are almost identical.

21

3.1. Introduction

0.5

Delta omega

0.5
X

1.5

10

15

20

25

Time (s)

Figure 3.5. Step response in the frequency after a step change, Pe

in the active power load.(X = Pe /K, T = 5 s)

'

'Z

Zp.u
ZN

P (p.u.)
PN

Figure 3.6. The speed droop of a frequency controller.

22

3.2

3. Primary Frequency Control

Turbine Control

This section gives an overview of the modelling of turbines, steam and hydro,
and their controllers. Their characteristics and behaviour are also briefly
discussed. The aim here is to give an understanding of the basic physical
mechanisms behind these models that are very commonly used in simulation
packages for the study of power systems dynamics. Figure 3.7 shows a
block diagram how these turbine models are integrated in the overall system
models.
Tie line power
AGC

Electrical system
- Loads
- Transmission lines
- Other generators

Frequency
Energy supply system:
Steam or water

Speed
changer

Speed
governor

Valves
or gates
Turbine
controls

Turbine

Generator

Speed
Other
signals

Figure 3.7. Functional block diagram of power generation and control

system. AGC = Automatic Generation Control, see chapter 4 (From
ref. [1])

3.2.1

Turbine Models

Steam Turbines
Figures 3.8 ,3.9, and 3.10 show the most common steam turbines and their
models.
It is outside the scope of these lecture notes to give a detailed derivation
and motivation of these models, only a brief qualitative discussion will be
provided. In a steam turbine the stored energy of high temperature and
high pressure steam is converted into mechanical (rotating) energy, which
then is converted into electrical energy in the generator. The original source

3.2. Turbine Control

23

of heat can be a furnace fired by fossil fuel (coal, gas, or oil) or biomass, or
it ca be a nuclear reactor.
The turbine can be either tandem compound or cross compound. In
a tandem compound unit all sections are on the same shaft with a single
generator, while a cross compound unit consists of two shafts each connected
to a generator. The cross compound unit is operated as one unit with one
set of controls. Most modern units are of tandem compound type, even if
the crossover compound units are more efficient and has higher capacity.
However, the costs are higher and could seldom be motivated.
The power outlet from the turbine is controlled through the position of
the Control Valves, which control the flow of steam to the turbines. The
delay between the different parts of the steam path is usually modelled by
a first order filter as seen in Figures 3.8 ,3.9, and 3.10. Certain fractions of
the total power is extracted in the different turbines, and this is modelled
by the factors FV HP , FHP , , FIP , FLP in the models. Typical values of
the time constant of the delay between the control valves and the highpressure turbine, TCH , is 0.10.4 s. If a re-heater is installed the time delay
is larger, typically TRH = 4 11 s. The time constant of delay between the
intermediate pressure and the low pressure turbines, TCO , is in the order
0.30.6 s.

24

3. Primary Frequency Control

Nonreheat

Valve
position

Control
Valves
Steam
Chest

Shaft
HP
To Condenser

Linear Model

PGV

1
1 + sTCH

PM

Tandem Compund, Single Reheat

Crossover

Reheater

Valve
position

Control
Valves
Steam
Chest

HP

IP

LP

LP

Shaft

To Condenser

Linear Model

FHP

PGV

1
1 + sTCH

1
1 + sTRH

FIP

FLP

PM

1
1 + sTCO

Figure 3.8. Steam turbin configurations and approximate linear models. Nonreheat and tandem compund, single reheat configurations.

25

3.2. Turbine Control

Tandem Compund, Double Reheat

Reheater
Control
Valves
Steam
Chest

Valve
position

VHP

Crossover

Reheater

HP

IP

LP

LP

Shaft

To Condenser
Linear Model

FVHP

FHP

FIP

FHP

PGV
1
1 + sTCH

1
1 + sTRH 1

1
1 + sTRH 2

1
1 + sTCO

Figure 3.9. Steam turbin configurations and approximate linear models. Tandem compound, double reheat configuration.

PM

26

3. Primary Frequency Control

Cross Compund, Single Reheat

Control
Valves
Steam
Chest

Valve
position

HP, LP Shaft
LP

HP

LP

Crossover

Reheater

IP, LP Shaft
IP

LP

LP

Linear Model

FLP

FHP

PGV

1
1 + sTCH

1
1 + sTRH

PM 1

1
1 + sTCO

FIP

FLP

PM 2

Figure 3.10. Steam turbin configurations and approximate linear

models. Cross compund, single reheat configuration.

27

3.2. Turbine Control

Step Response To illustrate the dynamics of a steam turbine, the configuration with tandem compound, single reheat, Figure 3.8, with the following
data will be studied:
TCH = 0.1 s, TRH = 10 s, TCO = 0.3 s
FHP = 0.3, FIP = 0.4, FLP = 0.3
As TCH TRH und TCO TRH , we can in an approximate analysis put
TCH = TCO = 0. Then a simplified block diagram according to Figure 3.11
can be used. For this system the step response is easy to calculate, and is
according to Figure 3.12.

0.3

PM

0.7

1
1 + s10

PGV

Figure 3.11. Simplified model of tandem compound, single reheat

system in Figure 3.9.

0.9

0.8

0.7
Power from turbine

FIP + FLP = 0.7

0.6

0.5
TRH = 10 s
0.4

0.3

0.2
FHP = 0.3
0.1

10

15

20

25
Time (s)

30

35

40

45

50

Figure 3.12. Step response of system in Figure 3.11.

28

3. Primary Frequency Control

P1
Length of
Penstock

= L
h = Head

Area = A
Velocity = v
P2

Effective Area= a
Output Velocity = v out
Figure 3.13. Schematic drawing of hydro turbine with water paths.

Hydro Turbines
Compared with steam turbines, hydro turbines are easier and cheaper to
control. Thus, frequency control is primarily done in the hydro power plants
if available. If the amount of hydrogenerated power in a system is not
sufficient, the steam turbines have to be included in the frequency control.
The power produced by a generator is determined by the turbine governor and the dynamic properties of the turbine. Thus, to be able to determine
the frequencys dynamic behaviour, models for the turbine as well as for the
turbine control are necessary.
Figure 3.13 depicts a hydro turbine with penstock and hydro reservoir
and defines the notation that will be used from now on. Bernoullis equation
for a trajectory between the points P1 and P2 can be written as
Z

P2
P1

v
1
dr + (v22 v12 ) + 2 1 +
t
2

The following assumptions are usually made:

P2

P1

1
dp = 0 .

(3.8)

v1 = 0, since the reservoir is large and the water level does not change
during the time scale that is of interest here.
The water velocity is nonzero only in the penstock.

29

3.2. Turbine Control

The water is incompressible, i.e. does not change with water pressure.
The water pressure is the same at P1 and P2 , i.e. p1 = p2 .
Further,
2 1 = gh .

(3.9)

The above assumptions together with eq. (3.9) make it possible to write
(3.8), with vout = v2 and the length of the penstock L, as
L

dv 1 2
+ vout gh = 0 .
dt
2

(3.10)

The velocity of the water in the penstock is v. The effective opening of the
penstock, determined by the opening of the turbines control valve (guide
vanes), is denoted a. If the penstocks area is A,
vout =

A
v
a

(3.11)

is valid and eq. (3.10) can be written as

 
1
1 A 2
dv
.
= gh
v
dt
L
2L a

(3.12)

The maximum available power at the turbine is

1
1 A3 v 3
3
P = avout
= 2 .
2
2 a
To get the system into standard form,

x = v ,

a
u =
,
A

y = P ,

(3.13)

(3.14)

are introduced. (Here, we have used the standard notation, i.e. x for state, u
for control signal, and y for output signal.) The system now can be written
as

1
gh

x2
,
x =

L
2Lu2
(3.15)

y = A
.
2u2
The system corresponding to eq. (3.15) can be described with the block
diagram in Figure 3.14.

30

3. Primary Frequency Control

x

u
*/*
xu

x u

1
-----2L

dx
dt

x
y = P = A --------2
2u
1
--s

gh
-----L
A
------2

Figure 3.14. Block diagram showing model of hydro turbine.

Eq. (3.15) is nonlinear and a detailed analysis is beyond the scope of

these lecture notes. To get an idea of the systems properties, the equations
are linearised, and small variations around an operating point are studied.
In steady state, x = 0, and the state is determined by x0 , u0 , and y0 , which
fulfils
p

x0 = u0 2gh ,

(3.16)

Ax30

y0 =
.
2u20
Small deviations x, u, and y around the operating point satisfy

1
2x20

x = 2x0
x
+
u ,

2Lu20
2Lu30

(3.17)

3
2

Ax0
Ax0

y = 3 2 x 2 3 u ,
2u0
2u0
which, using eqs. (3.16), can be written as

2gh
2gh

x =
x +
u ,

u0 L
u0 L

y =

(3.18)

3y0
2y0

u .
x
u0
u0 2gh

The quantity L/ 2gh has dimension of time, and from the above equations
it is apparent that this is the time it takes the water to flow through the

31

3.2. Turbine Control

penstock if a = A. That time is denoted T :
p
T = L/ 2gh .

(3.19)

If eqs. (3.18) are Laplacetransformed, x can be solved from the first

of the equations, leading to
x =

L/T
u ,
1 + su0 T

(3.20)

which, when inserted in the lower of eqs. (3.18) gives

y =

y0 1 2u0 T s

u .
u0 1 + u 0 T s

(3.21)

The quantity u0 T = a0 T /A also has dimension of time and is denoted Tw .

The time constant Tw is the time it takes for the water to flow through the
penstock when a = a0 or u = u0 , i.e. for the operation point where the
linearization is done. Eq. (3.21) can thus be written as
y =

y0 1 2Tw s

u .
u 0 1 + Tw s

(3.22)

It is evident that the transfer function in eq. (3.22) is of nonminimum

phase, i.e. not all poles and zeros are in the left half plane. In this case, one
zero is in the right half plane. That is evident from the step response to
eq. (3.22), depicted in Figure 3.15.
The system has the peculiar property to give a lower power just after
the opening of the control valve is increased before the desired increased
power generation is reached. The physical explanation is the lower pressure
appearing after the control valve is opened, so that the water in the penstock
can be accelerated. When the water has been accelerated, the generated
power is increased as a consequence of the increased flow. That property of
water turbines places certain demands on the design of the control system
for the turbines.

3.2.2

Turbine Governors

It is the task of the turbine governor to control the control valve such that
the desired power is produced by the generator in question. That power is
partly determined by the set value for the produced power and partly by
a contribution originating from the frequency control. In this context, the
latter is of interest.
There is a number of different turbine governors for steam turbines. A
quite general model that can describe most controllers in a satisfying way
is depicted in Figure 3.16. The gain K determines here the speed droop
characteristics of the controller, i.e. S = 1/K.

32

3. Primary Frequency Control

u
u 1

t(s)
0

10

15

20

y
y0
----- u 1
u0

t(s)

y0
- 2 ----- u 1
u0

Tw = 5 s

Figure 3.15. The variation of the produced power, y, after a step

change in the control valve.

A model of controller of a hydro turbine is given in Figure 3.17. The

control servo is here represented simply by a time constant Tp . The main
servo, i.e. the guide vane, is represented by an integrator with the time
constant TG . Typical values for these parameters are given in Table 3.1.
Limits for opening and closing speed as well as for the largest and smallest
opening of the control valve are given. The controller has two feedback loops,
a transient feedback loop and a static feedback loop. The transient feedback
loop has the amplification for high frequencies. Thus, the total feedback
after a frequency change is ( + ). In steady state, the transient feedback
is zero, and the ratio between the frequency deviation and the change in the

Parameter
TR
TG
Tp

Typical Values
2.5 7.5 s
0.2 0.4 s
0.03 0.06 s
0.2 1
0.03 0.06

Table 3.1. Typical values for some parameters of the turbine governor
for hydro power.

33

3.3. Role of Speed Droop

P up

P max

P0

( 1 + sT 2 )
K ----------------------1 + sT 1

1
-----T3

P down

P GV

1
--s

P min

Figure 3.16. General model of turbine governor for steam turbine.

control valve is given by

u =

1
.

(3.23)

Using eq. (3.22), the stationary change of power is obtained as

P =

1 P0
.
u0

(3.24)

Thus, the speed droop for generator i, Si , is

Si = i

u0i
,
P0i

(3.25)

and the total speed droop, S, in the system is given by

S 1 =

Si1 .

(3.26)

The transient feedback is needed since the water turbine is a nonminimum

phase system as discussed above. If the transient feedback is left out or
made too small, the system can become unstable. The transient feedback
causes the system to be slower; the transient frequency deviations become
considerably larger since the initial total feedback can be about ten times
larger than the static feedback, i.e. the speed droop is initially lower than
its stationary value.

3.3

Role of Speed Droop

It is of particular interest to study the actions of the frequency control in

steady state. Under the assumption that the turbine power controller has

34

3. Primary Frequency Control

u open
+

1
-----TG

1
----------------1 + Tps

u max
1--s

u +

u close

u min
u0

T R s
----------------1 + TRs

u = gate opening
Figure 3.17. Model of turbine governor for hydro turbine.

an integrating character ( 0 when t in Figure 3.18), it follows that

in steady state
1
(f0 f ) + (P0 P ) = 0(= )
(3.27)
S
which can be written as
S=
or in per unit

f f0
f0 f
=
Hz/MW
P0 P
P P0
f f0
f0
S=
P P0
P0

(3.28)

(3.29)

(Eg.(3.29) is equivalent with eq. (3.6) with S = 1/K.)

The speed droop characteristic, Figure 3.19, gives all possible operating
points (P, f ) of the turbine. The position and slope of the straight line
can be fixed by the parameters P0 , f0 and S. We have chosen to label the
horisontal axis with the power P with for small deviations of the frequency
around the nominal value is identical with the torque T . In the literature
the speed droop characteristics is sometime also described by instead of
by f .

35

3.3. Role of Speed Droop

P0

- e

Controller

Turbine

P
f

1
S

f0

Figure 3.18. Schematic block diagram of system of controller, turbine, generator, and power system.

f0

P0
f

und

sind Einstellwerte

Figure 3.19. Speed droop characteristic of turbine control.

f0

fN

P0
f

PN

36

f0 , P0

3. Primary Frequency Control

Pm
R

XL

Pe

Infinite Bus

f=fN=const.
U=UN=const.

f,P

Figure 3.20. Generator operating in a large interconnected system.

We will now study how the frequency control of a generator will act in
three different situations. Firstly, when the generator is part of a large interconnected system, and secondly when the generator is in islanded operation
feeding a load. The third system to be studied is a two machine system.
Generator in large system If a generator is embedded in a large interconnected system, it can with a very good approximation be modelled as
connected to an infinite bus as shown in Figure 3.20
In steady state the frequency is given by the one of the infinite bus, fN .
From the speed droop characteristics, Figure 3.19, the power produced by
the generator can then be determined. The turbine controller controls thus
only the power, not the frequency, see Figure 3.21.

f
f0 and P0 are nominal values

f0
fN

S
P=g(fN)
P0

PN

Figure 3.21. Speed droop characteristics for the case when the generator is connected to an infinite bus (large system).

37

3.3. Role of Speed Droop

Voltage Control
f0 , P0

Pm

Pe

f,P

Pe=

U2
= const.
R

Figure 3.22. Generator in islanded operation.

Islanded Operation In islanded operation the generator feeds a load, which

here is assumed to be a resistance load, Figure 3.22. By a voltage controller
the voltage U is kept constant and thus also P . In this case the turbine controller will control the frequency, not the power. The resulting frequency can
also here be determined from the speed droop characteristics, Figure 3.23.

f
f0 and P0 are nominal values

f0
f

S
f=h(P)
P0

Figure 3.23. Speed droop characteristics for the case when the generator is in islanded operation.

38

3. Primary Frequency Control

P0 2, f0 2, S2

P01, f01, S1
P2

P1
G1

G2
D P2

D P1

P=P1+P2 at fN

Figure 3.24. Two generator (subsystem) system.

Two Generator System The two generator system, Figure 3.24, provides a
simple model that is often used to study the interaction between two areas
in a large system. In this model the two generators could represent two
subsystems, and the speed droop is then the sum of all the individual speed
droops of the generators in the two subsystems, Figure 3.25. With the help
of the speed droop characteristics of the two systems, we will determine how
a change in load will compensated by the two systems. Thus, if we have a
change P in P , what will the changes in P1 , P2 , and f be?
This will be solved in the following way:
The quantities (P01 , f01 , S1 ) and (P02 , f02 , S2 ) describe the speed droop
characteristics of the two systems g1 and g2 .
From these the sum g3 = g1 + g2 is formed.
From the given P we can determine P1 , P2 and fN from g3 .
In a similar way: From P + P can P1 + P1 , P2 + P2 and fN + f
be determined, and thus P1 , P2 und f .
All these steps are shown in Figure 3.25.

39

2
2

P1 + P1

P2
P1
f N + f

fN

f 01

f 02

S1 , g1

P01

P02

S2 , g 2

S3 , g 3

P2 + P2

P + P

3.3. Role of Speed Droop

Figure 3.25. Speed droop characteristics for a two machine (subsystem) system.

40

3. Primary Frequency Control

4
Load Frequency Control
In this chapter the secondary, or load-frequency, control of power systems
will be discussed. Simple models that enable the simulation of the dynamic
behaviour during the action of frequency controllers will also be derived and
studied.

In the previous chapters the role of the primary frequency control was dealt
with. It was shown that after a disturbance a static frequency error will
persist unless additional control actions are taken. Furthermore, the primary
frequency control might also change the scheduled interchanges between
different areas in an interconnected system, see the Two Machine example in
section 3.3. To restore the frequency and the scheduled power interchanges
additional control actions must be taken. This is done through the LoadFrequency Control (LFC). The LFC can be done either manually through
operator interaction or automatically, in which latter case it is often called
Automatic Generation Control (AGC). The characteristics of AGC will be
studied in the subsequent sections, both during steady state and dynamic
conditions.

4.1

Automatic Generation Control - Static Model

The overall goal of the LFC is to

Keep the frequency in the interconnected power system close to the
nominal value.1
Restore the scheduled interchanges between different areas, e.g. countries, in an interconnected system.
How this is achieved by the use of AGC will now be demonstrated.
Consider a two area system as depicted in Figure 4.1. The two controllers, R1 and R2 , will send new reference values of the power to the
1
In many systems deviations of up to 0.1 Hz from the nominal value ( 50 or 60 Hz)
is deemed as acceptable in steady state. In some systems, North America, even tighter
tolerance bands are applied, while recently in UK the tolerance band has been relaxed
somewhat.

41

42

4. Load Frequency Control

Area 1
G

Area 2

PT 2

PT 1

G
G

f1

f2

R1

PT 1

R2

Tie-line power for Area 1 =

j
T1

j:1

Sum over all tie-lines

Figure 4.1. Two area system with AGC.

generators participating in the AGC. In an N area system there are N con' diagram of such a controller
trollers Ri , one for each area i. A block
is given
'
in Figure 4.2. A commonway is to implement this as a PI-controller:
Pci = (Cpi +

1
)ei
sTN i

(4.1)

where Cpi = 0.1 . . . 1.0 and TN i = 30 . . . 200 s. The error ei is called

Area Control Error, ACEi for area i.
We will now consider a system with N areas. The ACEs are in this case:
X j f
(4.2)
ACEi =
(PT i,ref PTj i ) + Bi (fref f ) i = 1, 2, . . . , N
ji

ACEi = PT i + Bi f

i = 1, 2, 3, 4

(4.3)

The set i consist of all areas connected to area i for which the tie-line
powers should be controlled to the set value PTj i,ref The constants Bi are
called frequency bias factors [MW/Hz]. It is assumed that the frequency
references are the same in all areas, i.e. fref,i = fref for all i, and f is also
the same in steady state forX
all areas. The goal is to bring all ACEi 0.
The variables are thus
PTj i (N variables) and f , i.e. in total N + 1
ji

variables. Since we have N equations ( ACEi = 0), we need one more

43

4.1. Automatic Generation Control - Static Model

PTi ,ref

PTi

'ei

Ri

'Pc ,i

Bi


f ref

Figure 4.2. Controller for AGC (ei = Error = ACEi =

Area Control Error for area i)

equation. As fifth equation we have the power balance:

X X j
(
PT i ) = 0
i

(4.4)

ji

and consequently a solution can be achieved.

In steady state f is identical for all areas, and we assume that the frequency is controlled back to the reference value, i.e. fref = f . If the sum of
the reference values of the tie line powers PTj i,ref is 0, then the system will
settle down to an operating point where PTj i,ref = PTj i for all tie line powers.
The time constants of the AGC is chosen so that it reacts much slower than
the primary frequency control.
Selection of Frequency Bias Factors
Consider the two area system in Figure 4.1. The load is now increased with
Pl in area 2. If the tie line power should be kept the same, the generation
must be increased in area 2 with Pl aftre the AGC has reacted. Before the
AGC has reacted we have a frequency deviation of f in both areas, which
means that for area 1
f = S1 PT 1
(4.5)
and for area 2

f = S2 (Pl PT 1 )

(4.6)

44

4. Load Frequency Control

and PT 1 = PT 2 . The two ACEs can now be written as

ACE1 = PT 1 + B1 f = PT 1 + B1 (S1 PT 1 ) = PT 1 (1 B1 S1 ) (4.7)
and
ACE2 = PT 2 + B2 f = PT 2 + B2 (S2 (Pl PT 1 )) =

PT 2 (1 B2 S2 ) B2 S2 Pl

(4.8)

In this case it is desirable that the AGC controller in area 1 does not react.
If we set B1 = 1/S1 we see from eq. (4.7) that ACE1 = 0. This is called
Non Interactive Control. If B2 = 1/S2 is chosen the ACE in area 2 becomes
ACE2 = Pl

(4.9)

This means that only controller 2 reacts and the load increase Pl is compensated for in area 2, cf eq. (4.1).
However, as long as the controller in eq. (4.1) has an integrating part all
positive values of Bi will guarantee that all ACEi 0. The choice according
to Non Interactive Control is through a number of investigations found to
give the best dynamic performance. In a multi-area case this corresponds
to selecting Bi = 1/Si for all areas..

4.2

Dynamic Model

Up to now we have mostly studied the static part of the frequency control,
i.e. what can be concluded from the speed droop characteristics of the
system. In this section we will study the dynamic behaviour during the
action of the frequency controllers. Two cases will be analyzed. The first
case corresponds to a system where all the generators and loads are strongly
coupled to each other, which is the case in a highly meshed system. In this
case we can model the system by a one node system, where all the generators
and loads are connected at the same node. Secondly, we will study a two
area system with a tie line between the two areas. This latter case will be
studied with and without AGC implemented.

4.2.1

One Node System

Assume there are n generators connected in the system, which is modelled

as one node. For the individual generator i the swing equation applies:
2Hi SBi

d(i /0 )
= Pti Pei ,
dt

i = 1, . . . , n

(4.10)

(The normally used version of the swing equation, where the quantities are
in p.u., have here been multiplied by the power base SBi so the quantities are

45

4.2. Dynamic Model

expressed in S.I. units.) Assuming that the generators are strongly coupled
(i = ), gives
X
d(/0 ) X
=
(Pti Pei )
2(
Hi SBi )
dt

(4.11)

2HSB
where
SB =

d(/0 )
= Pt Pe
dt

(4.12)

SBi

(4.13)

Total rating

P
i Hi SBi
H= P
i SBi
X
Pt =
Pti

Total inertia constant

(4.14)

Total turbine power

(4.15)

Total generator power

(4.16)

Pe =

Pei

from /0 = (2f )/(2f0 ) = f /f0 it follows that

2HSB df
= Pt Pe
f0 dt

(4.17)

As these differential equations are linear, they are also valid for the quantities
Pt = Pt Pt0

Pe = Pe Pe0
f

and consequently

(4.18)

= f f0

(4.19)

2HSB df
= Pt Pe
f0
dt

(4.20)

A block diagram of eq. (4.20) is given in Figure 4.3.

Turbine and Controller Consider the block diagram in Figure 4.4. From
this figure it follows that
Pt (s) =

Gt (s)
1
Gt (s) + s
K

[Pt0

1
(f f0 )]
S

(4.21)

If the dynamics of the turbine is neglected (Gt (s) = 1), one obtains
Pt =

1
1
[Pt0 (f f0 )]
1 + sT
S

T =

1
K

(4.22)

46

4. Load Frequency Control

'Pe

'f

(2 HS B / f 0 ) s


'Pt

Figure 4.3. Block diagram for the dynamics of n generators connected

at one node.
f0

Droop

1
S
Controller

Pt0

K
s

Turbine
Pt

Gt(s)

K
s

Figure 4.4. Block diagram of the dynamics for turbine and turbine control.

This equations is also valid for the quantities:

'f
1
1
Pt =
[Pt0 f ]

1 + sT
S

and the block diagram in Figure 4.5is valid.
In steady state we have
'
1
(Pt0 = 0)
Pt = f
S

(4.23)
'

(4.24)

and for the case of several controllers

Pti =

'f

X
i

1
f
Si

Pti =

i = 1, . . . , n
X 1
f
S
si i

(4.25)
'

(4.26)

where
Pt =

X
i

Pti

Total change in the turbine power

(4.27)


47

4.2. Dynamic Model

'f

1
S

1
1  sT

'Pt

'Pt 0

Figure 4.5. Block diagram of the system when the dynamics of the
turbine is neglected.

By defining
'f


X 1
1
=
S
Si
i

we thus have

'

(4.28)

1
Pt = f
S

(4.29)

Loads The loads are either frequency dependent or frequency independent.

Furthermore, kinetic energy can be stored (rotating masses in motors). A
load model that captures this is given by
Pl Pl0 = Pl = Kl f + g(f)

(4.30)

where
Pl0 : Load power when f = f0
Kl : Frequency dependency
g(f): Function the models the load with rotating masses
The function g(f) will now be derived. The rotating masses have the
following kinetic energy:
1
W (f ) = J(2f )2
2

(4.31)

The change in the kinetic energy is given by

PM =
and
PM =

dW
dt

(4.32)

dW
dt

(4.33)


'
48

4. Load Frequency Control

1
Dl
'f
2W0
s
f0

'Pl

Figure 4.6. Block diagram of the dynamic load model.

W can be approximated by
W (f0 + f ) = 2 2 J(f0 + f )2 =
W0 + W

W
PM

= 2 2 Jf02 + 2 2 J2f0 f + 2 2 J(f )2

2W0
W0
= W0 +
f + 2 (f )2
f0
f0
2W0
f

f0
2W0 df
2W0

=
f
f0 dt
f0

(4.34)

The frequency dependency is given by

1
Pl
f =
f = Kl f
f
Dl

(4.35)

The block diagram in Figure 4.6 describes the load model.

Total System The system in Figure 4.7 can now be described by the block
diagram in Figure 4.8.

49

4.2. Dynamic Model

Pe  ' Pe

Pt  ' P t

Pt 0  'Pt 0
R

Pl  ' Pl

'P

Load

Disturbance

Figure 4.7. System with all generators and loads connected to one node.

 '

'

 '

 '
f

 '

'

50

4. Load Frequency Control

'P

1
Dl

2W0
s
f0

'Pl

'Pe




1
(2 HS B / f 0 ) s

'Pt

1
S




1
1  sT

'Pt 0

Figure 4.8. Block diagram of the system in Figure 4.7.

'f

51

4.2. Dynamic Model

Dynamic Behaviour From the block diagram in Figure 4.8 it is straightforward to derive the transfer function between P and f (Pt0 = 0):
f =

1 + sT
P
1
1
2W0 2HSB
+
(1 + sT ) + (
+
)s(1 + sT )
S Dl
f0
f0

(4.36)

The step response for

P (s) =

z
s

(4.37)

is given in Figure 4.9. The frequency deviation in steady steady state is

f = lim (s f ) =
s0

z
z
=
= z DR
1
1
1
+
S Dl
DR

(4.38)

with
1
1
1
= +
DR
S Dl

(4.39)

In order to calculate an equivalent time constant, Tf , T is put to 0. This

can be done since for realistic systems
T TM =

f0 2W0 2HSB
(
+
)
SB f0
f0

(4.40)

This means that the transfer function in eq. (4.36) can be approximated by
a first order function
f =

1
P
=
DR P
SB
1
SB
+ TM
s
s
1 + TM D R
DR
f0
f0

(4.41)

or
1

f =

1 + TM D R

SB
s
f0

(4.42)

with
Tf = T M D R

SB
f0

(4.43)

as the equivalent time constant.

TM =

f0 2W0 2HSB
(
+
)
SB f0
f0

(4.44)

52

4. Load Frequency Control

1.5

Power

0.5

0.5

5
Time

10

5
Time

10

Frequency deviation

0.05

0.05

Figure 4.9. The step response for a one node system. The upper diagram shows the step function, the increase in turbine power (solid), and
the load variation (dashed). The lower diagram shows the frequency
deviation.

Example
SB = 4000 MW
f0 = 50 Hz
S = 4% = 0.04 SfB0 =
Dl =

50
4000

0.0450
4000

Hz/MW

Hz/MW

z = 400 MW
TM = 10 s
Then follows
f =

1
400 = 0.192 Hz
4000 4000
+
2
50

(4.45)

53

4.2. Dynamic Model

Area 1

Area 2

PT

S , Dl ,W0

S , Dl ,W0
TM , S B , f 0

U1 ,M1

U 2 ,M 2

TM , S B , f 0

f1
f2

f0

Figure 4.10. Model of a two area system. System 1 is much smaller

than system 2.

and
Tf = 10

4.2.2

4000
1
= 0.38s
4000 4000 50
+
2
50

(4.46)

Two Area System

We will now study the behaviour in a two area system. Each area could be
modelled as in the previous subsection. The two areas are connected with
a tie line over which power can be exchanged.
We will first consider the case without AGC. It is further assumed that
one of the areas is much smaller than the other. The bigger of the two areas
can then be regarded as an infinite bus in our analysis. It will now studied
the behaviour after a load change in the smaller system, system 1. The
system to be studied, with notation, is depicted in Figure 4.10. dargestellt.
The tie line power is given by
PT =

U1 U2
sin(1 2 )
X

(4.47)

where X is the (equivalent) reactance of the tie line. For small deviations
one gets (U1 and U2 are constant)
PT =
or
with

PT
PT
U 1 U2
1 +
2 =
cos(10 20 )(1 2 ) (4.48)
1
2
X
PT = PT (1 2 )

(4.49)

U1 U2
cos(10 20 )
PT =
X

(4.50)

54

4. Load Frequency Control

Tie line power flow

Pt  ' P t

Pt 0  'Pt 0
T

Pe  ' Pe

PT  ' PT

Pl  ' Pl

'P

Load

Disturbance

Figure 4.11. Power flow in Area 1 of the two area system.

As system 2 is very big (infinite bus) it follows that

(

TM SB
TM SB
)2 (
)1 f2 = constant 2 = 0
f0
f0

(4.51)

and consequently
PT = PT 1 = 2 PT

f1 dt

(4.52)

Figure 4.11 shows the power flow and Figure 4.12 the block diagram of the
system.
Without secondary frequency control (AGC), i.e. Pt0 = 0, the following transfer functions apply
f =

s
P
1
TM SB 2
1
+
)s +
s
2 PT + (
Dl S(1 + sT )
f0
PT =

PT =

2 PT
f
s

2 PT
P
1
1
TM SB 2

2 PT + (
+
)s +
s
Dl S(1 + sT )
f0

(4.53)

(4.54)
(4.55)

The response for

P (s) =

z
s

(4.56)

55

4.2. Dynamic Model

'P

1
Dl

'Pl

2W0
s
f0

'PT

2S PT
s

'Pe




'f

1
(2 HS B / f 0 ) s

'Pt

1
S

1
1  sT


'Pt 0

Figure 4.12. Block diagram of Area 1 of the two area system.

is shown if Figure 4.13. The steady state frequency deviation is

f = lim (s f ) = 0
s0

(4.57)

and the steady state deviation of the tie line power is

PT = lim (s PT ) = z

s0

(4.58)

The infinite bus brings the frequency deviation f back to zero, by increasing the tie line power so the load increase is fully compensated.
With secondary control (AGC) one obtains the step response in Figf load increase is in this case fully compensated by the generaure 4.14. The
tors in Area 1.

56

4. Load Frequency Control

1.5

Power

0.5

0.5

5
Time

10

5
Time

10

Frequency deviation

0.05

0.05

Figure 4.13. Step response for the system in Figure 4.11 without
AGC. The upper diagram shows the step in load, the turbine power
(solid) and the tie line power (dashed). The lower diagram shows the
frequency deviation.

57

4.2. Dynamic Model

1.5

Power

0.5

0.5

10

15
Time

20

25

30

10

15
Time

20

25

30

Frequency deviation

0.05

0.05

Figure 4.14. Step response for the system in Figure 4.11 with AGC.
The upper diagram shows the step in load, the turbine power (solid)
and the tie line power (dashed). The lower diagram shows the frequency deviation.

58

4. Load Frequency Control

5
Model of the Synchronous Machine
Almost all energy consumed by various loads in an electric power system is
produced by synchronous machines, or, more correctly, the conversion from
the primary energy sources, like water energy, nuclear energy, or chemical
energy, to electrical energy is done in synchronous machines with a mechanical intermediate link, the turbine. This is true in larger power systems, but
not always in smaller systems like isolated islands, power supply of equipment in deserts, or other smaller systems. In these systems, the energy
can come from asynchronous generators, for example in wind generation
units, batteries, or some other source of electrical energy. In systems with
synchronous generators, these have an extremely important part in many
dynamic phenomena. Thus, it is very important to develop usable and realistic models of the synchronous machines. In the previous chapters, mainly
the mechanical properties of the synchronous machines have been modelled
using the swing equation, while a very simplistic model of the electrical properties of the synchronous machine has been used. In this chapter, a more
general, detailed model of the electric parts of the synchronous machine will
be derived. The simple models used earlier will be justified. It should be
emphasized that the description here aims towards the development of models usable for studying dynamic phenomena in the power system. It is not
the purpose of these models to give a detailed and deep understanding of the
physical functions of the synchronous machine. Of course, it is desirable to
have a good insight into the physics of the synchronous machine to be able to
derive appropriate models. For a detailed discussion of these aspects, books
and courses dealing with the theory of electrical machines should be studied.

5.1

Parks Transformation

Parks transformation is a phase transformation (coordinate transformation)

between the three physical phases in a three phase system and three new
phases, or coordinates, that are convenient for the analysis of synchronous
machines. This transformation is also known as the dqtransformation or
Blondels transformation. A reason why the transformation is suitable can
be derived from Figure 5.1.
It is obvious that the phase quantities in the a, b, and cphases will
vary periodically in steady state. Further, the self and mutual inductances
between stator circuits and rotor circuits will vary with the rotor position.
Instead of performing all computations in the fixed stator system, the stator
59

60

5. Model of the Synchronous Machine

a-axis
d-axis

ib

fb n

sc

iQ
iD
Direction
of Rotation sa
iD
n

b-axis

iF

fc

iF

fa
ia

iQ
c-axis

ic

sb

q-axis
Figure 5.1. Definition of quantities in Parks transformation.

quantities voltages, currents, and fluxes can be transformed to a system that

rotates with the rotor. Thus, two orthogonal axes are defined as shown in
Figure 5.1: One along the axis in which the current in the rotor windings
generates a flux, and one in an axis perpendicular to this. The first is the
direct axis (daxis), and the other is the quadrature axis (qaxis). From now
on, the denominations daxis and qaxis will be used. To make the system
complete, a third component corresponding to the zero sequence must be
defined.
Figure 5.1 is a simplified picture of a synchronous machine and should
only be viewed as an intuitive basis for the transformation given below. The
machine in Figure 5.1 has one pole pair, but Parks transformation can, of
course, be applied to machines with an arbitrary number of pole pairs.
Parks transformation is, as a consequence of the reasoning above, time
dependent, and the connection between the phase currents and the trans-

61

5.1. Parks Transformation

formed

i0

id

i
q

currents is given by
r
1
=
(ia + ib + ic ) ,
3
r 




2
2
2
=
ia cos + ib cos
+ ic cos +
,
3
3
3
r 




2
2
2
=
ia sin ib sin
ic sin +
.
3
3
3

(5.1)

If the aaxis is chosen as reference,

= t + 0

(5.2)

is obtained and the time dependence in the transformation is obvious. It

should be pointed out that ia , ib , and ic are the real physical phase currents
as functions of time and not a phasor representation of those. Now,
xabc = (xa , xb , xc )T
x0dq = (x0 , xd , xq )T

(5.3)

can be defined. x can here be an arbitrary quantity, like voltage, current,

or flux. With this notation, Parks transformation can be written as

with

x0dq = P xabc

(5.4)

r 1/2
1/ 2
1/ 2
2
.
P =
cos
cos( 2
cos( + 2
3 )
3 )
3
2
2
sin sin( 3 ) sin( + 3 )

(5.5)

xabc = P 1 x0dq ,

(5.6)

The inverse transformation is then given by

and it can easily be shown that

P 1 = P T .

(5.7)

A mnemonic for Parks transformation can be obtained from Figure 5.1 by

projecting the a, b, and caxes onto the d and qaxes in the figure.
Equation (5.7) implies that Parks transformation is an orthonormal
transformation. This is reflected in the expression for the momentary power
that is produced in the stator windings
p = ua ia + ub ib + uc ic = uTabc iabc =
(P 1 u0dq )T P 1 i0dq = (P T u0dq )T P 1 i0dq =
uT0dq P P 1 i0dq = uT0dq i0dq =
u0 i0 + ud id + uq iq .

(5.8)

62

5. Model of the Synchronous Machine

Here, Equations (5.6) and (5.7) have been used. Equation (5.8) can therefore
be written as
p = ua ia + ub ib + uc ic = u0 i0 + ud id + uq iq .

(5.9)

Equation (5.9) shows that the introduced transformation is power invariant,

which is a consequence of Equation (5.7).
It should be pointed out that there are several different variants of Parks
transformation appearing in literature. They can differ from the form presented here by the direction of the qaxis and by constants in the transformation matrix. When using equations from some book or paper, it is thus
important to make sure that the definition of Parks transformation used is
the same as ones own. Otherwise, wrong results might be obtained.
The rotor windings produce a flux linkage that mainly lies in the direction
of the daxis. That flux induces an emf, E, which is lagging by 90 , hence
in the direction of the negative qaxis. For generator operation, the phasor
for E leads by an angle before the phasor for the terminal voltage U . At
t = 0, the negative qaxis thus leads by an angle before the phasor for the
voltage along the aaxis, cf. Figure 5.1. For t > 0, the d and qaxes have
moved by an angle t with the angular speed of the rotor . The rotors
daxis will hence be in position
= t + +

.
2

(5.10)

It is particularly of interest to study how zero sequence, negative sequence,

and positive sequence quantities are transformed by Parks transformation.
It is comparatively easy to show that a pure zero sequence quantity only
leads to a contribution in x0 with xd = xq = 0. A pure positive sequence
quantity

sin( + )

(5.11)
xabc (+) = 2x sin( + 2
3 )
2
sin( + + 3 )
is transformed to

x0dq (+) =

0
3x sin() ,
cos()

(5.12)

i.e. pure DC-quantities (time independent) in the dqsystem with the zero
sequence component equal zero. A pure negative sequence quantity gives
rise to quantities in d and qdirections that vary with the angular frequency
2. The zero sequence component vanishes also in this case. (Show this!)

63

5.2. The Inductances of the Synchronous Machine

5.2

The Inductances of the Synchronous Machine

In the following, a synchronous machine with one damper winding in d and

one in qaxis and, of course, a field winding is considered. Quantities related
to these windings are denoted with the indices D, Q, and F, respectively.
The flux linkages in the stator windings and in the windings F, D, and Q
depend on the currents in these windings according to


 

iabc
Labc,abc
Labc,F DQ
abc
.
(5.13)
=
iF DQ
LF DQ,abc LF DQ,F DQ
F DQ
Labc,abc , . . ., LF DQ,F DQ are 3 3 matrices with self and mutual inductances
as matrix elements. These matrices will depend on the rotor position, hence
they are time dependent. It can be shown that the inductances in these
matrices can be approximated by
Labc,abc :

Laa = Ls + Lm cos 2 ,

Lbb = Ls + Lm cos(2 4

3 ) ,

Lcc = Ls + Lm cos(2 + 4
3 ) ,
(5.14)
L
=
L
=
M

L
cos(2
+ 3 ) ,

s
m
ab
ba

Lbc = Lcb = Ms Lm cos(2 + ) ,

Lac = Lca = Ms Lm cos(2 + 5

3 ) .
Labc,F DQ and LF DQ,abc :

LF DQ,F DQ :

LaF = LF a = MF cos ,
LbF = LF b = MF cos(
LcF = LF c = MF cos( +

2
3 )
2
3 )

,
,

LaD = LDa = MD cos ,

L = LDb = MD cos( 2
3 ) ,
bD
2

L
=
L
=
M
cos(
+

cD
Dc
D

3 ) ,

LaQ = LQa = MQ sin ,

LbQ = LQb = MQ sin( 2

3 ) ,

L = L = M sin( + 2
cQ
Qc
Q
3 ) .

LF F = LF ,

DD = LD ,

LQQ = LQ ,

LF D = LDF = MR ,

L
= LQF = 0 ,

FQ
LDQ = LQD = 0 .

(5.15)

(5.16)

64

5. Model of the Synchronous Machine

All inductances with only one index in Equations (5.14)(5.16) are constants
and depend on the design of the synchronous machine. The resulting inductances are of course, as mentioned before, not quite exact. They can be
called exact in an ideal machine, where spatial harmonics and other unsymmetries are neglected. For a real synchronous machine, the approximations
are usually very good and lead to fully acceptable results for the computations and analyses treated here. It should be emphasized that the model
developed here is for use in computations where the synchronous machines
are part of a larger system. The model is not primarily aimed at studies of
the internal quantities in the generator.
It is now natural to transform the abccomponents in Equation (5.13)
to 0dqcomponents. For this, an extended transformation given by


P 0
,
(5.17)
Pex =
0 I
with P according to (5.5) and a 3 3 unit matrix I is used. The result is

 


0dq
L0dq,0dq
L0dq,F DQ
i0dq
=
,
(5.18)
F DQ
LF DQ,0dq LF DQ,F DQ
iF DQ
with the inductance matrix given by


L0dq,0dq
LF DQ,0dq

L0dq,F DQ
LF DQ,F DQ

P
0

0
I



Labc,abc
LF DQ,abc

Labc,F DQ
LF DQ,F DQ



P 1
0


0
.
I
(5.19)

The virtue of the Parks transformation is apparent in the following equation, where the inductance matrix in (5.18) is computed

L0 0
0
L0dq,0dq = 0 Ld 0 ,
(5.20)
0
0 Lq

with

and

L = Ls 2Ms ,

0
Ld = Ls + Ms + 32 Lm ,

Lq = Ls + Ms 23 Lm ,
L0dq,F DQ

where

0
= kMF
0
k=

and, of course,

0
0
kMD
0 ,
0
kMQ

3
2

LF DQ,0dq = LT0dq,F DQ .

(5.21)

(5.22)

(5.23)
(5.24)

5.3. Voltage Equations for the Synchronous Machine

65

d-axis
d

D
F
q-axis
q

Figure 5.2. Schematic picture of the transformed system. (One

damper winding in the daxis and one in the qaxis.)

LF DQ,F DQ has, of course, not changed, but for completeness it is repeated here.

LF MR 0
0 .
LF DQ,F DQ = MR LD
(5.25)
0
0 LQ

Two important observations can be made from Equations (5.20)(5.25):

The inductances in the inductance matrix in Equation (3.18) are not

dependent on time.
The quantities in d and qdirections are decoupled. (The induction
matrix is block diagonal: one 2 2 matrix and one 1 1 matrix.)
The second observation above leads to a picture of the transformed system according to Figure 5.2.

5.3

Voltage Equations for the Synchronous Machine

For the three stator circuits and the three rotor circuits the following relations can be written:

a ,
ua = ra ia
b ,
(5.26)
u = rb ib
b
c ,
uc = rc ic

66

5. Model of the Synchronous Machine

ra

ia
+

rF

iF

uF
-

Figure 5.3. Sign convention for stator and rotor circuits.

and

F ,
uF = rF iF +

0 = rD iD + D ,

Q .
0 = rQ iQ +

(5.27)

Equations (5.26) and (5.27) can be written more compactly in vector

form,
 





abc

iabc
Rabc
0
uabc

.
(5.28)
=
F DQ
iF DQ
0
RF DQ
uF DQ

The vector uF DQ is defined as

uF DQ = (uF , 0, 0)T ,

(5.29)

while the other vectors are defined as before. Rabc and RF DQ are diagonal
3 3 matrices.
If Equation (5.28) is multiplied by Pex according to Equation (5.17), all
quantities are transformed to the dqsystem, i.e.
 





abc
P
i0dq
P Rabc P 1
0
u0dq

. (5.30)
=
iF DQ
0
RF DQ
uF DQ
F DQ
The matrix P Rabc P 1 is denoted R0dq , and if ra = rb = rc = r, which in
most cases is true,
R0dq = Rabc = rI
(5.31)
is valid.

5.3. Voltage Equations for the Synchronous Machine

67

To get Equation (5.30) expressed solely in dqquantities, also the last

term on the right hand side must be expressed in these. Since P is time
dependent, it is important to remember that
P 6= 0 ,

(5.32)

abc ,
0dq = d (P abc ) = P abc + P

dt

(5.33)

abc =
0dq P abc =
0dq P P 1 0dq .
P

(5.34)

which leads to

and thus
Equation (5.30) can hence be written as
 
 

0dq

P P 1 0dq
=

+
.
0
F DQ
(5.35)
1

Some trivial computations show that the matrix P P can be expressed as

u0dq
uF DQ

R0dq
0
0
RF DQ

P P 1



i0dq
iF DQ

0 0 0
= 0 0 .
0 0

(5.36)

The voltage equations in the dqsystem can thus be written in component

form as

0 ,
u0 = ri0
d + q ,
(5.37)
u = rid
d
q d ,
uq = riq

and

F ,
uF = rF iF

0 = rD iD D ,

Q .
0 = rQ iQ

(5.38)

In the previous section, expressions for the dependencies of the flux linkages
on the currents in the different windings were derived. To further simplify
the expressions that were obtained, the per unit system for the different
windings is now introduced so that all mutual inductances in the daxis
are equal, and all in the qaxis are equal. (In our case, only one damping
winding in the qaxis was considered, but in a more general case several
damper windings can be considered.) We introduce
q

3
MF
q2

3
2 MQ

3
2 MD

= LAQ .

= MR = LAD ,

(5.39)

68

5. Model of the Synchronous Machine

The different fluxes can then be written as

0 = L0 i0 ,
= Ld id + LAD (iF + iD ) ,
d
q = Lq iq + LAQ iQ ,

and

F = LF iF + LAD (id + iD ) ,
= LD iD + LAD (id + iF ) ,
D
Q = LQ iQ + LAQ iq .

(5.40)

(5.41)

Equations (5.37) and (5.38) together with Equations (5.40) and (5.41) now
describe the electrical dynamics of a synchronous machine completely. These
equations together with a description of the external system unequivocally
determine the behaviour of the synchronous machine during different disturbances. In Figure 5.4, a graphical description of these equations is given.
In Equation (5.37), we observe that the emf in d and qdirection consists
of two terms: one that is a time derivative of the absolute value of the flux
linkage and one that arises because the field winding is rotating. The first
of these is usually called stator transient and the other rotational emf. In
steady state, the first of these vanishes, and the whole emf is created by the
d and
q are
rotation of the field winding. It can be shown that the terms
in most applications much smaller than d and q , which justifies that
the first ones are often neglected.

5.4

Synchronous, Transient, and Subtransient Inductances, and Time Constants.

The complete description of the synchronous machine given in the previous

section can be simplified and made more physical if a number of new parameters that can be expressed in the already defined ones are introduced.
In steady state and a sufficiently long time after a disturbance, the currents
in the damper windings vanish, and the inductances treated in this section
describe how the currents are linked with the fluxes directly after a disturbance and in steady state. Further, the time constants that specify how
fast the currents in the damper windings decay are derived. Earlier, the
synchronous inductances Ld and Lq were defined. They are repeated here
for completeness.

Ld = Ls + Ms + 32 Lm
(5.42)
Lq = Ls + Ms 23 Lm
These inductances describe the synchronous machine in steady state. For a
synchronous machine with salient poles, like a hydro power generation unit,
Lm > 0 and thus Ld > Lq , while Lm 0 for machines with round rotor
leading to Ld Lq .

5.4. Synchronous, Transient, and Subtransient Inductances, and Time Constants.69

rF
+
LF

uF
-

rd

id

iF

+
L AD

rD

Ld

LD

ud

q
+
rq

iq

iD

+
L AQ

rQ

Lq

LQ

uq

d
+
r0

i0

iQ

+
L0

Rotor Circuits

Stator Circuits

Figure 5.4. Graphical description of the voltage equations and the

coupling between the equivalent circuits.

If all rotor windings are short circuited, and a symmetrical three phase
voltage is applied on the machines terminals, the flux linkage in the daxis
initially depends on the subtransient inductance and after a couple of periods
on the transient reactance. A voltage is applied to the stator windings,

ua
cos

ub = 2U cos( 2 ) c(t) ,
(5.43)
3
2
cos( + 3 )
uc

with a step function c(t), i.e. c(t) = 0 for t < 0 and c(t) = 1 for t > 0. If the
voltage vector in Equation (5.43) is Parktransformed,

u0
0

ud = 3U c(t)
(5.44)
uq
0

u0

70

5. Model of the Synchronous Machine

is obtained. For t = 0+, that is, directly after the voltage is applied on
the terminals, the flux linkages F and D are still zero, since they cannot
change instantly:

0 = LF iF + LAD (id + iD )
(5.45)
0 = LD iD + LAD (id + iF )
From (5.45), iD and iF can now be expressed in id ,

LF LAD L2AD

id ,
i
=

D
LF LD L2AD

LD LAD L2AD

id .
iF =
LF LD L2AD

(5.46)

The expression for iD and iF can be related to d in (5.40), giving



LD L2AD + LF L2AD 2L3AD
(5.47)
id = Ld id ,
d = Ld
LF LD L2AD
with the subtransient inductance Ld in the daxis
Ld = Ld

LD + LF 2LAD
.
LF LD /L2AD 1

(5.48)

The subtransient inductance in the daxis, Ld , is thus defined as the initial

flux linkage of the stator current in per unit, when all rotor windings are
shortcircuited and with no current flowing in them before the disturbance.
If there is no damper winding in the daxis, or if the current in the
damper winding has decayed to zero, i.e. iD = 0, the same reasoning and
the same computations as above lead to


L2AD
(5.49)
id = Ld id ,
d = Ld
LF
with the transient inductance in the daxis, Ld , defined as
Ld = Ld

L2AD
.
LF

(5.50)

Show this!
The transient inductance in the daxis, Ld , is thus defined as the flux
linkage in the daxis in per unit of the stator current with the assumptions
above. For a machine with damper windings, that is the flux some periods after the disturbance when the current in the damper winding has
vanished. The time constant for this is derived later.
An equivalent analysis to that above can be performed for the qaxis,
but, since no field winding exists in the qaxis, the terminology is somewhat

5.4. Synchronous, Transient, and Subtransient Inductances, and Time Constants.71

different. For a machine with salient poles and damper winding in the qaxis,
the effective inductance after the current in the damper winding has decayed
is practically equal to the synchronous inductance. Hence it is sometimes
said that, for machines with salient poles, the transient and synchronous
inductances in the qaxis are equal.
If the reasoning above is repeated with the modification that cos in
Equation (5.43) is changed to sin etc., one obtains that
Lq = Lq

L2AQ
LQ

(5.51)

for a synchronous machine with an equivalent damper winding in the qaxis.

According to the reasoning above,
Lq = Lq .

(5.52)

If one additional damper winding is introduced into the qaxis, a value of

Lq can be defined that is different from Lq . In general, for the inductances
defined,
Ld < Ld < Ld ,
(5.53)
Lq < Lq < Lq .
(For the derivation of relationships defining Lq and Lq if more than one
damper winding in the qaxis is modelled, we refer to books on the theory
of electrical machines.)
The time constants that determine how fast transients in the rotor windings decay after a disturbance are easiest to compute if it is assumed that the
stator circuits are open, i.e. ia = ib = ic = 0, or equivalently i0 = id = iq = 0.
This is not consistent with the derivation above, but the obtained results
can be used to compute the time constants that are valid when the generator
is connected to a network. The time constants that will be derived here are
those normally specified by the manufacturer of the generator and are easily
obtained by measurements.
Suppose now that the stator windings are open and look at a step in the
field voltage at the time t = 0, i.e. uF = UF c(t).

F = UF c(t)
rf if +
(5.54)
D =0
rD iD +
Since id = 0,


D = LD iD + LAD iF ,
F = LF iF + LAD iD .

(5.55)

At t = 0+, D = 0 and thus

iF (0+) =

LD
iD (0+) .
LAD

(5.56)

72

5. Model of the Synchronous Machine

xd (p.u.)
xq (p.u.)
xd (p.u.)
xd (p.u.)
xq (p.u.)
xq (p.u.)
(s)
Tdo
(s)
Tdo
(s)
Tqo
(s)
Tqo
H (s)

Round Rotor
1.0 2.3
1.0 2.3
0.15 0.4
0.12 0.25
0.3 1.0
0.12 0.25
3.0 10.0
0.02 0.05
0.5 2.0
0.02 0.05
3 5 (n = 3000 rpm)
5 8 (n = 1500 rpm)

Salient Pole
0.6 1.5
0.4 1.0
0.2 0.5
0.15 0.35

0.2 0.45
1.5 9.0
0.01 0.05

0.01 0.09
1.5 5

Table 5.1. Typical values of some parameters of synchronous machines.

The current iF (0+) can now be eliminated from the equations, which leads
to
LAD /LF
rD
diD
iD = UF
.
(5.57)
+
2
dt
LD LAD /LF
LD L2AD /LF
Here, it has been assumed that rD rF , which generally is true. The
, can now
subtransient time constant of the open circuit in the daxis, Tdo
be defined as
LD L2AD /LF

Tdo
=
.
(5.58)
rD
When iD has vanished, i.e. when iD = 0, the field current is determined
solely by the upper of Equations (5.54), and the transient time constant of
, is given by
the open circuit, Tdo

Tdo
= LF /rF .

(5.59)

Accordingly, for a synchronous machine with a damper winding in the q

axis,

Tqo
= LQ /rQ
(5.60)
is defined.
The quantities introduced in this section are important parameters of a
machine and are usually given by the manufacturer of the machine. The reasons for this are that they are easily measured and that they are introduced
in a natural way into the simplified models we will derive in the next section.
In Table 5.1, typical values for some of the discussed parameters for different
types and sizes of synchronous machines are given. Reactances instead of
inductances are given in Table 5.1, i.e. xd = 0 Ld etc., with 0 = 2f0 .

73

5.5. Simplified Models of the Synchronous Machine

5.5

Simplified Models of the Synchronous Machine

A complete and exact model of the synchronous machine, considering the

assumptions and approximations made, was presented in Sections 3.2 and
3.3. If the synchronous machine is simulated in, for example, a stability
program, these are often the equations used to model the synchronous machine. Nevertheless, it is, for several reasons, often meaningful to use models
that comprise more simplifications and approximations than those derived
earlier. If a better insight into a problem is wanted or if a problem is analyzed without using computer simulations, simplified representations of the
synchronous machine are often the only possibility. Here, such a model will
be derived under the following assumptions:
d=
q = 0.
The stator transients are neglected, i.e. we set
The daxis contains no damper windings.
In the qaxis, one damper winding is modelled.
With these assumptions, the synchronous machine is described by the
equations

d = Ld id + LAD iF ,

q = Lq iq + LAQ iQ ,
(5.61)
F = LF iF + LAD id ,

Q = LQ iQ + LAQ iq ,

F ,
uF = rF iF +
(5.62)
Q ,
0 = rQ iQ +

ud = rid + q ,
(5.63)
uq = riq d .
For completeness, some relations defined above are repeated here.

Ld = Ld L2AD /LF
Lq = Lq L2AQ /LQ

(5.64)

The goal is now to eliminate all quantities with indices F and Q, except for
uF , from the equations above to get a model that can be used to represent
the synchronous machine as a component in a system. That means that the
only quantities that should be present in the model are stator voltages, the
stator current, and uF , which is a control variable that can be changed by
the excitation system as described in Chapter 6.
From the Equations (5.61), iQ is eliminated, which leads to
q

LAQ
Q = Lq iq .
LQ

(5.65)

74

5. Model of the Synchronous Machine

Now, by defining
ed =

LAQ
Q ,
LQ

(5.66)

Equation (5.65) can be written as

q ed / = Lq iq .

(5.67)

Equivalent derivations can be done for the rotor circuit in the daxis,
i.e. for the field winding.
d

LAD
F = Ld id
LF

(5.68)

is obtained, which, using the definition

LAD
F ,
LF

(5.69)

d + eq / = Ld id .

(5.70)

(5.71)

eq =
can be written as
Now, the quantities

ed = LAQ iQ ,
eq = LAD iF

are introduced. From the equations above, the relationships

(
ed = ed (xq xq )iq ,
eq = eq + (xd xd )id ,

(5.72)

with xd = Ld etc. can be obtained.

According to Equations (5.71) and (5.72),
iQ =

ed (xq xq )iq
.
LAQ

(5.73)

Substituting this into the second equation of (5.62), which together with
(5.66) give
e (xq xq )iq
LQ
+
e = 0 ,
(5.74)
rQ d
LAQ
LAQ d
where it has been assumed that = 0 or = 0 . Equation (5.74) can also
be written as

Tqo
e d + ed (xq xq )iq = 0 .
(5.75)
For the rotor circuit in the daxis, i.e. for the exciter winding, accordingly

Tdo
e q + eq + (xd xd )id =

LAD
uF
rF

(5.76)

5.5. Simplified Models of the Synchronous Machine

75

is obtained, which, using

eF =

LAD
uF ,
rF

(5.77)

can be written as

e q + eq + (xd xd )id = eF .
Tdo

(5.78)

If Equations (5.75) and (5.78) are Laplacetransformed and the voltage

equation (5.63) is rewritten with the introduced quantities, we obtain


   

rd xq
ed
id
ud
,
(5.79)
+
=
eq
iq
uq
xd rq
with ed given by
ed =

(xq xq )iq

1 + sTqo

(5.80)

and eq from
eq =

eF + (xd xd )id
.

1 + sTdo

(5.81)

This model, i.e. Equations (5.79)(5.81), together with the swing equation,
are often called the fourth order model. This model of the synchronous
machine demands, with the assumptions made here, four state variables: ed
and eq and the mechanical quantities and .
Often, also the damper winding in the qaxis can be neglected. That
leads to a third order model, which can be written as
 


 

0
rd xq
ud
id
+
=
,
(5.82)
eq
xd rq
uq
iq
with eq according to Equation (5.81).

76

5. Model of the Synchronous Machine

6
The Excitation System of the Synchronous
Machine
In this chapter, a short description of the function and design of the most
common excitation systems is given. Models of these systems for implementation in simulation programs and for analytical studies are also given.

6.1

Construction of the Excitation System

The main purpose of the excitation system is to feed the field winding of the
synchronous machine with direct current so that the main flux in the rotor
is generated. Further, the terminal voltage of the synchronous machine is
controlled by the excitation system, which also performs a number of protection and control tasks. A schematic picture of a generator with excitation
system is depicted in Figure 6.1. Below, a short description of the functions
of the different blocks in Figure 6.1 is given:
The exciter supplies the field winding with direct current and thus
comprises the power part of the excitation system.
The controller treats and amplifies the input signals to a level and
form that is suited for the control of the exciter. Input signals are pure
control signals as well as functions for stabilizing the exciter system.
The voltage measurement and load compensation unit measures the terminal voltage of the generator and rectifies and filters it.
Further, load compensation can be implemented if the voltage in a
point apart from the generator terminals, such as in a fictional point
inside the generators transformer, should be kept constant.
The power system stabilizer, PSS, gives a signal that increases
the damping to the controller, cf. Chapter 7. Usual input signals for
the PSS are deviations in rotor speed, accelerating power, or voltage
frequency.
The limiter and protection can contain a large number of functions
that ensure that different physical and thermal limits, which generator and exciter have, are not exceeded. Usual functions are current
limiters, overexcitation protection, and underexcitation protection.
Many of these ensure that the synchronous machine does not produce
or absorb reactive power outside of the limits it is designed for.
77

78

6. The Excitation System of the Synchronous Machine

Limiter and
Protection
Voltage
Measurements
Load Compensation

Synchronous
Regulator

Exciter

Machine

PSS

= To the Power System

Figure 6.1. Schematic picture of a synchronous machine with excitation system with several control, protection, and supervisory functions.

Today, a large number of different types of exciter systems is used. Three

main types can be distinguished:
DC excitation system, where the exciter is a DC generator, often
on the same axis as the rotor of the synchronous machine.
AC excitation system, where the exciter is an AC machine with
rectifier.
Static excitation system, where the exciting current is fed from a
controlled rectifier that gets its power either directly from the generator terminals or from the power plants auxiliary power system,
normally containing batteries. In the latter case, the synchronous machine can be started against an unenergised net, black start. The
batteries are usually charged from the net.
Below, a more comprehensive treatment of some of the functions described above is given.

6.2

Compensation Equipment

Figure 6.2 shows the block diagram of a compensation circuit, consisting of

a converter for measured values, a filter, and a comparator.
There are several reasons for the use of compensation in voltage control
of synchronous machines. If two or more generators are connected to the

79

6.3. DC Excitation Systems

UT
IT

U c = U T + ( R c + j X c )I T

Uc

1
-----------------1 + sT R

U err

U ref
Figure 6.2. Block diagram of compensating circuit.

same bus, the compensation equipment can be used to create an artificial

impedance between those. That is necessary to distribute the reactive power
in an appropriate way between the machines. The voltage is measured
somewhat inside the generator, corresponding to positive values of Rc
and Xc in Figure 6.2. If a machine is connected with a comparatively large
impedance to the system, which usually is the case since the generators
transformer normally has an impedance in the order of magnitude of 10%
on basis of the machine, it can be desirable to compensate a part of this
impedance by controlling the voltage somewhat inside of that impedance.
This then corresponds to negative values of Rc and Xc . As a rule, Xc is
much larger than Rc .

6.3

DC Excitation Systems

Today, hardly any DC excitation systems are being installed, but many of
these systems are still in operation. Generally, it can be said that there is
a large number of variants of the different excitation systems listed above.
Every manufacturer uses its own design, and demands that depend on the
application often lead to considerable differences in the detailed models of
the devices in each group. Here, typical examples for models will be given. In
reality, the models given by the manufacturers and power suppliers must be
used. One example of a DC excitation system, the IEEE type DC1 system,
is given in Figure 6.3. The input signal for the controller is the voltage
error Uerr from the compensation equipment. The stabilizing feedback UF
is subtracted, and sometimes a signal from the PSS is added. Both these
signals vanish in steady state. The controller is mainly described by the
dominating time constant TA and the amplification KA . The limits can
represent saturation effects or limitations of the power supply. The time
constants TB and TC can be used to model internal time constants in the
controller. These are often small and can then usually be neglected.
The output signal from the voltage controller, UR , controls the exciter.
The exciter consists of a DC machine that can be excited independently or
shunt excited. For shunt excited machines, the parameter KE models the

80

6. The Excitation System of the Synchronous Machine

U PSS
U RMAX

+
+

U err

KA
------------------1 + sT A

1 + sT
------------------C1 + sT D

UR

1
--------sT E

EF

U RMIN

UF

U FE

SE ( E F ) + K E

sK F
------------------1 + sT F

Figure 6.3. Model of DC exciter system (IEEE Type DC1).

setting of the field regulator. The term SE represents the saturation of the
exciter and is a function of the exciters output voltage, EF . If saturation is
neglected, that is SE = 0, the effective time constant of the exciter becomes
TE /KE , and its effective amplification is 1/KE .

6.4

AC Excitation Systems

For AC excitation systems, the exciter consists of a smaller synchronous machine that feeds the exciter winding through a rectifier. The output voltage
of the exciter is in this case influenced by the loading. To represent these
effects, the exciter current is used as an input signal in the model. In Figure 6.4, an example of a model of AC exciter systems is shown (IEEE type
AC1). The structure of the model is basically the same as for the DC excitation system. Some functions have been added. The rectifier of the exciter
prevents (for most exciters) the exciter current from being negative. The
feedback with the constant KD represents the reduction of the flux caused
by a rising field current IF . That constant depends on the synchronous and
transient reactances of the exciter. The voltage drop inside the rectifier is
described by the constant KC , and its characteristic is described by FEX ,
which is a function of the load current.
DC and AC excitation systems are sometimes called rotating exciters,
since they contain rotating machines. That distinguishes them from static
excitation systems, which are described in Section 6.5.

6.5

Static Excitation Systems

In static excitation systems, the exciter winding is fed through a transformer

and a controlled rectifier. By far most exciter systems installed today are of

81

6.5. Static Excitation Systems

U PSS
U RMAX

U err

KA
------------------1 + sT A

1 + sT C
------------------1 + sT D

UR

+
-

1
--------sT E

U FE

sK F
------------------1 + sT F

F EX = f ( I H )

K E + SE

EF

F EX

U RMIN

UF

UE

K CI F
I H = ------------UE

IF

KD

Figure 6.4. Model of an AC exciter system (IEEE Type AC1).

U PSS
U T U RMAX - K C I FE

U IMAX

U err

UF

1 + sT
------------------C1 + sT D

KA
-----------------1 + sT A

EF

U T U RMIN - K C I FE

U IMIN
sK F
-----------------1 + sT F

Figure 6.5. Model of a static exciter system.

that type, and a large number of variants exists. The primary voltage source
can be a voltage transformer that is connected to the generator terminals,
but even a combination of voltage and current transformers can be found.
With the latter arrangement, an exciter current can be obtained even if the
voltage at the generator terminals is low, for example during a ground fault
in or near the power plant. Sometimes, it is possible to supplement these
voltage sources by using the auxiliary power of the power plant as voltage
source. That makes it possible to start the generator in an unenergised net.
An example of a model of a static exciter system is shown in Figure 6.5.
Static excitation systems can often deliver negative field voltage and
even negative field current. However, the maximum negative field current is
usually considerably lower than the maximum positive field current.
The time constants are often so small that a stabilizing feedback is not
needed. The constant KF can then be set to zero. Since the exciter system

82

6. The Excitation System of the Synchronous Machine

is normally supplied directly from the generator bus, the maximum exciter
voltage depends on the generators output voltage (and possibly its current).
This is modelled by the dependency of the limitations of the exciter output
on the generators output voltage. The constant KC represents the relative
voltage drop in the rectifier.

7
Damping in Power Systems
In this chapter, a short introduction to damping in a power system is given.
Different sources of positive and negative damping are discussed, and methods to improve the damping are given.

7.1

General

What damping in the context of electromechanical oscillations in a power

system means is quite selfevident. Normally, two different kinds of electrical torques appear at a generator rotor that is oscillating: a synchronizing
torque Ts and a damping torqueTd . The synchronizing torque Ts is
in phase with the deviation in rotor angle , whereas the damping torque
Td is in phase with the deviation in rotor speed . The synchronizing
torque, also called the synchronizing power, strives, if it is positive, to bring
the rotor back to the stable equilibrium in which the mechanical power is
equal to the electrical power. When the generator has reached an operating
point where the synchronizing power no longer can return the system to
the stable equilibrium, the generator will fall out of phase. The variation
of the synchronizing torque with the rotor angle determines, together with
the machines moment of inertia the frequency of rotor oscillations. The
partitioning into synchronizing and damping torque is shown in Figure 7.1.
Damping is neglected in the classical model. Therefore, the system will,
after a disturbance, either fall out of phase (instability) or oscillate with unchanged amplitude. This is not realistic, since real systems contain damping.
The damping torque depends on the time derivative of the rotor angle in
such a way that the oscillation is damped. Normally, the damping torque is
rather small and thus influences the oscillation frequency only marginally.
It mainly influences the amplitude. In a synchronous machine, the main
contributors to damping are the damper windings and the field winding.
If the modes of oscillation in a system are determined by computing
the eigenvalues of the linearised systems Jacobian matrix, changes in the
synchronizing and damping torques will become apparent as follows: An
increase of the synchronizing torque moves the eigenvalue parallel to the
imaginary axis towards larger values. This corresponds to an increase in the
spring constant in a mechanical analogy. If the damping torque is increased
instead, the eigenvalue will move parallel to the real axis to the left. In
the classical model, all eigenvalues will be situated on the imaginary axis.
83

84

7. Damping in Power Systems

T e

T d

T s

Figure 7.1. Partitioning of electrical torque in synchronizing and

damping components.

Necessary for stability is that no eigenvalues are situated in the right half
plane. This corresponds to positive Ts as well as positive Td in Figure 7.1.

7.2

Causes of Damping

As mentioned earlier, the internal damping of a generator comes from the

windings in the rotor circuit. That damping is determined by phases and
amplitudes of the oscillating torques caused by induced currents in exciter
winding and damping windings. Further, some loads contribute with positive damping. These contributions originate from the frequency dependency
of the loads, but also their voltage dependency contributes.
Generally, the inner damping of the generators decreases with decreasing
frequency of the oscillations. The currents in the damping windings decay,
and hence, for very slow oscillations, their contribution is small.
Low or negative damping in a power system can lead to spontaneous appearance of large power oscillations. This can, in the worst case, necessitate
tripping of lines and it must be avoided. That type of instability is called
small signal instability or instability caused by low damping. (Earlier, that
type of instability was called dynamic instability.)
A common reason for low damping is the use of voltage controllers with
high gain. That was experienced in generators feeding a strong net through
a line. Such a configuration can also be analyzed comparably easily. It can
be shown that an eigenvalue with positive real part can occur when large
amounts of power are transmitted and voltage controllers with high gains are
used. Before the reason behind this phenomenon was known, the problem
was solved by operating the generator with manual voltage control, or by
making the voltage controller slower or decreasing its gain.
To explain that mechanism in detail is beyond the scope of this com-

85

7.3. Methods to Increase Damping

K STAB

sT W
-------------------1 + sT W

1 + sT 1
-----------------1 + sT 2

U ref

Figure 7.2. Block Diagram of a simple PSS.

pendium, but in summary, it can be said that the rotor angle influences
the generator voltage, which through the voltage controller influences the
transient emf, which influences the electrical torque. Now it turns out that,
when the load on the machine is high, the phase angle can be such that a
contribution with negative damping is obtained. If the amplification in the
voltage controller is high, that negative contribution can be significant.

7.3

Methods to Increase Damping

Several methods for increasing the damping in a power system are available.
The simplest and usually cheapest way is the installation of power system
stabilizers, PSS, in the generators. The operating principle for these is very
simple. To increase the damping in the system, a signal is added to the
reference voltage of the generators voltage controller. The phase of this
signal should of course be such that it results in a positive contribution to
the damping. Thus, the same physical mechanism in the system of generator
and voltage controller that above resulted in negative damping is used to
obtain positive damping.
Such a power system stabilizer usually utilizes the rotor deviation from
the synchronous frequency as input signal. Sometimes, other signals
that contain the same information can be used, like Pe or Te . A diagram
illustrating the principle mode of operation of a PSS is given in Figure 7.2.
The input signal, in this case , first passes a highpass filter to ensure that
permanent frequency deviations do not contribute. The next filter shifts the
phase appropriately for the critical oscillation frequency so that a positive
contribution to damping is obtained. The constant Kstab determines the
size of that contribution. That constant should of course not be chosen
larger than necessary to obtain the needed damping, since this could lead
to undesired side effects.
Other possibilities for increasing the damping in a system are different
types of controllable equipment that may be installed in the system, such as
HVDC (High Voltage Direct Current) or SVC (Static Voltage Condensers).
These components can often give large contributions to damping, but they
are usually too expensive to install them only to increase the damping, and
the existing equipment is not always located optimally for damping purposes.

86

7. Damping in Power Systems

8
Load Modelling
Since, neglecting losses, an equal amount of power is consumed in the loads
in the system as is generated in the generators, the load characteristics are on
principle just as important for the system properties as the generators. That
is, however, not reflected in the level of detail and the accuracy usually used
in load models for analyzing system stability. This chapter discusses briefly
how load characteristics influence the system stability and which problems
arise in the derivation of appropriate load models. The most common load
models are presented.

8.1

The Importance of the Loads for System Stability

The characteristics of the loads influence the system stability and dynamics
in many different ways. The voltage characteristics of the loads have a
direct influence on the accelerating power for generators nearby and are
thus very important for the behaviour during the first oscillation after a
fault. It has been shown in section 2.2.1 that the frequency dependency of
the loads influences directly how large the frequency deviation after different
system disturbances will become. The frequency dependency of the loads
also influences the system damping. The same is true for their voltage
dependency since it influences the voltage control.
This compendium concentrates on what is usually called angular stability, or synchronous stability, that is, the ability of the generators to stay
synchronized after disturbances. Another important property of a power
system is the ability to keep the voltages in the system within acceptable
limits during disturbances. This is a measure for the voltage stability of
the system. Voltage stability is highly dependent on the balance of reactive
power in the system, but also the active power has some influence here. It
is obvious that the voltage dependency of the loads is of high importance
for the systems voltage stability.
It is for several reasons difficult to derive good load models. (Of course,
deriving models for single load objects is formally not very difficult. Loads
here are, however, lumped loads as they are perceived from a bus in the
high voltage grid.) First, it is difficult to estimate the composition of the
loads, since it varies during the day as well as during the year. Further,
this composition varies from bus to bus. Thus, sometimes different load
models have to be used at different buses, depending on the composition of
the loads, for example industrial loads, domestic loads, and rural loads.
87

88

8.2

8. Load Modelling

Load Models

For studies of angular stability, loads are usually modelled with static models. Sometimes, large induction motors have to be represented individually
by special models to obtain the correct dynamic behaviour. Dynamic load
models for lumped loads have begun to be used during the last few years,
especially for studying voltage stability, but those are expected to be used
in the future more widely and even for other types of studies. The frequency
dependency is generally modelled according to Section 4.2, page 47.

8.2.1

Static Load Models

For traditional stability studies, where the investigated time frame is at

most around 10 s after the disturbance, the most commonly used model
types are static models. They are called static since they describe the load
using only algebraic equations. The modelled load dynamics are in these
cases so fast that they can be considered instantaneous compared with other
phenomena, like rotor oscillations, that are modelled. The most common
model for voltage dependency is
 
U
,
P = P0
U0
(8.1)
 
U
.
Q = Q0
U0
However, the load can also be modelled as an arbitrary polynomial in
(U/U0 ),
X  U i
ki
.
(8.2)
P = P0
U0
i

In Equations (8.1) and (8.2), U0 is the nominal voltage at nominal load,

P0 and Q0 . In Equations (8.1), the voltage exponents and are often
different.
If = 0, the load is called constant power load; if = 1, it is called
constant current load; and if = 2, it is a constant impedance load. It is
very common to model the (active) load for stability studies as consisting of
these three parts. Usually, a somewhat larger voltage exponent is used for
the reactive load.
Some examples for the voltage exponents of different loads are:
Electric heating: = 2, Q = 0.
Light bulbs: 1, 6, Q = 0.
Fluorescent tubes: 0, 9, 2.
To be able to include active and reactive losses in the underlying distribution
grid, the above models have to be modified.

89

8.2. Load Models

X = L

rs

i
+

+
e

Figure 8.1. Representation of induction motor.

A load model often used is the so called ZIPModel. The ZIP-Model is

a special case of the model in Equation (8.2), which contains three terms
with:
1 = 2, i.e. constant impedance (Z) load
2 = 1, i.e. constant current (I) load
3 = 0, constant power (P ) load

8.2.2

Motor Loads

Around half of all electric power used by the industry is used for operation
of motors. Sometimes, the load in certain nodes is dominated by electric
motors. It can then be justified to model those explicitly.
For small changes in voltage, a motor load behaves approximately like a
constant power load. For larger voltage changes, it can be necessary to use
a more accurate representation. Synchronous machines are then modelled
according to the models derived in Chapter 5, with the mechanical part Pm
depending on the characteristic of the mechanical load. A large part of the
motor load consists of induction motors that can be modelled as follows:
An induction motor is basically a synchronous machine with shortcircuited
exciter coil. If the exciter coil rotates with an angular speed different from
the rotating fluxes generated by the three phase coils, a current that generates a flux is induced in the exciter coil. Between the rotating synchronous
flux generated in the phase windings and the flux from the exciter winding, energy is exchanged. This is the basis for the function of the induction
motor.
The induction motor can, according to Figure 8.1, be described by a
voltage source behind an impedance. The value of L can be obtained from
the equivalent circuit of the induction motor shown in Figure 8.2.
In Figure 8.2
rs and Ls are the stator resistance and inductance,
Lm is the magnetizing inductance,
rr and Lr are rotor resistance and inductance.

90

8. Load Modelling

Lr

Ls

rr
---s

rs

Lm

Figure 8.2. Equivalent circuit for induction motor.

The slip s is defined by

s=

0
0

(8.3)

and thus L is given by

L = Ls +

Lm Lr
.
Lm + Lr

(8.4)

The dynamics are described by

de
1
+ (e + j0 (Ls + Lm L )i) + j0 se = 0 ,
dt
0
i=

e e
,
rs + j0 L

(8.5)

(8.6)

d
1
=
(Te Tl ) ,
dt
2Hm

(8.7)

Te = (e i ) .

(8.8)

Here,
is the machines angular speed,
0 is the systems angular speed,
Tl is the load torque,
0 = (Lr + Lm )/rr are no load operation constants.

8.2.3

Equivalent Dynamic Loads

The load models presented above are, as mentioned, valid for studying phenomena that do not last longer than about ten seconds after a disturbance.
If phenomena taking place in a longer time frame should be studied, slow
dynamics in the system have to be accounted for. These dynamics originate mainly from two different sources: The tap changers installed at lower
voltage levels that try to restore the voltage to the desired value and the
controllers installed at the loads.

91

8.2. Load Models

The control of tap changers can be done in several different ways, but
common to most systems are that tap changers are stepped, typically in
intervals of some tens of seconds, until the voltage is restored . Since this
control exists at different voltage levels (cascade coupled controllers), undesirable overshoots in the control can occur if the control loops are not
coordinated. Generally, the control has to be slower the lower the voltage
level is.
In Sweden, a large part of the load, at least in winter, consists in many
areas of heating loads. The changes in this type of load are determined by
thermostats. Hence, it takes some time until, for example, a voltage drop
becomes apparent. That time is determined by the thermal time constants
for what is heated, such as houses, and by the design of the thermostats.
Summarizing, it can be said that the dynamics determined by tapchanger control and load dynamics are highly complicated. Measurements are
needed to get reliable results. Measurements of load characteristics have
during the last few years become very important, and much work is being done in many utilities to investigate load characteristics under different
loading conditions.
A typical example of a load behaviour after a voltage drop is shown in
Figure 8.3. It is clearly visible how the load drops momentarily, as described
by the load models from Section 8.2.1, to recover later to a considerably
higher level. A rather general description is given by
Tp

dPr
+ Pr = Ps (U ) Pt (U ) ,
dt
Pl (t) = Pr + Pt (U ) ,

where
Pr (t) is a state variable,
Ps (U ) is a static model for the long term load behaviour,
Pt (U ) is a static model for the transient load behaviour,
Pl (t) is the value of the active load at the time t.
Of course, U = U (t) in the equations above.
For the reactive load, similar behaviour and equations are valid.

(8.9)
(8.10)

92

8. Load Modelling

Load Recovery Factor Definition

1.05

Active Power [pu]

Ps

0.95
Pt

0.9

PRF =

Pt - Ps
Pt

0.85

0.8

0.75

-5

10
Time [minutes]

15

Figure 8.3. The Transient Behaviour of the Load in Equations (8.9)

and (8.10) after a Step in the Voltage.

20

References
[1] P. Kundur, Power System Stability and Control, McGraw-Hill Inc.,
New York, 1994 (ISBN 0-07-035958-X)
[2] Dynamic Models for Steam and Hydro Turbines in Power System Studies, IEEE Trans. Power Appar. Syst. 1904-1915, Nov./Dec. 1973
[3] G. Andersson, Modelling and Analysis of Electric Power Systems, ETH,
2004

93

94

REFERENCES

Appendix A
Connection between per unit and SI
Units for the Swing Equation
If nothing else is given after a quantity, that quantity is in SIunits. If a
quantity is expressed in per unit, p.u. is given in brackets after the quantity (p.u.). For simplicity, it is assumed that the nominal electrical and
mechanical frequencies are equal, (rad/s).
In SIunits,
d2
J 2 = T ,
(A.1)
dt
with
J = moment of inertia for rotor turbine (kgm2 ),
= angle (rad),
= angular velocity (rad/s),
T = effective torque on the rotor turbine (Nm).
When using electrical degrees, Equation (A.1) is usually written as
M d2

= T ,
0 dt2

(A.2)

with

M = moment of inertia = J0 180

(Js/el ).
The H-factor, or constant of inertia, for synchronous machine i is defined
by
1
J 2
Hi = 2 0 ,
(A.3)
Si
with
Hi = constant of inertia for synchronous machine i (s),
Si = rated power of synchronous machine.
The per unit base for torques at synchronous machine i, Tbas,i , is given by
Tbas,i =

Si
,
0

(A.4)

leading to

Si
.
0
Using (A.3) and (A.5), (A.1) can be written as
T = T (p.u.)

Si
2Hi Si d2
2 =
T (p.u) ,
dt
0
02
95

(A.5)

(A.6)

96

A. Connection between per unit and SI Units for the Swing Equation

or

0
d2
=
T (p.u) .
dt2
2Hi

(A.7)

Equation (A.7) can also be written as

d
0
() =
T (p.u) ,
dt
2Hi

(A.8)

which is the same as

0
d
() =
T (p.u) ,
dt
2Hi
or

(A.9)

T (p.u)
d
((p.u.)) =
.
dt
2Hi

(A.10)

P = T m ,

(A.11)

Generally,
with the actual mechanical angular speed of the rotor m that, according
to the assumptions, is equal to the electrical angular speed . With the
equations above, this gives
P (p.u) = T (p.u) (p.u) .

(A.12)

Equation (A.10) now becomes

(p.u.)

P (p.u.)
1

.
2Hi
(p.u.)

(A.13)

For rotor oscillations, (p.u.) 1 and (A.13) can be approximated by

(p.u.)

=
or
=

P (p.u.)
,
2Hi

0
P (p.u.) .
2Hi

(A.14)

(A.15)

The most common equations in literature are (A.9), (A.10), (A.14), and
(A.15). Of these, (A.9) and (A.10) are exact if is the actual angular
frequency. Equations (A.14) and (A.15) are good approximations as long as
0 . That is valid for normal oscillations in power systems.

Appendix B
Influence of Rotor Oscillations on the
Curve Shape
If the relative movement between the field winding of a synchronous machine
and its phase windings is a purely rotating motion with constant angular
speed, the resulting induced voltages in the phase windings will be shaped
ideally like a sinusoid. From now on, it is assumed that the field winding is in
the rotor, while the phase windings are on the stator, but since the relative
motion determines the voltage in the phase windings, it is even possible
to think of stationary field windings and rotating phase windings. In all
modern larger synchronous machines, the field winding is on the rotor, so
the assumption above does have a practical background. However, almost
all relationships and conclusions are independent of this assumption.
For simplicity, consider a single phase synchronous machine according
to Figure B.1. A three phase machine has two more phase windings shifted
120 relative to the phase winding in the figure. The phase winding and
the exciter winding are arranged so that the flux linkage through the phase
winding is sinusoidally shaped as a function of the angle m in Figure B.1:
(t) = 0 cos m = 0 cos 0 t ,

(B.1)

with the angular speed of the rotor 0 according to the systems electrical
frequency. That flux induces a voltage in the phase winding that is given by
U (t) =

d
sin 0 t .
= 0 0 sin 0 t = U
dt

(B.2)

Now, we shall study how the flux linkage through the phase windings will
be influenced when rotor oscillations appear in the system. If the balance
between power into the generator and power from the generator, i.e. between
mechanical torque and electrical power, is disturbed, the rotor will start to
oscillate relative to an undisturbed reference rotor that continues to rotate
with the angular speed 0 . The rotor position can generally be described
by
m (t) = 0 t + (t) ,
(B.3)
where (t) is a solution of the swing equation. It has earlier been mentioned
that stable solutions of the swing equation for a synchronous machine connected to a strong grid consist of oscillations that are nearly sinusoidal with
frequencies on the order of magnitude of some tenths of a Hertz to some
97

98

B. Influence of Rotor Oscillations on the Curve Shape

m
N

Figure B.1. Schematic picture of single phase synchronous machine.

Hertz. To investigate how the linked flux, and thus the voltage and the
current, look during an oscillatory movement, a rotor motion according to
m (t) = 0 t + sin(r t + r ) ,

(B.4)

with the angular speed r corresponding to the oscillation frequency and

the amplitude of the oscillatory movement , is assumed. The flux linkage
can now be written as
(t) = 0 cos m (t) = 0 cos(0 t + sin(r t + r )) ,

(B.5)

which implies that the oscillatory movement contains a phaseangle modulation of the flux linkage. The momentary angular frequency, (t), is defined
for (t) as
d
(t) = (0 t + sin(r t + r ))
(B.6)
dt
and varies between 0 + r and 0 r . It can be shown (cf. text books
on modulation theory) that Equation (B.5) can be written as
(t) = 0

n=
X

Jn () cos((0 + nr )t + nr ) .

(B.7)

n=

Jn () is a Bessel function of the first kind with the argument and degree
n, as given by
Z
1
Jn () =
cos( sin x nx)dx .
(B.8)

An important property of Jn () that will be used later is
Jn () = (1)n Jn () .

(B.9)

99
1
0.9
0.8

J0()

0.7
0.6
0.5

J1()

0.4
0.3
0.2
0.1

J2()

J3()
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Figure B.2. Bessel functions of the first kind of order 0 to 3.

From Equation (B.7) follows that, in spite of that the momentary angular
frequency for (t) is between 0 +r and 0 r , (t) will have infinitely
many side bands with the frequencies 0 nr beside the fundamental
frequency 0 . A relevant question is how large the amplitudes of these
side bands are. Generally, the coefficients Jn () decay rapidly when the
order n becomes larger than the argument . In Figure B.2, values for
the first four Bessel functions are shown for the argument between 0
and 1. It should be observed that is measured in radians, so that 1
corresponds to approximately 57 , which in this context is quite a large
amplitude. Figure B.2 shows that side bands with n = 3 and larger can
be neglected even for amplitudes as large as = 1. (t) can thus be
approximated quite accurately by
(t) 0

n=2
X

Jn () cos((0 + nr )t + nr ) .

(B.10)

n=2

Typical oscillation frequencies are, in most cases considerably, lower than

3 Hz, so that practically the whole energy spectrum for (t) lies in the
frequency area f0 6 Hz, i.e. 50(60) 6 Hz. This justifies the usual
representation of the grid with the traditional phasor model with a constant
frequency corresponding to f0 .
If the amplitude is very small, say less than 0.2, corresponding to
an amplitude of approximately 10 , side bands with n = 2 and higher can

100

B. Influence of Rotor Oscillations on the Curve Shape

--2

--2

Figure B.3. Vector with superimposed oscillation, according to the text.

be neglected, see Figure B.2. Further, it can be shown that, if is small,

J0 () 1

2
,
4

(B.11)

,
J1 ()
2
are valid, and (t) can be written as
2

) cos 0 t + cos((0 +r )t +r ) cos((0 r )t r )) .

4
2
2
(B.12)
Equation (B.9) has here been used. That approximation is used even in a
context of power system analysis other than rotor oscillations, namely when
studying socalled subsynchronous oscillations, SSO. If there is a resonance
phenomenon at a subsynchronous frequency it is called subsynchronous resonance, SSR. The most common cause of SSO are torsional oscillations on
the axis connecting the turbine(s) and the generator rotor. The frequencies
of the natural oscillations on that axis are typically 5 Hz and higher. (For
an axis with n distinct masses, including generator rotor and, maybe,
exciter, there are n 1 different eigenfrequencies.) For the side band in
Equation (B.12), i.e. for 0 r , the frequency can deviate significantly
from the nominal frequency, so that it is normally not possible to look only at
the component with nominal frequency. Since the electrical damping for the
lower side band, the subsynchronous frequency, can be negative, due to, for
example, series compensation, the partitioning according to Equation (B.12)
has to be kept. The damping in the upper side band, the supersynchronous
frequency, is almost always positive.
To increase the understanding for the partitioning in Equation (B.12), a
more intuitive derivation than the stringent mathematical one using Bessel
(t) 0 ((1

101
functions can be given. Consider a vector with amplitude 1 that performs
small oscillations with an angular frequency r and amplitude . This can
be illustrated geometrically according to Figure B.3. The vector with filled
arrowhead oscillates symmetrically around the horizontal axis with the frequency r and the amplitude . That vector can now be partitioned into
the three vectors with unfilled arrowheads. One vector does not move and
lies along the horizontal axis. Two vectors with amplitude /2 rotate with
the angular frequency r according to Figure B.3. It is easily observed
that the sum of the vectors with hollow arrowhead is at all times equal to
the vector with filled arrowhead. Since the vectors rotate with the angular
frequency 0 with respect to a stationary system, Equation (B.12) is obtained directly from the projection of the vectors on to the horizontal axis,
with the modification that the factor for the fundamental frequency is one.