Professional Documents
Culture Documents
laboratory
Ekaterina Telegina
Master Thesis
PSL1510
iii
iv
Acknowledgements
First and foremost, I would like to thank my supervisor Theodor Borsche for
his continuous support and guidance during my work on this master thesis.
Thank you for offering such an interesting research topic. It has been a
pleasure working with you.
I would also like to thank Professor Dr. Goran Andersson for giving me
the opportunity to write a master thesis at the Power System Laboratory.
The Power System Analysis and Power System Dynamics and Control
courses that he taught further improved my knowledge on the subject of
power system operation and stability which was pivotal for the successful
completion of the present work.
My sincere appreciation goes to my friends for their patience and love.
Special thanks to Elena for her invaluable support during the hard times
and to Anton for his encouragement and understanding.
Finally, I am deeply grateful to my family for their constant love and
support. You always motivated me to work hard and do my best.
v
vi
Contents
List of Tables x
List of Symbols xv
1 Introduction 1
1.1 Background and Literature Overview . . . . . . . . . . . . . . 1
1.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
vii
3.4.1 Small-Signal Stability . . . . . . . . . . . . . . . . . . 28
3.4.2 Transient Stability . . . . . . . . . . . . . . . . . . . . 36
5 Simulation Results 55
5.1 IEEE Two-Area Test System . . . . . . . . . . . . . . . . . . 55
5.1.1 System Description . . . . . . . . . . . . . . . . . . . . 55
5.1.2 Small-Signal Stability Analysis . . . . . . . . . . . . . 56
5.1.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . 67
5.1.4 Transient Stability Analysis . . . . . . . . . . . . . . . 74
5.2 IEEE South East Australian Test System . . . . . . . . . . . 77
5.2.1 System Description . . . . . . . . . . . . . . . . . . . . 77
5.2.2 Small-Signal Stability Analysis . . . . . . . . . . . . . 78
5.2.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . 80
5.2.4 Transient Stability Analysis . . . . . . . . . . . . . . . 84
5.3 Discussion of Simulation Results . . . . . . . . . . . . . . . . 85
A Runge-Kutta Methods 91
Bibliography 107
viii
List of Figures
ix
C.1 A shunt connected to bus k [3] . . . . . . . . . . . . . . . . . 95
C.2 Lumped-circuit model of a transmission line [3] . . . . . . . . 96
C.3 Unified branch model [3] . . . . . . . . . . . . . . . . . . . . . 97
E.1 IEEE South East Australian five-area test system [4] . . . . . 102
x
List of Tables
xi
5.19 Values of the inertia constants M and damping coefficients
KD on 900 MVA base in the two-area test system (Case 5) . 72
5.20 Optimization results of the two-area test system (Case 6) . . 72
5.21 Values of the inertia constants M and damping coefficients
KD on 900 MVA base in the two-area test system (Case 6) . 73
5.22 Optimization results of the two-area test system (Case 7) . . 73
5.23 Values of the inertia constants M and damping coefficients
KD on 900 MVA base in the two-area test system (Case 7) . 73
5.24 Optimization results of the two-area test system (Case 8) . . 74
5.25 Values of the inertia constants M and damping coefficients
KD on 900 MVA base in the two-area test system (Case 8) . 74
5.26 Steady-state operating condition of the five-area test system . 78
5.27 Rotational inertia constants M and damping coefficients of
the five-area test system generators in Base Case and Low-
Inertia Case, calculated on 100 MVA base . . . . . . . . . . . 78
5.28 Results of transient overshoot computation in the five-area
test system in Base Case . . . . . . . . . . . . . . . . . . . . . 79
5.29 Parameters of the optimization program for the five-area test
system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.30 Optimization results of the five-area test system (Case 1) . . 81
5.31 Values of the inertia constants M and damping coefficients
KD on 100 MVA base in the five-area test system (Case 1) . . 81
5.32 Optimization results for the five-area test system (Case 2) . . 82
5.33 Values of the inertia constants M and damping coefficients
KD on 100 MVA base in the five-area test system (Case 2) . . 82
5.34 Optimization results for the five-area test system (Case 3) . . 83
5.35 Values of the inertia constants M and damping coefficients
KD on 100 MVA base in the five-area test system (Case 3) . . 83
E.1 Power flow input data for IEEE South Australian test system
[4] calculated on 100 MVA base . . . . . . . . . . . . . . . . . 103
E.2 Parameters of the branches of IEEE South Australian test
system [4] calculated on 100 MVA base . . . . . . . . . . . . . 104
E.3 Parameters of the aggregated synchornous machines of IEEE
South East Australian . . . . . . . . . . . . . . . . . . . . . . 104
E.4 Eigenvalues of the South East Australian test system in Base
Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
E.5 Eigenvalues of the South East Australian system in Low-
Inertia case . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
xii
List of Acronyms
AC Alternating Current
AVR Automatic Voltage Regulator
BESS Battery Energy Storage System
HVDC High Voltage Direct Current
IEEE Institute of Electrical and Electronics Engineers
PFC Primary Frequency Control
PSS Power System Stabilizer
RES Renewable Energy Sources
R-K Runge-Kutta
SMIB Single Machine Infinite Bus
SVC Static VAR Compensator
xiii
xiv
List of Symbols
xv
S The droop of PFC
TR The AVR time constant
Tt The turbine time constant
TW The PSS washout block time constant
T1 A PSS phase compensation block time constant
T2 A PSS phase compensation block time constant
Xfd The inductance of the field circuit of a synchronous machine
X1d The inductance of the d-axis damping circuit of a synchronous machine
X1q The inductance of the first q-axis damping circuit of a synchronous machine
X2q The inductance of the second q-axis damping circuit of a synchronous machine
Xadu The unsaturated mutual inductance between the stator and d-axis rotor circuits
of a synchronous machine
Xads The saturated mutual inductance between the stator and d-axis rotor circuits
of a synchronous machine
Xaqs The saturated mutual inductance between the stator and q-axis rotor circuits
of a synchronous machine
xvi
Vk The magnitude of the nodal voltage at bus k
k The angle of the nodal voltage at bus k
Y The admittance matrix of a transmission network
Ykm The magnitude of Y element in the k-th row and m-th column
km The angle of Y element in the k-th row and m-th column
xvii
xviii
Chapter 1
Introduction
1
2 CHAPTER 1. INTRODUCTION
1.3 Structure
This thesis is organized as follows: Chapter 2 briefly reviews the power sys-
tem stability fundamentals. Chapter 3 presents modelling of synchronous
machine, transmission network, and aggregated load for the rotor angle sta-
bility studies. System equations are formulated and system state matrix is
derived. Chapter 4 develops an optimization algorithm focused on improve-
ment of the damping of system modes under a transient frequency overshoot
constraint. Sensitivities of damping ratio and frequency overshoot to iner-
tia and damping changes are derived. Furthermore, implemenation of the
algorithm in MATLAB is described. Chapter 5 investigates the small-signal
stability of two test systems for various RES penetration cases and im-
plements the developed optimization algorithm. The impact of rotational
inertia changes on stability of the test systems is illustrated by providing
the results of transient simulations. Finally, a conclusion and an outlook of
the present work are given in Chapter 6.
4 CHAPTER 1. INTRODUCTION
Chapter 2
5
6 CHAPTER 2. POWER SYSTEM STABILITY FUNDAMENTALS
Rotor angle stability analysis involves the analysis of the effect of small
disturbances on the system of interest (small-signal stability) and the dy-
namic behaviour of the system subjected to a large disturbance (transient
stability).
Small-signal stability is the ability of the power system to maintain
synchronism under small disturbances. A great number of small distur-
bances occur in a system during its normal operation. They are primarily
caused by the constant variation of demanded and generated power. The
disturbances are considered to be sufficiently small to enable linearization
of the system equations for the purposes of analysis.
Small-signal stability problems could be divided in two groups, local and
global. Local problems are associated with rotor angle oscillations of a small
part of the system. As an example, generators of a certain power plant may
be oscillating against the rest of the power system. This type of oscillations
is called local plant mode oscillations. Other local problems that might
occur in a power system include interplant mode oscillations, control modes
and torsional mode oscillations [2].
Global small-signal stability problems are caused by oscillations involv-
ing a large group of generators. The oscillations of generators in one area
swinging against generators in another area are reffered to as interarea
mode oscillations. In large power systems, usually there are two forms of
interarea oscillations [2]:
An oscillation mode with a very low frequency (0.1-0.3 Hz) that in-
volves all the generators in the system. Generators of the intercon-
nected system are split in two groups, with one of the groups swinging
against another.
x = f (x, u, t)
(2.1)
y = g(x, u, t)
where x is the state vector with the state variables as elements, u is the
vector of inputs to the system, y is the vector of output variables, f and g
are vectors of nonlinear functions relating x and y to x and u. With n as
the order of the system of differential equations, r as the number of inputs,
and m as the number of output variables, the vectors have the following
form
x1 u1 f1
x2 u2 f2
x= ... u = ... f = ...
(2.2)
xn ur fn
y1 g1
y2 g2
y=
... g = ...
(2.3)
ym gm
In the rotor angle stability analysis, Equations (2.1) should represent
the dynamics of the power system in the time-scale relevant to rotor swings
(0.01 s - 10 s). The dynamic behaviour of the power system components,
namely generators, transmission network, static and dynamic loads, static
VAR compensators (SVC), etc., should be reflected adequately to the analy-
sis scope. Among the mentioned components, modelling of the synchronous
generators plays certainly the most important role for the investigation of
the rotor angle stability. Quite often the dynamic behaviour of a system
is described only by the differential equations associated with synchronous
generators, whereas all the other components are represented by algebraic
equations. For instance, the transient processes occuring in transmission
8 CHAPTER 2. POWER SYSTEM STABILITY FUNDAMENTALS
x = Ax + Bu (2.4)
y = Cx + Du (2.5)
where
f1 f1 f1 f1
x1 ... xn u1 ... ur
A = ... ... ... B = ... ... ...
fn fn fn fn
x1 ... xn u1 ... ur
g1 g1 g1 g1
x1 ... xn u1 ... ur
C = ... ... ... D = ... ... ... (2.6)
gm gm gm gm
x1 ... xn u1 ... ur
If all the eigenvalues have negative real parts, the original system is
asymptotically stable, i.e. it returns to the original state after being
subjected to a small perturbation.
If at least one of the eigenvalues has a positive real part, the original
system is unstable.
det(A I) = 0 (2.7)
Ai = i i (2.8)
i A = i i (2.9)
as
xi (t) = i1 c1 e1 t + i2 c2 e2 t + ... + in cn en t (2.12)
= j (2.13)
= (2.14)
2 + 2
The damping ratio of a decaying oscillatory mode should stay within the
limits
0<<1 (2.15)
d2 m
J = Tm Te (2.17)
dt2
where
J is the total moment of inertia of the synchronous machine
m is the mechanical angle of the rotor
Tm is the mechanical torque on the rotor
Te is the electrical torque on the rotor
to isolate the fault and thus ensure normal operational conditions for
as much equipment as possible,
d2 m
m J = Pm Pe (3.1)
dt2
13
14 CHAPTER 3. MODELLING OF POWER SYSTEM
where S is the MVA rating of the machine. The inertia constant shows how
much time it would take for a machine to decelerate from synchronous speed
to standstill if rated power is extracted from it and no mechanical power is
fed into it [3].
Another quantity that is broadly used in the literature is called the
mechanical starting time M , defined as
M = 2H (3.3)
2H d2
= Pm Pe KD (3.4)
0 dt2
where
KD - damping coefficient in p.u. torque/p.u. speed deviation
0 - synchronous electrical angular velocity of the rotor
Equation (3.4) is commonly reffered to as swing equation, as it repre-
sents swings in rotor angle during disturbances.
Using the following notation for the relative angular velocity in p.u.
1 d
r = (3.5)
0 dt
the swing equation can be rewritten in the form of a system of first order
differential equations:
1
pr = (Pm Pe KD r ) (3.6)
M
p = 0 r (3.7)
x = [ r ]T (3.8)
one can analyze the stability of a system. This modelling approach was
widely used in the early stability studies. Therefore, it is often referred to
as Classical Model. However, such a model does not take into account
the electromagnetic dynamics of the machine, such as dynamics of the rotor
circuits and effects of the voltage control devices on the field voltage. To in-
corporate the specified dynamic effects in the model, additional differential
equations are formulated further in this section.
0 Rfd
pfd = Efd 0 Rfd ifd (3.9)
Xadu
p1d = 0 R1d i1d (3.10)
p1q = 0 R1q i1q (3.11)
p2q = 0 R2q i2q (3.12)
where the subscripts fd, 1d, 1q, 2q stand for the quantities of the field
circuit, d-axis damping circuit, and q-axis damping circuits respectively.
denotes the flux linkage of a circuit, i designates the circuit current, R is the
resistance of a circuit, Efd is the exciter output voltage, 0 is the synchronous
angular velocity, and Xadu stands for the unsaturated mutual impedance.
Thus, the state vector should be augmented by the flux linkages of the
rotor circuits
x = [ r fd 1d 1q 2q ]T (3.13)
Vref
Terminal voltage
EFMAX
transducer Exciter
1 v1
Et KA E fd
1 sTR
EFMIN
The first block of the diagram represents the terminal voltage transducer.
It measures terminal voltage of the machine (Et ), rectifies and filters it with
an output
1
v1 = Et (3.14)
1 + pTR
Equation (3.14) could be rearranged to get the time derivative of v1 at the
left side:
1
pv1 = (Et v1 ) (3.15)
TR
This differential equation supplements the swing equation and Equations
(3.9-3.12) in modelling of the dynamic behaviour of a synchronous machine.
The voltage v1 should be therefore added to the state vector (3.13)
x = [ r fd 1d 1q 2q v1 ]T (3.16)
Vref
Terminal voltage
EFMAX
transducer Exciter
1 v1
Et KA E fd
1 sTR
EFMIN
Phase
Gain Washout compensation
sTW v2 1 sT1 vs
r KSTAB
1 sTW 1 sT2
Figure 3.2: Thyristor excitation system with AVR and PSS [2]
A gain block senses the value of the angular velocity deviation from
the synchronous speed (r ,) and with the gain KSTAB , it sets the level
of damping introduced by the PSS. The output signal of the gain block is
processed by the washout block with a time constant TW that serves as a
high-pass filter.
18 CHAPTER 3. MODELLING OF POWER SYSTEM
KSTAB 1
pv2 = (Pm Pe KD r ) v2 (3.22)
M TW
A phase compensation block serves to compensate for the phase lag be-
tween the exciter input and the air-gap torque of the generator. The phase
characteristic of the system depends on its state, and the settings of PSS
should be acceptable for a wide range of possible system conditions.
From Figure 3.2,
1 + pT1
vs = v2 (3.23)
1 + pT2
Hence
T1 1 1
pvs = pv2 + v2 vs (3.24)
T2 T2 T2
With pv2 given by (3.22), (3.24) can be rewritten as
T1 KST AB 1 T1 1 1
pvs = (Pm Pe KD r ) + ( )v2 vs (3.25)
T2 M T2 T2 TW T2
The value of vs is subject to a constraint
A new expression for the exciter output voltage according to Figure 3.2 is
Thus, the differential equation for the flux linkage of the field winding should
be adjusted once more:
0 Rfd
pfd = KA (Vref + vs v1 ) 0 Rfd ifd (3.28)
Xadu
The system of the synchronous machine differential equations is now
expanded with (3.22) and (3.25), and v2 and vs should be added to the state
vector
x = [ r fd 1d 1q 2q v1 v2 vs ]T (3.29)
3.1. SYNCHRONOUS MACHINE MODELLING 19
turbine time constants should not exceed 10-15 s. Typical values of Tt of the
high-pressure steam turbine are 0.1-0.4s, a re-heater has a larger time delay
(4-11s). The time constant of the delay between the intermediate and low
pressure turbines is in the order of 0.3-0.6s [3]. It should be noted, that (3.30)
describes only one turbine stage and therefore represents a simplified model
of a turbine control. A faster primary frequency control can be provided by
Battery Energy Storage Systems, as shown in [7].
The mechanical power output change Pm completes the state vector
that now consists of 10 state variables:
x = [ r fd 1d 1q 2q v1 v2 vs Pm ]T (3.31)
boundary conditions should relate the internal variables of the machine with
the demanded power output and, thus, with the rest of the power system.
This is achieved if boundary conditions are represented by the stator voltage
equations (3.33).
Stator Voltage Components
ed = Ra id + Xl iq aq
(3.33)
eq = Ra iq Xl id + ad
The demanded power output and the terminal voltage magnitude set-
point determine the generator currents id and iq , and internal variables of
the machine.
Equations (3.32) and (3.33) should be expressed in terms of the state
variables, currents, and terminal voltage magnitudes. Thus, the internal
variables of the machine (rotor currents, flux linkages ad and aq , and
electrical power demand Pe ) should be eliminated from (3.32) and (3.33) by
means of Equations (3.34),(3.35) and (3.37).
Rotor Currents
1
ifd = (fd ad )
Xfd
1
i1d = (1d ad )
X1d
1 (3.34)
i1q = (1q aq )
X1q
1
i2q = (2q aq )
X2q
Flux Linkages
00 fd 1d
ad = Xads (id + + )
Xfd X1d
(3.35)
00 1q 2q
aq = Xaqs (iq + + )
X1q X2q
where
00 1
Xads = 1 1 1
Xads + Xfd + X1d
(3.36)
00 1
Xaqs = 1 1 1
Xaqs + X1q + X2q
22 CHAPTER 3. MODELLING OF POWER SYSTEM
Electrical Torque
Since, as already mentioned, in p.u Pe = Te ,
Te = Pe = ad iq aq id (3.37)
1 00 fd 1d 00 1q 2q
pr = (Pm + Pm Xads (id + + )iq + Xaqs (iq + + )id
M Xfd X1d X1q X2q
KD r )
p = 0 r
0 Rfd 1 00 fd 1d
pfd = KA (Vref + vs v1 ) 0 Rfd (fd Xads (id + + ))
Xadu Xfd Xfd X1d
1 00 fd 1d
p1d = 0 R1d (1d Xads (id + + ))
X1d Xfd X1d
1 00 1q 2q
p1q = 0 R1q (1q Xaqs (iq + + ))
X1q X1q X2q
1 00 1q 2q
p2q = 0 R2q (2q Xaqs (iq + + ))
X2q X1q X2q
1
pv1 = (Et v1 )
TR
KSTAB 00 fd 1d
pv2 = (Pm + Pm Xads (id + + )iq +
M Xfd X1d
00 1q 2q 1
+ Xaqs (iq + + )id KD r ) v2
X1q X2q TW
T1 KSTAB 00 fd 1d
pvs = (Pm + Pm Xads (id + + )iq +
T2 M Xfd X1d
00 1q 2q 1 T1 1 1
+ Xaqs (iq + + )id KD r ) + ( )v2 vs
X1q X2q T2 T2 TW T2
1 1
pPm = Pm r
Tt STt
00 1q 2q
ed = Ra id + Xl iq Xaqs (iq + + )
X1q X2q
00 fd 1d
eq = Ra iq Xl id + Xads (id + + )
Xfd X1d
(3.38)
The system (3.38) models the dynamic behaviour of a synchronous gen-
erator but it should be supplemented by the initial values of the machine
state variables, since stability of a system significantly depends on its initial
operational state. The expressions for the calculation of the synchronous
machine initial setpoint are presented in Appendix B.
3.2. TRANSMISSION NETWORK MODELLING 23
Y = G + jB (3.39)
From Kirchhoffs Current Law, the expression for nodal current injec-
tions can be derived as
I =YE (3.40)
where
I is the current injection vector with elements Ik , k = 1, 2, ..., N
E is the nodal voltage vector with elements Uk ejk
The complex value of the current injection at bus k is given by
X
Ik = Vm Ykm ej(m +km ) (3.41)
mK
where
Ykm and km are the magnitude and angle of the complex element of ad-
mittance matrix in k-th row and m-th column.
The admittance matrix Y is usally very sparse but its size can be re-
duced by means of network reduction. There are several network reduction
techniques, one of the most common is application of Krons reduction for-
mula.
24 CHAPTER 3. MODELLING OF POWER SYSTEM
where
V = VV0 is the relative voltage magnitude at the load bus, P and Q are active
and reactive components of the load when the bus voltage magnitude is V ,
and the subscript 0 stands for their values at the initial operating point.
This model is composed of the following components:
constant power (does not vary with changes in the voltage magnitude)
P P
KD = = (3.51)
f r
where
P is the change of active power demand due to the change of the bus
frequency f or relative angular velocity r , which are equal in p.u. Since
the voltage frequency is a derivative of the voltage angle,
0
p = 0 r = P (3.52)
KD
26 CHAPTER 3. MODELLING OF POWER SYSTEM
X
P = PL PL0 = Vk Vm Ykm cos(k m km ) PL0 , (3.53)
mK
the differential equation for the load bus voltage angle can be rewritten as
0 X
pk = (Vk Vm Ykm cos(k m km ) PL0 ) (3.54)
KDk
mK
Hence, the load bus voltage angle becomes a state variable and its changes
are described by the differential equation (3.54).
I
q
Et
EI
eq r
d
ed
0
ER R
Figure 3.3: Reference frame transformation
ed = Et sin( ) (3.55)
eq = Et cos( ) (3.56)
00 1q 2q
Et sin( ) = Ra id + Xl iq Xaqs (iq + + ) (3.57)
X1q X2q
00 fd 1d
Et cos( ) = Ra iq Xl id + Xads (id + + ) (3.58)
Xfd X1d
Network equations (3.45 and 3.46) for the generator buses should be
rewritten considering
and thus
that results in
The network equations for the load buses should be adjusted to include the
static load characteristrics (3.47, 3.48). If the active power component of
the load demand is modelled by a constant current characteristic, and the
reactive power component is represented by a constant impedance, (3.47,
3.48) become
X
Pk0 ( VVk0 ) = Vk Vm Ykm cos(k m km )
k
mK
(3.64)
X
Q0k ( VVk0 )2 = Vk Vm Ykm sin(k m km )
k
mK
Thus, for the purpose of stability analysis, a power system can be mod-
elled by 10 ng differential equations and 4 ng + 2 nL algebraic equations,
where ng is the number of the generator buses, and nL is the number of the
load buses. If the load damping modelling approach described in Section
3.3.2 is adopted, the number of differential equations becomes 10 ng + nL ,
whereas the number of the algebraic equations is reduced to 4 ng + nL .
The system equations are expressed in terms of the state variables, the
generator currents id and iq , the complex bus voltages with the magnitude
Vk and angle k , and the parameters of the system components.
For the small-signal stability analysis, the system equations should be
linearized to take the form of Equations (2.4). The results of the linearization
are presented in Section 3.4.1. The transient stability analysis by means of
the formulated system equations is shortly discussed in 3.4.2.
No Load Damping
Linearization of the system equations results in the following set of equations
expressed in terms of the perturbed variables:
x = Ax + F1 Ig + F2 Vg + Bu (3.65)
0 = C1 x + G1 Ig + G2 Vg (3.66)
0 = C2 x + G3 Ig + G4 Vg + G5 VL (3.67)
0 = G6 Vg + G7 VL + Du (3.68)
Apart from the differential equations (3.65), this system includes stator volt-
age equations (3.66), generator bus network equations (3.67), and load bus
3.4. OVERALL SYSTEM EQUATIONS 29
network equations (3.68). The state vector is composed by the state vectors
of ng synchronous machines:
x1
x2
x = . (3.69)
..
xng
Ag1 0 0
..
0 Ag2 0 .
A= .
.
(3.70)
.
. 0 . . 0
0 0 Agng
The non-zero entries of each Agi matrix are expressed in terms of the
30 CHAPTER 3. MODELLING OF POWER SYSTEM
C1 and C2 are block diagonal matrices with the block elements C1g and
C2g respectively, where
" X 00 X 00 #
V cos( ) 0 0 0 Xaqs Xaqs 0 0 0 0
C1g = X 00 00
Xads
1q 2q
id V cos( ) iq V sin( ) 0 0 0 0 0 0 0 0 0
C2g =
id V sin( ) iq V cos( ) 0 0 0 0 0 0 0 0 0
(3.73)
In (3.72) and (3.73), V and denote the terminal voltage magnitude and
angle of the generator in question.
The generator current vector Ig is given by 1
id1
iq1
id2
Ig = iq2 (3.74)
..
.
idn
g
iqng
The matrices F1 , G1 , and G3 , that all also have a block diagonal structure,
are comprised by individual generator matrices of the form
0 0
1 (X 00 iq aq ) 1 (ad + Xaqs 00 i )
d
M ads M
Rfd 00
0 X
Xfd ads 0
R1d 00
0 X X 0
1d ads
R1q 00
0 0 X1q Xaqs
F1g = (3.75)
R 00
0 0 X2q X
2q aqs
0 0
(2,1) T1 (2,2)
KSTAB F1g1 F
T2 1g1
T (8,1) T (8,2)
T2 F1g1 T2 F1g1
1 1
0 0
00
Ra Xl + Xaqs V sin( ) V cos( )
G1g = 00 G3g = g
Xl Xads Ra Vg cos( ) V sin( )
(3.76)
The elements of the generator voltage vector Vg are the voltage angles
1
The notation for the generator current vector and the voltage vectors was adapted
from [15], whereas in other sources (e.g.[2]) the preference is given to the real and imaginary
components of current and voltage.
32 CHAPTER 3. MODELLING OF POWER SYSTEM
(2k1,2m1)
G4 = Vk Vm Ykm sin(k m km )
(2k1,2m)
G4 = Vk Ykm cos(k m km )
(3.81)
(2k,2m1)
G4 = Vk Vm Ykm cos(k m km )
(2k1,2m)
G4 = Vk Ykm sin(k m km )
3.4. OVERALL SYSTEM EQUATIONS 33
The diagonal entries of the matrix are defined by the following expressions:
(2k1,2k1)
G4 = idk Vk cos(k k ) + iqk Vk sin(k k )
X
Vk Vm Ykm sin(k m km )
mK
(2k1,2k)
G4 = idk sin(k k ) + iqk cos(k k )
X
Vm Ykm cos(k m km )
mK
(3.82)
(2k,2k1)
G4 = idk Vk sin(k k ) + iqk Vk cos(k k )
X
Vk Vm Ykm cos(k m km )
mK
(2k,2k)
G4 = idk cos(k k ) iqk sin(k k )
X
Vm Ykm sin(k m km )
mK
(2k1,2k1)
X
G7 = Vk Vm Ykm sin(k m km )
mK
(2k1,2k)
X
dPLk
G7 = dVk Vm Ykm cos(k m km )
mK
(3.83)
(2k,2k1)
X
G7 = Vk Vm Ykm cos k m km )
mK
(2k,2k) dQLk
X
G7 = dVk Vm Ykm sin(k m km )
mK
dQLk
where dP
dVk and dVk are the sensitivities of the static load characteristics to
Lk
the voltage at the corresponding load bus. With the constant current and
the constant impedance characteristics for the active and reactive power
34 CHAPTER 3. MODELLING OF POWER SYSTEM
dPLk
= Pk0 /Vk0 (3.84)
dVk
dQLk
= 2Q0k Vk /(Vk0 )2 (3.85)
dVk
The input vector u may contain different quantities, e.g. the turbine
setpoint changes Pm set . Another example would be u as a vector of
the load changes PL set which explains the appearance of the D matrix in
(3.68). Since the load power demand is not explicitly given in (3.65), the last
term of (3.65) should disappear in this case. However, u can be defined
as
Pm set
u = (3.86)
PL set
which yields non-zero elements in the both matrices B and D.
To enable the small-signal analysis, the system state matrix AS should
be derived. The successive elimination of Vl and Ig while assuming
u = 0 will give
x = A0 x + F 0 Vg (3.87)
0 0
0 = C x + G Vg (3.88)
where
A0 = A F1 G1 1 C1 (3.89)
0 1
F = F2 F1 G1 G2 (3.90)
0 1
C = C2 G3 G1 C1 (3.91)
0 1 1
G = G4 G3 G1 G2 G5 G7 G6 (3.92)
1
Vg = G0 C 0 x (3.93)
Thus
AS = A0 F 0 G01 C 0 (3.95)
x = Ax + F1 Ig + F2 Vg + F3 VL + Bu (3.98)
0 = C1 x + G1 Ig + G2 Vg (3.99)
0 = C2 x + G3 Ig + G4 Vg + G5 VL (3.100)
0 = C3Q x + G6Q Vg + G7Q VL (3.101)
The matrices A and F2 should be extended to account for the new differ-
ential equations. The block AL contains nL nL elements of the following
form
(k,m) 0
AL = Vk Vm Ykm sin(k m km ) (3.102)
KDk
where k = ng + 1, ng + 2, ..., ng + nL , m = ng + 1, ng + 2, ..., ng + nL and
m 6= k
and
(k,k) 0 X
AL = Vk Vm Ykm sin(k m km ) (3.103)
KDk
mK
where m = 1, 2, ..., ng + nL
AL should be added at the diagonal of A (3.70),
The matrices F2 and F3 should be augmented with nL additional rows
each. The elements of these rows are as follows:
(k,2m1) 0
F2L = Vk Vm Ykm sin(k m km ) (3.104)
KDk
(k,2m) 0
F2L = Vk Ykm cos(k m km ) (3.105)
KDk
where k = ng + 1, ng + 2, ..., ng + nL ,m = 1, 2, ..., ng The first 10ng nL
entries of the new matrix F3 are zeros, since the derivatives of the generator
36 CHAPTER 3. MODELLING OF POWER SYSTEM
with m = 1, 2, ..., ng + nL
Besides, nL rows with zero entries should be added to the matrix F1 .
Now the system state matrix can be computed by using a similar ap-
proach as in the no damping case:
x = f (x, Ig , V ) (3.116)
0 = g(x, Ig , V ) (3.117)
5. To find the values of the state variables x at the next time instant, R-
K numerical integration method can be applied (see Appendix A) by
using the time derivative values from the previous step in the formulae
(A.2).
6. The steps 3-5 should be repeated to obtain the time response of the
system to the given disturbance. If the fault is assumed to be cleared
at some time point, the admittance matrix will be changed again, and
the network variables will change stepwise, but not the state variables.
According to [2], the advantages of this approach are its flexibility, sim-
plicity, reliability, and robustness. However, it might become numerically
unstable if the time step is bigger than the time constant of the smallest
eigenvalue.
38 CHAPTER 3. MODELLING OF POWER SYSTEM
Chapter 4
In this chapter, sensitivities of the system state matrix to inertia and damp-
ing are derived. They are further implemented in computation of damping
ratio sensitivities and sensitivities of transient frequency overshoot based on
[1]. In the following sections, an algorithm for inertia and damping opti-
mization and its implementation in MATLAB are presented.
where
(i) (i)
= Re( ) (4.2)
39
40 CHAPTER 4. IMPACT OF INERTIA AND DAMPING
(i) (i)
= Im( ) (4.3)
(i) AS (i)
= (i)T (4.4)
Thus, to find the sensitivities of the damping ratios to the inertia and
damping coefficients, the expressions for the sensitivity of the system state
matrix should be derived.
No Load Damping
The sensitivity of the system state matrix to the inertia of one of the
syncronous machines, M , is determined by
AS A0 F 0 0 01 0 A0 F 0 0 01 0
= F G C = F G C (4.5)
M M M M M
with
A0 A F1
= (A F1 G1 1 C1 ) = G1 1 C1 (4.6)
M M M M
F 0 F1
= (F2 F1 G1 1 G2 ) = G1 1 G2 (4.7)
M M M
C 0
= 0 (4.8)
M
G0
= 0 (4.9)
M
A
In (4.6), M and F
M are block diagonal matrices. However, their only
1
blocks that contain non-zero elements are the ones that correspond to the
generator in question, since the inertia level of a generator does not explicitly
influence the state variables of the other generators. The non-zero elements
4.1. SENSITIVITY OF DAMPING RATIO 41
Ag
of the submatrix M can be derived from (3.71) that yields
A0 A F1
= G3 1 C4 (4.12)
M M M
F 0 F1
= G3 1 G8 (4.13)
M M
The non-zero elements of M A
and F
M can be determined by means of
1
where m = 1, 2, ..., ng + nL
F2 (9ng +k,2m1) 0
= 2 Vk Vm Ykm sin(k m km ) (4.21)
KDk KDk
F2 (9ng +k,2m) 0
= 2 Vk Ykm cos(k m km ) (4.22)
KDk KDk
F3
Finally, non-zero entries of KDk are defined by
F3 (9ng +k,m) 0
= 2 Vk Ykm cos(k m km ) (4.23)
KDk KDk
F3 (9ng +k,k) 0 X
= 2 Vk Ykm cos(k m km ) (4.24)
KDk KDk
mK
where m = 1, 2, ..., ng + nL
(k) X (k) X
= ckj (j) , = dkj (j) (4.25)
jN jN
44 CHAPTER 4. IMPACT OF INERTIA AND DAMPING
where N is the set of the system modes, and the off-diagonal elements ckj
and dkj can be expressed as
(j)T A
S (k)
ckj = k 6= j (4.26)
((k) (j) ) (j)T (j)
(k)T A
S (j)
dkj = k 6= j (4.27)
((k) (j) ) (j)T (j)
x = AS x + bk u (4.30)
y = cl x (4.31)
where cl is a matrix mapping the frequency of node l on the output ylk and
bk shows the contribution of a disturbance at the k-th node to the deviation
of the state variables. Hence, the transfer function G(s) can be expressed
as
Y (s)
G(s) = = cl (sI A)(1) bk
U (s)
= cl (sI )(1) T bk (4.32)
X Rk
li
= (i)
iN
s
The matrix bk is not as easily derived as in [1] because in this thesis, a more
complex model of a power system is employed. The derivation of bk will be
covered later in this section.
4.2. SENSITIVITY OF TRANSIENT OVERSHOOT 45
k
Rli
With the short-hand Klik = (i)
, the step response Y (s) is given by
1 X (i)
Ylk (s) = G(s) = Klik (4.34)
s s(s (i) )
iN
(i) t
X
ylk (t) = L1 [Ylk (s)] = Klik (1 e ) (4.35)
iN
After dividing the eigenvalues into real and complex conjugate, represented
by the sets 0 and Lambda+ , respectively, the time-domain response can be
rewritten as
(i) t (i)t
X X
ylk (t) = Klik (1 e )2 (KlikRe e ||Klik || sin( (i) t lik ))
i0 i+
(4.36)
with
lik = arctan(KlikRe , KlikIm ) (4.37)
i = argmaxKljk (4.38)
j
The first peak time and magnitude of the dominating mode i could be ap-
proximated by
1
tkpl = (0.5 lik ) (4.39)
(i)
k
Mpl = ylk (tkpl ) (4.40)
l to inertia and
The next step would be to estimate the sensitivity of Mpk
damping. The derivative of the transfer function residue Rlik to inertia or
damping of the j-th generator is given by
x = Ax + bk Pk (4.47)
r = cl x (4.48)
x = Ax + F1 Ig + F2 Vg (4.49)
0 = C1 x + G1 Ig + G2 Vg (4.50)
0 = C2 x + G3 Ig + G4 Vg + G5 VL (4.51)
k
0 = G6 Vg + G7 VL + d Pk (4.52)
VL = G7 1 (G6 Vg + dk Pk ) (4.53)
1
Ig = G1 (C1 x + G2 Vg ) (4.54)
x = A0 x + F 0 Vg (4.55)
0
0 = C x + G Vg 0
G5 G1 k
7 d Pk (4.56)
4.2. SENSITIVITY OF TRANSIENT OVERSHOOT 47
x = AS x + F 0 G01 G5 G1 k
7 d Pk (4.57)
bk = F 0 G01 G5 G1
7 d
k
(4.58)
dk(2k1,2k1) = 1 (4.59)
x = A0 x + F 0 Vg (4.60)
0 0 k
0 = C x + G Vg d Pk (4.61)
that yields
x = AS x + F 0 G01 dk Pk (4.62)
and
bk = F 0 G01 dk (4.63)
bk bk
Now Mj and KD should be calculated to be used in (4.41)
bk F 0 01
= G G5 G1
7 d
k
(4.64)
Mj Mj
bk F 0 01 k
= G d (4.65)
Mj Mj
bk
= 0 (4.66)
KD
x = AS x + bk Pk (4.67)
48 CHAPTER 4. IMPACT OF INERTIA AND DAMPING
bk
= 0 (4.69)
Mj
(10ng +k,k)
bk 0
= 2 (4.70)
KD KDk
bk F 0 01 k
= G d (4.71)
Mj Mj
bk F 0 01 k
= G d (4.72)
KD KD
F0 0
F
with M j
and K D
as defined by (4.18).
By comparing Equations (4.64) and (4.71), it could be concluded that
the load damping model, suggested in Section 3.3.2, does not adequately
reflect the sensitivity of the generator frequency response to the rotational
inertia for the load bus case.
The expressions for bk and its derivatives can be used to calculate the
transient overshoot Mpl k and its sensitivities.
Knowing how the changes of inertia and damping affect the eigenvalues
of the system state matrix and the magnitude of the transient overshoot,
it is now possible to formulate an optimization program focused on the
improvement of the system stability.
associated with a cost. The sets K and M include the nodes with damping
and inertia, respectively. The total amount of inertia and damping that can
be added is assumed to be limited.
The sensitivities of the damping ratios and the overshoot are non-linear,
therefore, each step of the optimization will be associated with a solution
of a linearized problem. Superscript denotes the number of the iteration
with 0 being the first iteration. After each iteration, the system state matrix,
along with the sensitivities of damping ratios and transient overshoot should
be computed all over again.
The change of the damping from the current iteration to the next is given
by
+1 +1
KDj = KDj KDj (4.73)
A similar expression is applied to the value of inertia. To enable the calcula-
+1
tion of the absolute change in inertia and damping, KDj and Mj+1 should
be split in positive parts, KDj+
and Mj+ , and negative parts,KDj and Mj ,
as
(
+1 0 +1 0 >0
+ KDj KDj if KDj KDj
KDj = +1 0 0 (4.74)
0 if KDj KDj
(
+1 0 >0
0 if KDj KDj
KDj = +1 0 | if K +1 K 0 0 (4.75)
|KDj KDj Dj Dj
The objective function penalizes the minimal damping ratio with cost
c 0. The purpose of the slack variable kpl is to ensure the feasibility of
the problem, and it is penalized by the cost c . Procurement of additional
inertia and damping implicates economic costs cMi and cKi . However, the
accurate calculation of cMj and cKj would be a complex task without a
well-developed inertia and damping procurement market.
XX X X
min [c min + (c kpl ) + cKj |KDj | + cMj |Mj |] i +
KDj ,Mj
k l iK iM
(4.80)
50 CHAPTER 4. IMPACT OF INERTIA AND DAMPING
s.t.
X X
i+1 = i + i +1
KDj + i
Mj+1 (4.81)
KDj Mj
jK jM
min
i+1 (4.82)
X
tot
|KDj | KD (4.83)
jK
X
|Mj | M tot (4.84)
jM
min +1 max
KDj KDj KDj (4.85)
Mjmin Mj+1 Mjmax (4.86)
min +1 max
KDj KDj KDj (4.87)
Mjmin Mj+1 Mjmax (4.88)
+
0 KDj , 0 KDj (4.89)
+ +1 0
KDj KDj = KDj KDj (4.90)
0 Mj+ , 0 Mj (4.91)
Mj+ Mj = Mj+1 Mj0 (4.92)
k
X Mpl k
X Mpl
k
fp Mpl Pk Pk Mj+1 + +1
Pk KDj kpl (4.93)
Mj KDj
jM jK
With (4.81), the value of the damping ratios are computed at the step + 1
using the previous value i and the changes related to the adjustment of Mj
and KDj . The constraint (4.82) serves to set min to the lowest damping
ratio value. The inequality constraints (4.83) and (4.84) limit the total
change of inertia and damping to the values available for procurement. The
individual values of Mj and KDj at each bus are limited by (4.85) and
k computed at each iteration are valid only
(4.86). Sensitivities of and Mpl
for a small range of values around the initial Mj and KDj . Therefore, the
+1 +1
steps KDj and Mj should be limited by (4.87) and (4.88). Equalities
(4.89-4.92) split KDj and Mj as in (4.78 and 4.79).
The magnitude of the transient overshoot after a disturbance Pk is
given by Mpl k P . The change of inertia and damping by K +1 and
k Dj
+1
Mj results in additional terms of the overshoot as shown in (4.93).
The total magnitude of the overshoot should not exceed the limit fp .
At each iteration , the program finds the optimal levels of damping and
inertia. These values are used to calculate the system state matrix and its
eigenvalues, the new values of sensitivities, and the approximate magnitude
4.4. IMPLEMENTATION IN MATLAB 51
Optimization Program
START
DATA.mat
Calculation of the
initial state of the
generators
1
NO
Estimation of Calculation of Computation of
Optimization of
the transient the eigenvalue the system state
min (dm;dk) <0,1 the inertia and
overshoot and and eigenvector matrix and
damping levels
sensitivities sensitivities eigenvalues
YES
END
The flow chart of the developed MATLAB program for inertia and damp-
ing optimization is presented in Figure 4.1. The input data for the optimiza-
tion program should be saved in the file DATA.mat and should include the
following arrays:
Transient Simulations
The master file of the transient simulations is called run transient. It re-
quires the same input data as the optimization program and starts the com-
putations by calling reindexation.m, pf.m, and kron.m. The next step is
estimation of the initial values of the generator variables in initial x gen.m.
Following this, the bus and the branch where a disturbance occurs is speci-
fied, and the admittance matrix of the system is altered to involve a shunt
element at the faulty bus. The new values of the network variables (voltages
and currents) are calculated at the next step by means of Newton-Raphson
iteration algorithm implemented in newtraph.m. The latter has two sub-
functions, algebraic.m, where the right-side parts of the network equations
(3.117) are evaluated, and jacobian.m computing the Jacobian associated
with these equations.
After that, the time step and the number of the time intervals before
the fault is cleared are specified. For each time interval, the values of the
state variables are computed by means of numerical integration, while the
values of the network variables are estimated in newtraph.m. Numerical
integration is carried out by the second order R-K method (rungekutta2.m)
54 CHAPTER 4. IMPACT OF INERTIA AND DAMPING
Simulation Results
55
56 CHAPTER 5. SIMULATION RESULTS
400 MW
7 8 9
G1 1 5 6 10 11 3 G3
110 km 110 km
25 km 10 km 10 km 25 km
C7 C9
L7 L9
2 4
G2 G4
Area 1 Area 2
values differ from the ones presented in [2]. This discrepancy could be ex-
plained by the difference between the implemented PSS model and a more
detailed model chosen by the author of [2].
Implementation of PFC introduces additional damping of the oscilla-
tions. Speed governors were not included into the model used for the small-
signal stability analysis of the two-area system in [2], therefore, the param-
eters of PFC were chosen at our own discretion as follows:
Droop S = 2%
The value of the turbine time constant is set in accordance with the maximal
time of full primary control reserve deployment allowed in interconnected
European power system which is 30 s. Normally, the values of Tt of non-
reheat steam turbines are significantly lower than 10 s, but to investigate
the worst-case scenario, the chosen value of Tt represents steam turbines,
equipped with a re-heater.
The eigenvalues of the state matrix of the given two-area system with
AVR, PSS, and PFC are presented in Table 5.3. In the case that will be
further referred to as Base Case, the rotational inertia and damping coeffi-
cients of the machines are set to the values given in Table 5.2 in accordance
with Example 12.6 of [2]. This case represents a system with conventional
generation, and thus conventional level of rotational inertia.
It should be noted, that the small-signal stability analysis of the test
system in Load Damping case, i.e. with incorporation of the frequency
dependency of the load modelled as in Section 3.3.2, yielded positive eigen-
values. This result contradicts the expectations from the effect of the load
damping on the stability of the system. Modelling of an aggregated load is
a complex task since it requires an adequate reflection of both voltage and
frequency dependency of the consumed power and due to the diversity of
the consumer devices. The model proposed in Section 3.3.2 does not seem
to offer an appropriate description of the voltage dependence of the active
component of demanded power and, thus, it should be further elaborated. It
is not included in the optimization analysis conducted in the present work.
In case of the high penetration of RES, the level of inertia is significantly
lower. For example, according to [5], in 2012, the share of inverter-connected
RES infeed in the German power system has reached maximal value of 50%.
Consequently, the aggregated inertia of the system lost half of its value
during the times with such a high RES share, changing from H = 6s to
H = 3 4s (or from M = 12s to M = 6 8s). This highly reduced inertia
scenario is reflected in the present thesis by Low-Inertia Case with all the
inertia constants reduced by 50% compared to Base Case.
5.1. IEEE TWO-AREA TEST SYSTEM 59
Generator G1 G2 G3 G4
KD 1 1 1 1
M base case [s] 13 13 12.35 12.35
M low inertia [s] 6.5 6.5 6.175 6.175
60 CHAPTER 5. SIMULATION RESULTS
Table 5.3: Eigenvalues of the two-area system in Base Case (left) and Low-
Inertia Case (right).
0.05
0.1
Transient Frequency [Hz]
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0 20 40 60 80 100
Time [s]
0.05
0.1
Transient Frequency [Hz]
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0 20 40 60 80 100
Time [s]
Another parameter that affects the value of the frequency nadir is the
damping coefficient KD . As already discussed in Chapter 3, KD represents
the relation of the electrical torque at the rotor of a generator to the fre-
quency deviation. From Equation (3.6), it is clear that the higher is KD ,
the smaller is the rate of the frequency deviation.The effect of the reduced
damping at the generator buses could be seen in Table 5.6 and Figure 5.4.
64 CHAPTER 5. SIMULATION RESULTS
0.05
0.1
Transient Frequency [Hz]
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0 20 40 60 80 100
Time [s]
If both the rotational inertia and damping coefficients of the machines are
reduced by 50% compared to Base Case, the absolute value of the transient
frequency overshoot increases by 15-17%, as shown in Table 5.7.
5.1. IEEE TWO-AREA TEST SYSTEM 65
0 0
0.05 0.05
0.1 0.1
Transient Frequency [Hz]
0.15 0.15
0.2 0.2
0.25 0.25
0.3 0.3
0.35 0.35
0.4 0.4
base case base case
0.45 0.45
lowinertia case lowinertia case
low inertia and damping low inertia and damping
0.5 0.5
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]
5.1.3 Optimization
The optimization of inertia and damping in the two-area system has been
accomplished in several stages on a simple-to-complex basis. The parame-
ters of the optimization in the investigated cases are presented in Tables 5.8
and 5.9.
Table 5.8: Parameters of the optimization program for two-area test system
(Case 1 - Case 4)
Table 5.9: Parameters of the optimization program for two-area test system
(Case 5 - Case 8)
0min 0.2074
min
opt 0.4024
Number of Iterations 90
Improvement 94%
Mpmin , [Hz] -0.258
The results of Case 1 implicate that a high share of RES should be seen
as a positive condition for the stability of the two-area test system with
5.1. IEEE TWO-AREA TEST SYSTEM 69
0min 0.2074
min
opt 0.3719
Number of Iterations 62
Improvement 79%
Mpmin , [Hz] -0.398
The minimal damping ratio grew by 79% with respect to Base Case. At
the same time, the maximal magnitude of the transient frequency overshoot
increased by roughly 14%.
70 CHAPTER 5. SIMULATION RESULTS
0min 0.2074
min
opt 0.4030
Number of Iterations 61
Improvement 94%
Mpmin , [Hz] -0.294
Case 4 To check how the availability of inertia at bus 1 affects the opti-
mal solution, Case 4 was designed, with the total available inertia at bus 1
reduced by 50% to 2M1base . The results of the Case 4 simulations can be
found in Tables 5.16 and 5.17.
0min 0.2074
min
opt 0.2829
Number of Iterations 47
Improvement 36%
Mpmin , [Hz] -0.312
The restriction imposed on the bus 1 inertia reserves reduces the im-
provement of the minimal damping ratio from 94% to 36%. The solution
is obtained by deploying the whole available inertia reserve at bus 1 and
reducing the rotational inertia at buses 3 and 4.
0min 0.2074
min
opt 0.3832
Number of Iterations 54
Improvement 85%
Mpmin , [Hz] -0.312
The obtained results (see Tables 5.18 and 5.19) are similar to those of
Case 4. The inertia of G1 is increased by 162%, while it is set to the minimal
value at bus 3 and close to the minimal value at bus 4. However, the inertia
level at bus 2 remains intact which can be seen as a reaction to the costs
imposed on the changes.
0min 0.2075
min
opt 0.3796
Number of Iterations 88
Improvement 83%
Mpmin , [Hz] -0.252
5.1. IEEE TWO-AREA TEST SYSTEM 73
Case 7 The optimization results in Case 7 are presented in Tables 5.22 and
5.23. The following parameters were used in this case: cMj = 0.015, cKj =
0.01, fp = 0.312 Hz. Since the absolute value of fp is smaller than the
magnitude of the frequency overshoot in Base Case, the slack variable was
introduced to ensure the feasibility of the frequency overshoot constraint.
The costs c associated with the slack variable were set at the value c = 15.
0min 0.2074
min
opt 0.3768
Number of Iterations 88
Improvement 82%
Mpmin , [Hz] -0.294
Case 8 The last optimization case for the considered test system, Case 8,
is based on Low-Inertia Case. The initial values of the rotational inertia are
already reduced by 50%, compared to the previous optimization cases. As
can be seen in Tables 5.24 and 5.25, improvement of the minimal damping
74 CHAPTER 5. SIMULATION RESULTS
0min 0.2926
min
opt 0.3840
Number of Iterations 190
Improvement 38%
Mpmin , [Hz] -0.312
KDj = 1 (Base Case, Low-Inertia Case, Case 2) and those with improved
damping KDj = 4 (Low Inertia and Low Damping, Case 1). The peaks of
the first group lines lay considerably higher than those of the second group
lines. This stands as a clear illustration of the influence of damping on the
transient frequency in the investigated system.
0.18
0.16
Transient Frequency [Hz]
0.14
0.12
0.1
0.08
0.06
Among the cases of the first group, Base Case implicates the highest val-
ues of the rotational inertia and, thus, demonstrates the strongest inertial
response. Therefore, as illustrated by the blue line in the graph, Base Case is
associated with the slowest rise and the smallest magnitude of the transient
frequency compared to the other cases in the group. Case 2, character-
ized by very low levels of inertia and an improved minimal damping ratio,
demonstrates the highest rate and magnitude of the frequency deviation.
However, the difference between the peak values of the transient frequency
in Base Case and Case 2 is merely 0.01 Hz which is 6.25% of the magnitude
in Base Case.
Figure 5.6 does not give a clear impression on the transients occuring
right after the short circuit. The transient frequency within the first 5
seconds after the fault is shown in Figure 5.7. This graph demonstrates that
the lower is the inertia, the steeper is the growth of the frequency after the
76 CHAPTER 5. SIMULATION RESULTS
0.12
0.1
Transient Frequency [Hz]
0.08
0.06
0.04
base case
0.02 lowinertia case
low inertia and high damping
case 1
case 2
0
0 1 2 3 4 5
Time [s]
The effect of the inertial response on the critical fault clearing time can
be observed by comparing two plots in Figure 5.8. Both plots represent the
rotor angles of the four generators after the considered disturbance. The
fault was cleared in both cases after 0.012 s, but the rotational inertia of the
generators in the left graph was set to the values of Base Case, whereas at
the right, the same event within Low-Inertia Case is modelled. With inertia
values of Base Case, the system remains stable. In contrast, in Low-Inertia
Case, the generators G3 and G4 fall out of step, which implicates islanding
of the two areas.
5.2. IEEE SOUTH EAST AUSTRALIAN TEST SYSTEM 77
0.8 8000
7000
0.6
Relative Rotor Angle [rad]
0.2 4000
0 3000
2000
0.2
1000
G1 G1
0.4 G2 G2
0
G3 G3
G4 G4
0.6 1000
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Time [s] Time [s]
Figure 5.8: Rotor angles of the generators G1-G4 of the two-area test system
after a short circuit at bus 9 in Base Case (left) and Low-Inertia Case (right)
Generator G1 G2 G3 G4 G5 G6 G7
Node 101 201 202 203 204 301 302
KD 9.999 33.335 22.224 16.668 26.668 53.336 17.776
M Base [s] 71.993 213.344 124.454 106.675 138.674 298.682 124.432
M Low Inertia [s] 35.996 106.672 62.227 53.338 69.337 149.341 62.216
Generator G8 G9 G10 G11 G12 G13 G14
Node 401 402 403 404 501 502 503
KD 17.776 9.999 17.776 19.998 6.666 10.000 8.335
M Base [s] 106.656 79.992 92.435 103.990 46.662 80.000 125.025
M Low Inertia [s] 53.328 39.996 46.218 51.995 23.331 40.000 62.513
five-area system has a very low minimal damping ratio, basemin = 0.096, as
could be seen in Table E.4. Furthermore, the system exhibits three more
oscillatory modes with damping ratios lower than 0.200. Thus, oscillations
occuring after small disturbances in this system decay at a very slow rate,
and changes of the operating state of the system might lead to its instability.
In Low-Inertia Case, the system is unstable with an expanding oscillatory
5.2. IEEE SOUTH EAST AUSTRALIAN TEST SYSTEM 79
mode = 1.812 39.368 (see Table E.5). This might cause considerable
concerns, if the system has a high installed capacity of RES.
Frequency response of the five-area test system was estimated in the
same way as it was done for the two-area system, i.e. by applying steplike
load changes to the generator buses. The results of the transient overshoot
approximation are shown in Table 5.28. The magnitude of the overshoot
depends on the node where disturbance occured. As could be seen in Ta-
ble 5.28, disturbances that take place in the same area cause frequency
deviations of the same amplitude. The magnitude of the overshoot has the
highest value in Area 5 and decreases in accordance with the power flow
direction.
5.2.3 Optimization
The optimization was carried out for three cases with different parame-
ters presented in Table 5.29.
Table 5.29: Parameters of the optimization program for the five-area test
system
0min 0.0960
min
opt 0.1743
Number of Iterations 185
Improvement 82%
Mpmin , [Hz] -0.047
Table 5.32: Optimization results for the five-area test system (Case 2)
0min 0.0960
min
opt 0.1511
Number of Iterations 150
Improvement 58%
Mpmin , [Hz] -0.049
Table 5.34: Optimization results for the five-area test system (Case 3)
0min 0.0960
min
opt 0.1590
Number of Iterations 150
Improvement 66%
Mpmin , [Hz] -0.045
0.5
0.02
0
0.5 0.03
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 5.9: Rotor angular velocity of the generators of the five-area test
system after a short circuit at bus 217 and disconnection of a circuit of the
line 217-215 in Base Case (left) and Case 1 (right)
Generator G1 has the electrically closest location to the faulty bus. The
fault leads to a large power imbalance at the rotor of this generator, with
mechanical power exceeding the power that can be transmitted from the
generator to other areas or consumed by the load at bus 102. In Case 1,
with inertia of G1 reduced by 60%, the disturnance leads to the acceleration
of G1 and, consequently, this generator loses synchronism with the grid.
This illustrates the importance of the transient frequency restriction for the
inertia optimization.
5.3. DISCUSSION OF SIMULATION RESULTS 85
Case analysis has demonstrated that damping of the critical oscillatory mode
in the IEEE two-area test system is significantly improved at reduced levels
of inertia. Thus, the minimal damping ratio could be increased by 79%
compared to the convenitional Base Case by reducing the inertia constants
by 65%-75%. However, a further reduction of the inertia would not lead
to any advancements which indicates that there is a specific level of the
RES penetration in the test system, optimal for damping of the interarea
oscillatory mode.
However, high shares of RES generation and low inertia levels raise a
common concern over the transient frequency response of the system. The
transient simulations have shown that the magnitude of the frequency over-
shoot after a short circuit greatly depends on the inertia levels in the system,
as it is defined by the inertial response of the synchronous machines. Rota-
tional inertia levels may become crucial for the system stability because low
inertia in the system leads to reduction of the critical fault clearing time.
The results of Case 3 and Case 5 show that the frequency overshoot
in the test system can also be limited by a considerable increase of the
rotational inertia at bus 1. Procurement of inertia at this bus could be
seen as an effective measure of securing an acceptable level of the frequency
overshoot. However, if the limitation imposed on the frequency magnitude
is too strict, it might become too expensive to comply with it by increasing
inertia solely. Thus, procurement of additional damping would become an
appropriate measure to limit the frequency deviations that develop in the
first minute after a large disturbance.
0.05
0.1
Transient Frequency [Hz]
0.15
0.2
0.25
0.3
0.35
0.4 Tt = 1s
0.45 Tt = 3s
Tt = 10s
0.5
0 20 40 60 80 100
Time [s]
Such a fast PFC could be provided by BESS [7]. The damping procure-
ment might be also accomplished by means of BESS [20]. However, if the
storage systems are used for damping, it requires from them an immedi-
ate reaction to any oscillations of the frequency, whereas PFC reserves are
activated outside of a dead-band around the nominal frequency. In Conti-
nental Europe, this dead-band is 10mHz [7]. The potential of the battery
systems in provision of these two services simultaneously should be further
investigated.
IEEE South East Australian test system has demonstrated very poor
damping of the oscillatory modes. Improvement of the minimal damping
ratio by 82% can be achieved by inertia and damping adjustment. The
solution implicates increased level of inertia at particular buses which does
not meet the expectations that high inertia would worsen the damping.
Furthermore, in Low-Inertia Case, with inertia constant reduced by 50%,
the test system becomes unstable.
The discrepancy between the results and the expectations can be orig-
inated in the simplicity of the employed AVR and PSS models and lack of
tuning of thier parameters. Alternative models of the control devices were
5.3. DISCUSSION OF SIMULATION RESULTS 87
89
90 CHAPTER 6. CONCLUSIONS AND OUTLOOK
case study was conducted for two test systems, IEEE two-area test system
and IEEE South East Australian test system. The case analysis of the for-
mer system included eight cases with different optimization parameters. It
was shown, that a reduction of rotational inertia significantly improves the
damping of the oscillatory modes. However, a simultaneous limitation of
the transient frequency deviations required the provision of additional iner-
tia and damping at particular nodes. Introduction of costs for inertia and
damping procurement affected the optimal solution by leaving only the most
effective changes of inertia and damping. The influence of PFC speed on
the frequency nadir was shown. The impact of inertia on frequency devi-
ations after a severe disturbance was illustrated by the results of transient
simulations implemented in MATLAB.
The minimal damping ratio of IEEE South East Australian test system
was optimized in three different cases. As an optimal solution in case of
non-zero costs, higher levels of rotational inertia and damping at particular
system nodes are suggested. However, the modelling accuracy in case of
this system should be improved by incorporation of realistic models and
parameters of AVR and PSS.
Detailed modelling of AVR, PSS, and turbine governors can be proposed
as as an objective for the future work. Implementation of the RES models
is another possible enhancement of the modelling framework. Furthermore,
the algorithm could be adjusted to consider the provision of inertia and
damping at buses with no generation.
Another possible field of research is the economical aspects of inertia and
damping procurement. Economical factors clearly have a great influence on
the optimal inertia and damping levels. Furthermore, pricing at potential
inertia-as-a-service markets requires an adequate estimation of the econom-
ical losses due to complications in system operation and outages caused by
a certain inertia level.
If the focus of analysis is to be shifted from the maximization of the
damping of oscillations to minimization of the transient frequency overshoot
in a system with low inertia, the optimization program could be adjusted
respectively, by setting the frequency overshoot magnitude as a main ob-
jective while restraining the minimal damping ratio. This would allow to
optimize the operation of the system during a high RES penetration with
regard to the frequency response. As shown in the case analysis of the IEEE
two-area test system, with the optimization program focused on the mini-
mal damping ratio, the already low inertia levels are proposed to be further
reduced. To find a proper balance between two objectives, the consequences
of both poor oscillatory mode damping and high frequency overshoot should
be evaluated for each particular power system.
Appendix A
Runge-Kutta Methods of
Numerical Integration
R-K methods used for numerical integration in the present work were pro-
posed for power system transient simulations in [2]. Depending on the
number of evaluations of the first derivative in Taylor series solution, R-
K methods of different orders could be used for this purpose. In this thesis,
the second order R-K method and Gills version of the fourth order R-K
method were implemented.
dx
= f (x, t)
dt
k1 = f (xn , tn )t
(A.2)
k2 = f (xn + k1 , tn + t)t
91
92 APPENDIX A. RUNGE-KUTTA METHODS
kj = aj [f (xj1 , t) bj qj1 ]
xj = xj1 + kj t (A.3)
qj = qj1 + 3kj cj f (xj1 , t)
Vt = V ej
It = Iej (B.1)
eq + Ra iq + Xds id
ifd = (B.7)
Xads
efd = Rfd ifd (B.8)
Xadu
Efd = (B.9)
Rfd efd
93
94 APPENDIX B. CALCULATION OF INITIAL STEADY STATE
Initial values of the flux linkages of the rotor circuits are defined by
r = 0 (B.16)
v1 = V (B.17)
v2 = 0 (B.18)
vs = 0 (B.19)
Pm = 0 (B.21)
Appendix C
Modelling of Transmission
Network Elements
With the sign convention from [3], the current injection from the shunt
is defined by
Iksh = yksh Ek (C.1)
95
96 APPENDIX C. TRANSMISSION NETWORK MODELLING
1
ykm = zkm = gkm + jbkm (C.4)
where
rkm
gkm = 2 (C.5)
+ x2km
rkm
xkm
bkm = 2 (C.6)
rkm + x2km
where akm is the turns ratio. For in-phase transformers, considered in this
thesis, tkm = akm which means that tkm R.
In [3], a unified branch model for lines, in-phase transformers, and phase-
shifting transformers was developed to facilitate the modelling routine. This
model is shown in Figure C.3.
97
The general expression for the branch current in this model is given by
Ikm = a2km (yk msh + ykm )Ek tkm tmk ykm Em (C.8)
Column Parameter
1 Bus number
2 Active power generation [p.u.]
3 Reactive power generation [p.u.]
4 Active power demand [p.u.]
5 Reactive power demand [p.u.]
6 Voltage magnitude [p.u.]
7 Damping coefficient
8 Shunt susceptance [p.u. injected at V=1.0 p.u.]
9 Slack bus = 1, Other buses =0
10 Area
11 Maximal reactive power [p.u.]
12 Minimal reactive power [p.u.]
13 Total MVA base of generator [p.u,]
14 Maximal active power [p.u.]
15 Base Voltage [kV]
99
100 APPENDIX D. STRUCTURE OF MATLAB INPUT ARRAYS
Column Parameter
1 Branch number
2 From bus number
3 To bus number
4 Transformer tap ratio
5 Branch resistance [p.u.]
6 Branch reactance [p.u.]
7 Total line charging susceptance [p.u.]
Row Parameter
1 Aggregated Generator No.
2 Node
3 Power rating [MVA]
4 Inertia constant H [s]
5 Number of generators
6 Number of generators on-line
7 Synchronous reactance Xd [p.u.]
8 Synchronous reactance Xq [p.u.]
9 Stator leakage inductance Xl [p.u.]
10 Stator resistance Ra [p.u.]
11 Transient reactance Xd0 [p.u.]
12 Transient reactance Xq0 [p.u.]
13 Subtransient reactance Xd00 [p.u.]
14 Subtransient reactance Xq00 [p.u.]
15 Transient OC time constanct Td0 0 [s]
Appendix E
Figure E.1: IEEE South East Australian five-area test system [4]
103
Table E.1: Power flow input data for IEEE South Australian test system [4]
calculated on 100 MVA base
Bus No. Pg [p.u.] Qg [p.u.] Pl [p.u.] Ql [p.u.] Vi [p.u.] Qshunt [p.u.] Area Base Voltage [kV]
101 4.788 1.632 0 0 1.000 0 1 15
102 0 0 -3.8 -0.38 0 1 330
201 28 1.945 0 0 1.000 0 2 20
202 19.2 2.42 0 0 1.000 0 2 20
203 13.5 2.472 0 0 1.000 0 2 20
204 15.84 0.712 0 0 1.000 0 2 20
205 0 0.418 -3.3 -0.33 1.055 0 2 330
206 0 0 -1.1 -0.11 0 2 330
207 0 0 -16 -1.6 0 2 330
208 0 0 -1.8 -0.18 0 2 330
209 0 0 0 0 0 2 330
210 0 0 0 0 0 2 500
211 0 0 -14.45 -1.45 0 2 330
212 0 0 -14.1 -1.4 1.5 2 330
213 0 0 0 0 0 2 500
214 0 0 0 0 0 2 330
215 0 0 -4.1 -0.4 0 2 330
216 0 0 -15.65 -1.55 1.5 2 330
217 0 0 -10.7 -1.1 0 2 330
301 46.8 11.28 0 0 1.000 0 3 20
302 15.32 2.532 0 0 1.000 0 3 20
303 0 0 0 0 0 3 500
304 0 0 0 0 0 3 500
305 0 0 0 0 0 3 500
306 0 0 -12.3 -1.23 0 3 500
307 0 0 -6.5 -0.65 0 3 500
308 0 0 -6.55 -0.66 0 3 500
309 0 0 -1.95 -0.2 0 3 330
310 0 0 0 0 0 3 330
311 0 0 0 0 0 3 330
312 0 0 -1.15 -0.12 0 3 220
313 0 1.294 -24.05 -2.4 1.015 0 3 220
314 0 0 -2.5 -0.25 0 3 220
315 0 0 0 0 0 3 275
401 14 4.66 0 0 1.000 0 4 20
402 8.7 0.942 0 0 1.000 0 4 20
403 14 1.888 0 0 1.000 0 4 20
404 14.64 2.388 0 0 1.000 0 4 20
405 0 0 -12.15 -1.2 0 4 275
406 0 0 -9.05 -0.9 0 4 275
407 0 0 0 0 0 4 275
408 0 0 -1.85 -0.2 0 4 275
409 0 0 -3.1 -0.3 0.6 4 275
410 0 0 -6.5 -0.65 0 4 275
411 0 0 -7 -0.7 0.3 4 275
412 0 0.639 -15.35 -1.55 1.000 0 4 275
413 0 0 0 0 0 4 275
414 0 0 0 0 -0.3 4 330
415 0 0 0 0 -0.6 4 330
416 0 0 0 0 -0.6 4 330
501 6 -0.176 0 0 1.000 0 5 20
502 8 2.12 0 0 1.000 0 5 15
503 6.9 1.845 0 0 1.000 0 5 15
504 0 0 -2 -0.4 -0.9 5 275
505 0 0 0 0 0 5 275
506 0 0 0 0 0 5 275
507 0 0.368 -7.1 -1.4 1.040 0 5 275
508 0 0 -5.2 -1.05 0 5 275
509 0 0.502 -0.7 -0.15 1.027 0 5 275
104 APPENDIX E. IEEE SOUTH EAST AUSTRALIAN SYSTEM
Table E.2: Parameters of the branches of IEEE South Australian test system
[4] calculated on 100 MVA base
Line No. From Bus To Bus Tap Ratio R [p.u.] X [p.u.] Bsh [p.u.] Line No. From Bus To Bus Tap Ratio R [p.u.] X [p.u.] Bsh [p.u.]
1 102 217 1 0.002022 0.016066 3.268 38 408 410 1 0.005500 0.064000 2.019
2 102 309 1 0.001865 0.014771 1.634 39 409 411 1 0.005150 0.035450 0.920
3 205 206 1 0.004800 0.038000 1.862 40 410 411 1 0.004300 0.053200 0.427
4 205 416 1 0.001850 0.023000 1.460 41 410 412 1 0.001075 0.013300 1.708
5 206 207 1 0.002250 0.017800 0.874 42 410 413 1 0.002000 0.024700 0.800
6 206 212 1 0.003300 0.026350 1.292 43 411 412 1 0.000600 0.012500 0.780
7 206 215 1 0.003300 0.026350 1.292 44 414 415 1 0.001000 0.012500 0.780
8 207 208 1 0.000900 0.007000 0.342 45 415 416 1 0.001850 0.023000 1.460
9 207 209 1 0.000800 0.006200 0.076 46 504 507 1 0.011500 0.075000 1.120
10 208 211 1 0.001033 0.008267 0.912 47 504 508 1 0.013000 0.009500 1.740
11 209 212 1 0.004500 0.035600 0.437 48 505 507 1 0.000800 0.008500 0.060
12 210 213 1 0.000500 0.007250 3.080 49 505 508 1 0.002500 0.028000 0.170
13 211 212 1 0.000700 0.005400 0.266 50 506 507 1 0.000800 0.008500 0.060
14 211 214 1 0.001900 0.015500 0.190 51 506 508 1 0.015000 0.110000 1.800
15 212 217 1 0.007000 0.055800 0.684 52 507 508 1 0.002000 0.019000 0.090
16 214 216 1 0.001000 0.007700 0.095 53 507 509 1 0.003000 0.022000 0.900
17 214 217 1 0.004900 0.038800 0.475 54 101 102 0.948 0.000000 0.012000 0.000
18 215 216 1 0.002550 0.020150 0.988 55 201 206 0.948 0.000000 0.004800 0.000
19 215 217 1 0.003600 0.028700 1.406 56 202 209 0.948 0.000000 0.007200 0.000
20 216 217 1 0.005100 0.040300 0.494 57 203 208 0.948 0.000000 0.010200 0.000
21 303 304 1 0.001000 0.014000 1.480 58 204 215 0.948 0.000000 0.006000 0.000
22 303 305 1 0.000550 0.008000 3.400 59 209 210 0.99 0.000000 0.006800 0.000
23 304 305 1 0.000300 0.004000 0.424 60 213 214 1 0.000000 0.006800 0.000
24 305 306 1 0.000200 0.003000 0.320 61 301 303 0.935 0.000000 0.003000 0.000
25 305 307 1 0.000150 0.002250 0.894 62 302 312 0.952 0.000000 0.008450 0.000
26 306 307 1 0.000100 0.001200 0.127 63 304 313 0.961 0.000000 0.016000 0.000
27 307 308 1 0.001150 0.016250 6.890 64 305 311 1 0.000000 0.012000 0.000
28 309 310 1 0.004500 0.035667 1.748 65 305 314 1 0.000000 0.012150 0.000
29 310 311 1 0.000000 -0.016850 0.000 66 308 315 0.96 0.000000 0.013500 0.000
30 312 313 1 0.002000 0.015000 0.900 67 401 410 0.939 0.000000 0.008450 0.000
31 313 314 1 0.000500 0.005000 0.520 69 403 407 0.952 0.000000 0.008450 0.000
32 315 509 1 0.003500 0.025000 0.380 70 404 405 0.952 0.000000 0.008500 0.000
33 405 406 1 0.001950 0.023750 0.762 71 413 414 1 0.000000 0.002667 0.000
34 405 408 1 0.005400 0.050000 0.189 72 501 504 0.952 0.000000 0.025500 0.000
35 405 409 1 0.006000 0.040667 2.370 73 502 505 0.93 0.000000 0.016000 0.000
36 406 407 1 0.000300 0.003800 0.124 74 503 506 0.93 0.000000 0.020000 0.000
37 407 408 1 0.004200 0.051300 0.412
Table E.4: Eigenvalues of the South East Australian test system in Base
Case
107
108 BIBLIOGRAPHY