Professional Documents
Culture Documents
Vol 9
ISBN 978-3-7376-0716-2
9 783737 607162
Gustav Lammert
were determined in collaboration with a manufacturer. The stability
analysis is performed in DIgSILENT PowerFactory using: i) a one-load
infinite-bus system; and ii) an IEEE multi-machine voltage stability
test system, namely the Nordic test system. The results show that
with the adequate control of photovoltaic systems, power system
dynamic performance can be significantly improved.
Energy Management and Power System Operation
Vol. 9
Edited by
Prof. Dr.-Ing. Martin Braun
University of Kassel
Gustav Lammert
kassel
university
press
This work has been accepted by the Faculty of Electrical Engineering / Computer Sciences of
the University of Kassel as a thesis for acquiring the academic degree of Doktor der
Ingenieurwissenschaften (Dr.-Ing.).
Abstract
This thesis investigates the impact of: i) the low voltage ride-through and dynamic
voltage support capability; ii) the active current recovery rate; iii) the local voltage
control; and iv) the plant-level voltage control of large-scale photovoltaic systems
on short-term voltage stability and fault-induced delayed voltage recovery as well
as transient and frequency stability. The power system dynamic performance is
analysed using state-of-the-art methods, such as phasor mode time-domain simula-
tions and the calculation of the critical clearing time that determines the stability
margin. Moreover, the recently developed Kullback-Leibler divergence measure is
applied to assess the quality of the voltage recovery. Drawbacks of this metric
are outlined and a novel metric, the so-called voltage recovery index, is defined that
quantifies the phenomenon of fault-induced delayed voltage recovery more systemat-
ically. The studies are performed with a generic photovoltaic system model, which
was developed by the Western Electricity Coordinating Council. Typical model
parameters are used that were determined in collaboration with a manufacturer.
The implemented model is successfully validated against the Renewable Energy
Model Validation tool that was developed by the Electric Power Research Institute.
Moreover, the implemented photovoltaic model is open-source software and can be
used by academia and industry. The stability analysis is performed in DIgSILENT
PowerFactory using: i) a one-load infinite-bus system in order to show the impact of
the photovoltaic system control modes on the fundamental concepts and principles
of short-term voltage stability; and ii) an IEEE multi-machine voltage stability test
system, namely the Nordic test system, that additionally illustrates interactions
with other power system components, such as synchronous generators. The results
show that without the low voltage ride-through capability, the multi-machine sys-
tem is short-term voltage and transient unstable. Only the low voltage ride-through
and dynamic voltage support capability help to avoid instability. The fastest active
current recovery rate achieves the best voltage recovery. However, this recovery rate
also leads to overfrequencies in the system that could cause disconnection of gener-
ation. Therefore, the rate should be tuned considering both, voltage and frequency
dynamics. In case of local constant voltage control, photovoltaic systems try to
restore their pre-fault voltage by increasing their reactive and reducing their active
currents. However, due to the rather stiff grid behaviour, restoring grid voltage
is impossible and the system collapses owing to the lack of active power produced
by the photovoltaic systems. To overcome this problem, adequate reactive power
limitation is required or the current limit logic needs to be changed. If plant control
is used, plant-level constant voltage and local coordinated reactive power/voltage
control should be applied. Finally, the results show that the voltage recovery in-
dex provides useful information about the delayed voltage recovery and helps to
compare different short-term voltage controls of photovoltaic systems.
ix
Acknowledgements
This dissertation was part of my work as a research assistant at the Department
of Energy Management and Power System Operation at the University of Kassel.
Throughout this time, I met remarkable people, whom I would like to thank in the
following paragraphs.
First and foremost, I would like to thank my supervisor Prof. Dr.-Ing. Martin
Braun, for giving me the opportunity to work under his guidance. He has al-
ways encouraged and trusted me and I am thankful that I have been part of his
research group. I would also like to express my gratitude to my co-supervisor
Prof. Dr. Thierry Van Cutsem, for his professional supervision during my external
research stay in Belgium. He helped me not only to develop valuable skills as a
researcher but also as a person. I wish to thank Prof. Dr.-Ing. Mike Meinhardt
for his continuous support over the least years. His lecture on power electronics
for renewable energy resources was outstanding at the University of Kassel. Many
thanks to Prof. Dr.-Ing. habil. Christian Rehtanz, for devoting his time to be mem-
ber of the examining committee and evaluate this thesis. I owe a special thanks to
Prof. Dr.-Ing. habil. István Erlich, who passed away unexpectedly in May 2018. He
was initially part of the examining committee and his exceptional knowledge and
the critical as well as passionate discussions gave the thesis the final shape.
A big thank you to all of my colleagues from the Department of Energy Man-
agement and Power System Operation at the University of Kassel. In particular,
I would like to thank my office colleagues Dr.-Ing. Dirk Fetzer and Dr.-Ing. Darı́o
Lafferte for the fruitful scientific and non-scientific discussions. Their help and mo-
tivation was fundamental to direct my research towards promising areas and to
avoid dead-ends. Needless to say that I also thank the other team members Kai
Fischbach, Christian Hachmann and Alexander Klingmann who supported me in
many different but always useful ways. Furthermore, I would also like to thank the
former colleagues Priyanka Chaudhari, Dr. Stefan Gehler, Dr. Jan Hegemann, Tina
Paschedag and Dr.-Ing. Robert Schmoll for the insightful discussions on technical
and non-technical topics alike.
I owe many thanks to my colleagues from the Fraunhofer Institute for Energy
Economics and Energy System Technology. Throughout the process of this disser-
tation, I have greatly benefited from the close collaboration with Dr.-Ing. Jan von
Appen, Dr. Thomas Degner, Malte Hof, Markus Kraiczy, Maria Nuschke and Luis
David Pabón Ospina. Not to forget, the former colleagues Dr.-Ing. Dominik Geibel,
Dr.-Ing. Tobias Hennig and Prof. Dr.-Ing. Thomas Stetz that I would like to thank
for their support and advise.
I express my deep gratitude to all members of the CIGRE C4/C6.35/CIRED
joint working group on Modelling and Dynamic Performance of Inverter-Based Gen-
eration in Power System Transmission and Distribution Studies. The participation
x
Publications
Journal publications
Conference publications
Other publications
Supervised theses
Contents
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Appendices 117
Nomenclature 165
Bibliography 175
1
Introduction
1
1.1 Motivation
Electrical power systems are undergoing fundamental structural and technological
changes. The term “structural changes” means that power systems are transformed
from vertically structured systems, with unidirectional power flows from the trans-
mission to the distribution system, to horizontally structured systems, with bidi-
rectional power flows. This includes also reverse power flows from the distribution
back to the transmission system as well as power flows among voltage levels. The
term “technological changes” means that conventional power plants equipped with
synchronous generators are replaced by generation connected via power electronic
converters. This replacement leads to a transition from a highly centralised elec-
tricity production of a few large power plants to a decentralised production of many
small units based on renewable energy resources with a stochastic nature. These
small units are also referred to as distributed generation and most of them are
inverter based and mainly consist of PhotoVoltaic (PV) and wind generation. In
Germany, for instance, inverter-based generation is connected to all voltage levels,
as shown in Fig. 1.1. It should be noted that PV generation is mostly connected
to the low and medium voltage level, whereas wind generation is predominant in
the medium and high voltage level. However, large-scale PV plants connected to
the high or extra high voltage level, such as the 349 MW PV park in Australia [1],
are also planned. The term “large-scale” refers to large PV or wind parks with a
nominal power of 50 MW and higher [2]. Aside from the change of the generation,
also the composition of the load is changing. In Germany, for instance, air condi-
tioning systems [3] or heat pumps [4] are being increasingly installed. Some of them
include induction motors that are directly coupled to the grid. This, in turn, re-
quires specific power system dynamic studies, such as short-term voltage stability or
Fault-Induced Delayed Voltage Recovery (FIDVR) studies. It should be noted that
there are also air conditioning systems or heat pumps, which are coupled via power
electronics, i.e., frequency converters, to the grid. This thesis focuses on the control
of large-scale PV plants and its impact on the dynamic performance of the power
2 Chapter 1. Introduction
system, specifically induction motor loads. Large-scale PV parks are similar in many
aspects with wind power plants that use full converter wind turbines. Therefore,
the achieved results in this thesis of the impact analysis of PV systems on power
system stability, can be transferred to full converter wind turbines due to the similar
modelling and control approach. Nevertheless, the dynamic characteristics of wind
and PV generation are different from conventional power plants, which may impact
the performance of the power system following disturbances. Moreover, generation
based on renewable energy resources that are mostly coupled via inverters, can al-
ready exceed load consumption in some cases. In Germany, for instance, on May 1,
2018, at 2 p.m., all of the load was covered by renewable energy resources, as shown
in Fig. 1.2. These extreme operating conditions require detailed power system dy-
namic studies that analyse the impact of inverter-based generation, specifically PV
generation, on stability of the overall system. The outcome of these studies are grid
code requirements that describe operational aspects of inverter-based generation,
such as the standard for North America [5], the network code for Europe [6], or, in
particular, the guideline for the high voltage level in Germany [7]. The grid codes
set the minimum technical requirements in terms of electrical performance, planning
and operation of an individual unit or park. However, two recent examples from
Europe illustrate the effects of not carefully designed grid codes. The first issue was
the non-existent Fault Ride-Through (FRT) requirement [8]. Initially, grid code
requirements defined the disconnection of inverter-based generation during voltage
dips, which is referred to as passive behaviour. Wind and PV generation would trip
in the wide area, where the fault occurs. This would lead to a considerable loss of
production after the fault, which could result in a significant frequency drop. After
120
Low voltage
Medium voltage
100 High voltage
Extra high voltage
Active power [GW]
80
60
40
20
0
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Year
power system dynamic studies [10] identified the problem, FRT requirements [11]
for inverter-based generation were defined that enhance the dynamic performance
of the system. The second issue was the 50.2 Hz problem [8]. Initially, grid codes
defined the disconnection of wind and PV generation for frequencies higher than
50.2 Hz. At that time, about 30 GW PV generation was installed in Germany at
all voltage levels. The simultaneous disconnection of generation in a situation of
high PV power production would exceed the total reserve power in Europe of 3 GW
and most probably lead to a blackout. This issue was solved by a system wide
retrofit program that changed the protection settings of PV units in Germany [12].
In general, these issues show how important it is to make right assumptions on the
future structure of the electrical power system. On the other hand, some of the
aforementioned issues are still under investigation, such as the FRT requirements
for PV systems [13]. Furthermore, recent incidents, such as the Southern California
event [14] or the South Australian blackout [15], both in 2016, highly motivate
the studies reported in this work. The former event was influenced by missing or
incorrect FRT requirements and too slow active current recovery rates [14]. The
driving forces for the latter event were the inability to ride through multiple faults
and the missing Dynamic Voltage Support (DVS) capability, i.e., no additional
reactive current injection in response to the voltage dip [15]. Both incidents require
further power system dynamic studies that determine the impact of different control
and protection methods of inverter-based generation on power system stability. In
this context, adequate model parameters and validated models are crucial in order
to receive a realistic system behaviour and to derive feasible conclusions [16]. All
of the aforementioned aspects are addressed in detail within this thesis.
100 Biomass
Hydro, pumped storage
90 Wind
80 Photovoltaic
Coal, nuclear, gas
Energy [GWh]
70 Consumption
60
50
40
30
20
10
0
00:00 04:00 08:00 12:00 16:00 20:00 24:00
Time [hh:mm]
Figure 1.2: Power generation and consumption in Germany on May 1, 2018 [17]
4 Chapter 1. Introduction
Multi-machine system
Fault-induced delayed
through and dynamic
Critical clearing time
Frequency stability
infinite-bus system
Transient stability
Low voltage ride-
Kullback-Leibler
voltage recovery
voltage stability
voltage support
voltage control
Active current
recovery rate
Short-term
Plant-level
Reference
Validated
One-load
Generic
[9]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
Thesis
6 Chapter 1. Introduction
1.3 Contributions
This section presents the scientific contributions of this thesis. According to the re-
search gaps identified in Section 1.2, the results of this work contribute significantly
to fill these gaps. The contributions are separated according to the main chapters
of this thesis, as presented in the sequel.
• For the impact analysis of PV system controls, different test systems are con-
sidered. First, a one-load infinite-bus system is used to present fundamental
concepts and principles of power system stability, specifically short-term volt-
age stability and FIDVR. Second, an IEEE multi-machine test system, i.e.,
the Nordic test system, is considered to study additionally the interactions
with other power system components, such as synchronous generators. The
latter test system is particularly suitable for voltage stability studies. Both
test systems enable a precise evaluation on how PV system control influences
the dynamic performance of the power system.
1.4 Structure
This section presents the structure of the thesis. The remainder of this work is
organised as follows.
Chapter 2: This chapter presents the relevant stability definitions and their meth-
ods for the analysis of power system dynamic performance. First, the time-domain
simulation method is introduced. Next, the calculation of the CCT, which deter-
mines the stability margin of the system, is presented. The delayed voltage recovery
is assessed using the KL divergence measure. The major steps for the computation
of this metric as well as the drawbacks are outlined and a novel metric, the so-called
VRI, is introduced that quantifies the phenomenon of FIDVR more systematically.
Finally, the conclusions are drawn and directions for future work are outlined.
Chapter 4: In this chapter, the stability analysis of the one-load infinite-bus system
is presented. First, the test system is introduced, which includes details about
the load and the PV system model, the operating point as well as the considered
contingency. Next, the case studies are presented, specifically the results of the
impact analysis of different PV system control modes on short-term voltage stability
and FIDVR. Finally, based on the achieved results, the conclusions are drawn and
directions for future work are outlined.
Chapter 5: This chapter presents the stability analysis of the multi-machine sys-
tem, namely the Nordic test system. First, an overview of the test system is given,
which includes details about the load model, the substitution of synchronous gen-
erators, the PV system model, the operating point as well as the considered con-
tingencies. Next, the case studies are introduced. In particular, the results of the
impact analysis of different PV system control modes on short-term voltage stabil-
ity and FIDVR as well as transient and frequency stability are illustrated. Finally,
the conclusions and directions for future work are presented.
Chapter 6: In this chapter, the general conclusions of the thesis are summarised
and the main directions for future work are outlined.
9
Power System
Stability
situation, any one form of instability may not occur in its pure form, especially in
highly stressed systems [49]. Therefore, the short-term control of PV systems does
not only influence short-term voltage stability, but also transient and frequency sta-
bility, which is also briefly investigated in this work. Hence, these stability criteria
are also highlighted in light grey in Fig. 2.1 and their definitions and analysis meth-
ods are given in Sections 2.2.2 and 2.2.3, respectively. Another important aspect
that is closely linked to short-term voltage stability is the delayed voltage recovery,
specifically the FIDVR phenomenon. This phenomenon can take place before insta-
bility is reached, as explained in Section 2.2.4. Furthermore, the distinction between
short-term voltage stability and transient stability is not always easy and both may
go hand in hand, which is presented in Section 2.2.5. It should be noted that the
presented power system stability categories and definitions are mainly based on re-
port [49] and represent synchronous generator dominated systems. However, as the
penetration level of inverter-based generation is going to increase, it is essential to
revisit some of the existing definitions. Eventually, new terms and definitions need
to be introduced, where appropriate that will adequately reflect and describe the
existing and future dynamic behaviour of evolving power systems [50].
Analysis methods: Frequency excursions may vary from a few seconds to several
minutes [49]. A single-bus equivalent with the representation of aggregated load and
generation can be used for simplified analyses. For detailed frequency studies, a full
network representation is advisable. A common method to analyse frequency sta-
bility is to perform numerical time-domain simulations. In this work, time-domain
analysis of the full network representation is used to study frequency stability, as
detailed in Section 2.3.
Transient overvoltages
Sub-synchronous resonance
Power system phenomena
Transient stability
Relevant
phenomena
Fault-induced delayed voltage recovery
Frequency stability
RMS simulation: RMS simulations are also called phasor mode or simply pha-
sor simulations. RMS models are mainly used to study power system stability
for large interconnected systems. This includes electromechanical oscillations, i.e.,
small-signal stability, transient stability, voltage stability and frequency stability, as
shown in Fig. 2.1. The power system is modelled by a set of non-linear DAEs [64].
Phasor simulation methods are used if the magnitude and phase of the voltages
and currents are of interest. It is not necessary to solve all the differential equa-
tions resulting from the interaction of resistances, inductances and capacitances.
Hence, the network is simulated with fixed complex impedances instead of differen-
tial equations [66]. RMS simulation tools consider phenomena with a bandwidth of
typically 0.1 Hz to 3 Hz, since the network model’s fidelity diminishes rapidly for
phenomena with frequencies significantly outside of this range. However, the con-
trol loops modelled on individual power plants may cover phenomena up to 10 Hz.
Another commonly used terminology for RMS type models is to call them positive-
sequence stability models. In essence, many of these models assume a perfectly
balanced network and consider only the positive-sequence components of all phe-
nomena. Converters are included in RMS programs using their averaged models.
They provide the behaviour of the converter ignoring fast switching transients and
any control with very small time constants compared to the time steps and the
phenomena that are considered. For instance, in stability analysis there is usually
no need to model the inverter in detail since its transients are much faster than the
dynamics being studied. Therefore, only the fundamental frequency outputs, i.e.,
50 Hz or 60 Hz, of the converter are modelled, which are mainly reflected in the
electrical control model, as shown in Section 3.2.3. Due to this averaged represen-
tation, the modelling of different technologies of inverter-based generation can be
unified using sub-models for each basic component according to their characteristics.
TM
Typical tools for RMS simulations are DIgSILENT PowerFactory R
, GE PSLF ,
PowerWorld Simulator, RAMSES, or Siemens PTI PSS E. It should be noted that
R
the use of different RMS software tools might lead to different results performing
the same stability studies [67, 68]. The investigations presented in this work fall
under the category of positive-sequence RMS simulations and the simulation tool
DIgSILENT PowerFactory [69] is used.
2.3. Time-domain simulation 17
• studies of weak system conditions with a very low short circuit ratio,
It should be noted that many RMS models are positive-sequence models. Typical
TM TM
programs for EMT simulations are PSCAD /EMTDC , EMTP-RV, OPAL-RT
R R
HYPERSIM , RTDS or DIgSILENT PowerFactory .
ẋ = f (x, y, t) (2.1)
0 = g(x, y, t) (2.2)
where x refers to the state vector, y to the vector of inputs and the time is denoted
as t. The derivative of the state vector is denoted by ẋ. With the initial values of
the DAE system:
x(t0 ) = x0 (2.3)
y(t0 ) = y0 (2.4)
that are usually obtained by performing a power flow computation with the initial
time t0 . It should be noted that the form of algebraic equations g can be twofold,
i.e., a current-injection or power-injection model [65]. The most common model
used in proprietary software packages is the current-injection one [73]. In order to
solve Eqs. (2.1) and (2.2), different numerical integration methods are available.
Explicit methods: Explicit numerical integration methods are well suited for non-
stiff ordinary differential equations. Many different explicit methods exist and for
power system applications, Runge-Kutta’s and the modified Euler’s method have
been widely used [64, 74].
This definition refers to the stability of the overall power system. However, previous
definitions of the CCT were based on transient stability, where the CCT is defined
as the maximum time between the initiation and the isolation of a fault, such that
the power system is transiently stable [75]. In general, the overall goal of the CCT
is to determine the stability margin of the system. Several calculation methods,
which depend on the stability criterion, exist, as shown in Table 2.1. In this work,
the CCT refers to both, short-term voltage stability and transient stability, and is
calculated through time-domain simulations using a trial-and-error strategy, which
is explained in the sequel.
4. Go back to Step 2 until system condition changes, i.e., stable → unstable then
() (+1)
tCCT := tCCT ; or unstable → stable then tCCT := tCCT .
As shown in the trial-and-error strategy, the calculation of the CCT considers both,
short-term voltage and transient stability. If only one of these is unstable, the fault
duration exceeds the CCT. The time step Δt is 0.01 s, as presented in Section 2.3.4.
2. Integral error-based methods: Metrics based on the integral [60, 61] cannot
distinguish between the waveform of the following two voltages. One with a
lower initial voltage drop that takes more time to recover. The other with a
higher initial voltage drop but showing fast recovery over a small period of
time.
• dynamic voltage control areas for placing static var and static synchronous
compensators [27].
The considered voltage time series data is from fault clearing instant tcl to the final
observation time tf with the time limits tcl < t1 < t2 < tf .
Vmax
Vpre
V3
WECC criterion
V2
V1
Voltage
Vgood
Vbad
Vmin
tcl t1 t2 tf
Time
Figure 2.4: WECC voltage violation criterion and sample voltage recoveries
22 Chapter 2. Stability definitions and analysis methods
1 KL
L
P KL = P (2.6)
A i=1 i
which leads to the PDF P KL in the form of a normalised distribution. The details
of the construction of the PDF P KL are shown in Algorithm 2.1.
1: procedure PDF for KL measure
2: initialise PiKL = 0 for all i
3: for each voltage interval i
4: for each discrete time t in i
5: PiKL (V (t)) := PiKL (V (t)) + 1
6: endfor
7: endfor
8: divide all PiKL by the total number of discrete times A
9: end procedure
Vmax
V3
V2 Pgood
KL
of Vgood
V1
Voltage
Pbad
KL
of Vbad
Vmin
0 0.05 0.1 0.15 0.2 0.25 0.3
Probability density function
Assuming a value for the parameter γ of 450 [27], the resulting PDF for the reference
voltage recovery Pref
KL
is shown in Fig. 2.6. The objective is to compare the distance
between the PDF of a given voltage recovery P KL and the PDF of the reference
voltage recovery Pref
KL
using the KL divergence measure K. In the same figure, the
PDF of the WECC voltage violation criterion is depicted, which consists of three
peaks at the voltage limits V1 , V2 and V3 accordingly.
Vmax
Pref
KL
Vpre
V3
V2
V1
Voltage
PWECC
KL
Vmin
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Probability density function
Figure 2.6: Probability density functions of reference voltage recovery and WECC
voltage violation criterion
24 Chapter 2. Stability definitions and analysis methods
L
L
K = ln W + PiKL ln PiKL + γ PiKL (Vi − Vpre )2 (2.10)
i=1 i=1
Analysing Eq. (2.10), it can be concluded that if the voltage recovery is delayed,
the KL divergence measure is high. This is due to the fact that PiKL is higher
for smaller i. The smaller Vi , the larger the weighting factor (Vi − Vpre )2 . This
would lead to a higher KL divergence measure K for a delayed voltage recovery. In
general, the value of the KL divergence measure is bounded below by zero, which
corresponds to the PDF of the voltage recovery P KL matching exactly the PDF
of the reference voltage recovery PrefKL
. Therefore, it can be concluded, the smaller
the KL divergence measure K, the better the voltage recovery. According to the
WECC criterion with the envelope VWECC given in Eq. (2.5) and the definitions of
Δt1 = t1 − tcl , Δt2 = t2 − t1 , Δt3 = tf − t2 , and Δt4 = tf − tcl , the critical value of
the KL divergence measure, denoted as KWECC , is calculated according to:
1
KWECC = (Δt1 log Δt1 + Δt2 log Δt2 + Δt3 log Δt3 )
Δt4
γ
+ (Δt1 (V1 − Vpre )2 + Δt2 (V2 − Vpre )2 + Δt3 (V3 − Vpre )2 ) (2.11)
Δt4
+ log W − log Δt4
If the KL divergence measure K is greater than KWECC , the WECC criterion is vio-
lated. It should be noted that the values of the concentration parameter γ and the
sub-intervals L significantly influence the resulting KL divergence measure. Consid-
ering the evolution of the sample voltage recoveries shown in Fig. 2.4 and the PDF
for the reference voltage recovery defined in Eq. (2.7), the KL divergence measures
are calculated accordingly, as shown in Table 2.2. Note that the aforementioned
procedure for the calculation of the KL divergence measure, along with the assumed
parameters, is considered for the stability analysis in Chapters 4 and 5, respectively.
Vmax
V3
V2
V1
Voltage
WECC criterion
Videal
Vosc
Vmin
tcl t1 t2 tf
Time
1 VRI+
M L
P VRI = (P + PjiVRI− ) (2.12)
A j=1 i=1 ji
where M denotes the number of voltage constraints and the index j refers to the
j-th voltage constraint. It should be emphasised that compared to the PDF for the
KL divergence measure P KL given in Eq. (2.6), the PDF for the VRI P VRI defined
in Eq. (2.12) does not only give information about the distribution of the voltage,
but it also shows its relation to the constraints, i.e., if and when voltage limits are
violated. The specific PDF with regard to the WECC criterion with M = 3, for
instance, consists of six PDFs in total, two according to the voltage violations, i.e.,
V1 , V2 , V3 , times three according to the time intervals, i.e., [tcl , t1 [; [t1 , t2 [; [t2 , tf [,
as given in:
L
1 VRI+ VRI+ VRI+
L L
P VRI+ = P1i + P2i + P3i (2.13)
A i=1 i=1 i=1
1
L
L
L
P VRI−
= P1i
VRI−
+ P2i
VRI−
+ P3i
VRI−
(2.14)
A i=1 i=1 i=1
For instance, Fig. 2.4 shows a good and a bad voltage recovery and the corresponding
PDFs are constructed in Fig. 2.8.
Vmax
V3
V2 Pgood
VRI
of Vgood
V1
Voltage
VRI+
P1i
Pbad
VRI
of Vbad VRI−
P1i
VRI+
P2i
VRI−
P2i
VRI+
P3i
VRI−
P3i
Vmin
0 0.05 0.1 0.15 0.2 0.25 0.3
Probability density function
− 2
ηj− = −e−λj (Vi −0) for j = 1, 2, ..., M and i = 1, 2, ..., L (2.16)
where the parameters and λ+
j λ−
control the width of the rewarding and penal-
j
ising weighting function, respectively. The parameter j refers to the j-th voltage
constraint. It should be noted that these distribution functions are very appropri-
ate because voltage oscillations but also overvoltages can be taken into account.
−
Furthermore, it should be emphasised that the values of λ+ j and λj significantly
influence the resulting VRI. As a general rule, the coefficients should be tuned ac-
cording to the voltage violation criterion. In case of the WECC voltage violation
criterion VWECC , the coefficients are tuned according to:
Assuming a value for the parameter σ of 0.01, and considering the WECC criterion
with M = 3, the so-defined weighting functions are depicted in Fig. 2.9. The corre-
− −
sponding values for the coefficients λ+
1 , ..., λ3 and λ1 , ..., λ3 are listed in Table A.3
+
of Appendix A.
Vmax
Vpre
V3
V2
V1
Voltage
η1+
η1−
η2+
η2−
η3+
η3−
Vmin
-1.0 -0.5 0.0 0.5 1.0
Weighting function
Future work: For very high penetration levels of inverter-based generation and
weak system conditions, RMS time-domain simulations might not be accurate or
suitable anymore. Instead, combined RMS and EMT, or pure EMT analysis is
required. In addition, the KL divergence measure could be improved if the PDF
of the reference voltage recovery is changed, specifically the concentration param-
eter of the density near the pre-fault voltage. Moreover, for the calculation of the
VRI, different weighting functions for overvoltages, i.e., voltages higher than the
normal operating voltage, could be introduced. Furthermore, the application of
other promising metrics, such as Lyapunov exponents or transient P–V curves, are
of interest for the analysis of short-term voltage stability.
31
Contributions: This chapter presents the WECC generic PV system model and
its implementation in DIgSILENT PowerFactory. Typical model parameters of the
PV system are provided that were determined in collaboration with a manufacturer.
The implemented PV system model is validated for several control modes against
another tool that represents the WECC specifications. In order to serve the need
of academia and industry, the implemented model is open-source software and can
be downloaded free of charge and used for any purpose.
3.2 Model
A generic PV system model is used to study the impact on power system stabil-
ity. The model follows the WECC specifications [82, 83] and is implemented in
DIgSILENT PowerFactory [69]. The model implementation is detailed in the fol-
lowing sections and additional information is given in reports [31, 84, 85].
3.2.1 Overview
The overall structure of the positive-sequence PV system model is depicted in
Fig. 3.1. The model includes 16 state variables and consists of three parts, namely:
It should be noted that the REGC A, REEC B, and REPC A models do not in-
clude any protection systems. However, voltage protection is modelled separately
as shown in the sequel. In this work, the phenomena of interest are short-term
9W 9W
REPC_A REEC_B REGC_A
9UHJ
9UHI Plant Level 4H[W ,TFPG’ ,TFPG ,T
4UHI Q Control
V/Q Control Current
4EUDQFK Generator Network
3UHI Limit
Model Solution
3EUDQFK Plant Level ,SFPG’ Logic ,SFPG ,S
3UHI P Control
)UHTBUHI P Control
)UHT
3T)ODJ
Figure 3.1: Block diagram of WECC generic photovoltaic system model [82]
3.2. Model 33
voltage stability and FIDVR, and subsequently transient and frequency stability, as
presented in Section 2.2 and shown in Fig. 2.3. In order to determine the impact of
PV system control on these phenomena, the relevant controls have to be identified.
Therefore, different power system controls, along with the three parts of the WECC
generic PV system model, i.e., the REGC A, the REEC B, and the REPC A model,
are depicted in Fig. 3.2. The fastest control is included in the REGC A model,
followed by the electrical control in the REEC B and the rather slow plant control
in the REPC A model. These relevant controls of the WECC generic PV system
model are highlighted with a solid arrow in Fig. 3.2. Moreover, other controls,
such as the synchronous generator control, the protection system, or the prime
mover control, are important to consider because they influence the aforementioned
power system phenomena. These controls are not investigated in detail in this
work, and hence highlighted with a dashed arrow in Fig. 3.2. The details about
the implementation and the necessary modifications of the REGC A, the REEC B,
and the REPC A model as well as the relevant control modes to study the impact
on power system stability, are outlined in the following sections.
FACTS control
Protection systems
REGC A 0: The REGC A 0 part represents the high voltage region and allows
reactive current management during high voltage events. In particular, the high
voltage reactive current management provides an additional reactive current de-
pending on the terminal voltage. Note that the additional reactive current injection
is not activated until the terminal voltage passes a certain over voltage threshold.
This provision of reactive current influences indirectly the terminal voltage of the
inverter. The REGC A 0 part also includes a ramp rate limiter that limits the
rate-of-change of reactive current output.
REGC A 1: The REGC A 1 part represents the low voltage region and enables
active current control during low voltage events. On the one hand, the low voltage
active current management emulates the response of the PLL controls during volt-
age dips. On the other hand, the low voltage power logic allows for a controlled
response of active current during and immediately following voltage dips. Note that
both controls involve a computation using a non-linear map, as shown in Fig. 3.3.
Moreover, the REGC A 1 part includes a ramp rate limitation of the active current
and a measurement delay of the terminal voltage.
REGC_A_0 ,TUPD[
,TFPG 1 ,T
1 + V7JT High Voltage
9W Reactive Current
,TUPLQ
Management
Network Interface
REGC_A_1
/93/ & UUSZU
,SFPG 1 ,S
1 + V7JS
J
Low Voltage
/93/ Active Current
/YSOVZ
0
Management
/YSO
1
1 + V7IOWU
1
]HUR[ EUNSW 9
REEC B 0: The REEC B 0 part exhibits the current limit logic and provides the
option to set the priority either to active or reactive current. To avoid an algebraic
loop, the current limitation is implemented as follows:
where Ipcmd and Iqcmd are the output signals of the REEC B 3 and REEC B 2
part, respectively. The maximum apparent current is denoted as Imax , and the
maximum active and reactive current limits are referred to as Ipmax and Iqmax ,
respectively. The value of Iqmin specifies the minimum reactive current limit. The
minimum active current limit, denoted as Ipmin, is zero.
GEG, GEG ,TK
REEC_B_1 1 9WBILOW -
9W .TY
1 + V7UY
+
(9W < 9GL
if (9W 9GLS)) or (9W
(9W > 9
9XS)
XS)) ,TO
9UHI
9ROWDJHBGLS = 1 1
else
9ROWDJHBGLS = 0 ,TLQM
9PD[
3I)ODJ ,TPD[
4PD[ 9)ODJ
1 9PD[ 4)ODJ + ,TPD[
3H × 1 +
1 + V7S .TS + .TL 1 +
V .YS + .YL 1 + ,TFPG
0 - V ,TFPG
SIDUHI WDQ 4PLQ 0 -
Freeze state i f 9PLQ 0
4JHQ 9PLQ
9PLQ 9ROWDJHBGLS = 1 Freeze state i f
,TPLQ
4H[W 9WBILOW ,TPLQ
,TPLQ 9ROWDJHBGLS = 1
0.01
D
1
÷ ,PD[ Current
1 + V7LT
V7LT
REEC_B_4 REEC_B_2 N Freeze state if
9ROWDJHBGLS = 1 Limit
Logic
3T)ODJ
) J
REEC_B_3
3PD[ & G3PD[ ,SPD[
1 N ,SFPG
3UHI ÷ ,SFPG
1 + V7SRUG
Freeze state i f
D ,SPLQ =0
3PLQ & G3PLQ 9ROWDJHBGLS = 1
9WBILOW
REEC_B_0
0.01
REEC B 1: The REEC B 1 part represents the DVS capability of the PV system
to control the terminal voltage in case of a voltage dip. The voltage is controlled
by the injection of an additional reactive current, i.e., the injection of an additional
capacitive reactive current leads to a voltage increase, whereas the injection of an
additional inductive reactive current leads to a voltage decrease considering the
consumer-oriented sign convention. Moreover, the REEC B 1 part includes the
voltage dip detection function that is activated (1) or deactivated (0) according to:
with the upper and lower voltage limits Vup and Vdip, respectively. This function
determines the state of the PV system. The fault condition is activated if one of the
voltage limits is exceeded. The PV system returns to the normal operating state if
the voltage is within the limits.
REEC B 2: The REEC B 2 part enables different local voltage control strategies,
i.e., local constant reactive power control, local constant voltage control, and local
coordinated reactive power/voltage control. These different local voltage control
modes apply Proportional Integral (PI) controllers that can be selected using the
flags QFlag and VFlag. It should be noted that the implementation of the anti-
windup limiters of the PI controllers is crucial for the dynamic performance of
the PV system. Therefore, it is explained in Appendix B.1.1. Moreover, some
of the control blocks are equipped with a freeze function that allows a distinction
between normal operation and fault condition, as given in Eq. (3.4). The detailed
implementation of the freeze function is also described in Appendix B.1.1.
REEC B 3: The REEC B 3 part provides active power control and generates an
active power reference signal, which is passed through a first order low-pass filter
with anti-windup and ramp rate limits. The generated active power signal is trans-
ferred into an active current signal using the terminal voltage. The details about the
anti-windup limits as well as the freeze function can be found in Appendix B.1.1.
REEC B 4: The REEC B 4 part allows constant power factor control, where the
reactive power is proportional to the active power. This control is modified by using
the tangent of the power factor angle reference pfaref according to:
Qext
tan(pfaref ) = (3.5)
Pref
with the initial active power reference Pref and reactive power reference Qext com-
puted from the power flow solution.
3.2. Model 37
REPC A 1: The REPC A 1 part controls either voltage or reactive power at plant
level using the flag RefFlag. The closed loop voltage or reactive power regulation
uses a PI controller with anti-windup limits and freeze function and the implemen-
tation is detailed in Appendix B.1.1. Note that the output signal of the REPC A 1
part is either a reference reactive power set point or the equivalent reactive power
needed to regulate the voltage at a specified bus.
REPC A 2: The REPC A 2 part represents the active power control at plant level
and provides a governor response derived from frequency deviation at a user-defined
bus. The frequency droop control is capable of being activated in both directions,
i.e., over and under frequency conditions. For under frequency conditions, power has
to be kept in reserve to allow for a quick active power injection. The active power
regulation uses a PI controller with anti-windup limits and its implementation is
presented in detail in Appendix B.1.1.
9FPS)ODJ 9UHI
,EUDQFK 1 1 - +
|9UHJ–(5F
|9UHJ – (5F + j;F)·,EUDQFK|
9UHJ 1 + V7YIOWU
V7Y
7YIOWU
0 4PD[BSODQW
[
GEG HPD[
+ 1
4EUDQFK
+ .S.L
.S .L 1 + V7IW
.F 5HI)ODJ 4H[W
V 1 + V7IY
REPC_A_0 0
HPLQ
Freeze state
Q
4PLQBSODQW if 9UHJ < 9IU]
1 -
1 + V7T
V7TIOWU
7 IOWU
+
REPC_A_1 4UHI
3ODQWBSUHI
[
3PD[BSODQW )UT)ODJ
IHPD[ 0
+
3EUDQFK 1 - .SJ + .LJ 1 3UHI
1 + V7S 0 V 1 + V7ODJ 1
+ IHPLQ
IGEG, IGEG 'GQ
- + 3PLQBSODQW
Q
)UHT
+ +
'XS
)UHTBUHI REPC_A_2
0
3.3 Control
The WECC generic PV system model can be configured in over 30 unique con-
trol modes [86]. Due to the numerous ways the model can be configured, selecting
the appropriate model structure is vital. Each unique model configuration corre-
sponds to a particular control mode. In this section, the relevant modes of operation
of PV systems that impact power system stability, specifically short-term voltage
stability and FIDVR, are identified. These control modes, i.e., LVRT and DVS ca-
pability, active current recovery rate, local voltage and plant-level voltage control,
describe state-of-the-art requirements and settings according to international stan-
dards and grid codes. The overall goal is to consider control modes, in particular
voltage control modes that influence short-term voltage dynamics. The relevant
control modes are shown in Table 3.1. For each control mode, different parameter
variations and/or flag combinations of the WECC generic PV system model are
incorporated. The relevant parameters and flag combinations, along with the as-
sociated models, are presented in the same table. For all control modes, reactive
current priority with PqFlag = 0 is considered according to international guidelines
and grid codes [7]. Moreover, it is demonstrated that the identified short-term con-
trol modes of PV systems do not only impact voltage stability, but also frequency
stability and transient stability. This is shown in detail for several case studies
considering a multi-machine system, as presented in Chapter 5. It should be noted
that only the positive-sequence control of PV systems is of interest in this thesis.
The relevant control modes of the PV system to study the impact on power system
stability are outlined in the sequel.
No low voltage ride-through capability (No LVRT): This mode uses the
under voltage protection system of the PV plant, as shown in Fig. 3.6. The focus
in this work is on voltage dips. Thus, only the under voltage function of the pro-
tection system is considered. Due to the voltage sag in the grid, the PV system
immediately disconnects at the Point of Common Coupling (PCC). This setting
represents old German grid code requirements [87, 88], which is still used in some
countries. The considered protection relay is depicted in Fig. 3.7. The measure-
ment block calculates the RMS values out of the momentary values and transfers
them to the under voltage protection function, where the voltage limits and time
delays are defined. The tripping signal is sent to the OR function that opens the
circuit-breaker and disconnects the PV plant. Note that in practice the PV plant
disconnects slightly delayed due to the measurement and inherent delay of the pro-
tection system. However, this time delay is neglected due to its marginal effect on
the dynamic performance of the power system.
External Step-up Photovoltaic
grid transformer system
PV
PCC Voltage
transformer
Under voltage
V < protection function
Va
Vb Trip
(1) V <
Vc
va
Voltage vb Trip Trip
Measurement (2) V < OR
transformer vc circuit-breaker
Trip
(n) V <
Low voltage ride-through and blocking mode (LVRT & block): This con-
trol mode considers two aspects, i.e., the LVRT and the blocking capability of the
PV system. The LVRT capability, or in general FRT capability, allows the PV
system to remain connected to the grid during voltage dips. The considered LVRT
characteristic is presented in Fig. 3.8 and based on the German grid code for the
high voltage level [7]. If the voltage during the fault-on period lies within the no trip
area, the PV system has to stay connected to the grid and ride through the fault.
On the other hand, the PV system may disconnect if the voltage is below the LVRT
characteristic. In practice, the LVRT capability can be enabled using a step-wise
under voltage protection system with n functions, as shown in Fig. 3.7. The second
aspect of this control mode is related to the behaviour of active and reactive current
during the fault-on period. With regard to this, blocking means that the PV sys-
tem does not inject any active or reactive current during the fault-on period. The
blocking mode is also called zero power or momentary cessation mode and is applied
in many inverters in the United States of America [14] and in Germany [89]. The
blocking mode is activated using the static gain in the REEC B 1 part of Fig. 3.4
with Kqv = 0 pu, and the low voltage power logic in the REGC A 1 part of Fig. 3.3
with Lvplsw = 1. The low voltage power logic utilises a non-linear map that reduces
the active current as a function of the voltage. It should be noted that an interme-
diate control option between the blocking mode and the DVS capability, which is
described in the next paragraph, is possible. Instead of reducing active and reactive
power to zero during the fault-on period, this control mode would try to keep them
at their pre-fault values. However, this control method is out of scope.
1.0
0.6
May trip area
0.4
0.2
0.0
Low voltage ride-through and dynamic voltage support (LVRT & DVS):
This control mode includes two aspects. The first aspect is similar to the former
mode, i.e., the LVRT capability. This capability allows the PV system to remain
connected to the grid during voltage dips according to the LVRT characteristic
shown in Fig. 3.8. The LVRT curve is based on the German grid code for the high
voltage level [7]. The second aspect of this control mode enables voltage support
during the disturbance, which is the DVS capability, also known as dynamic reactive
power support or dynamic voltage control. The DVS capability defines an additional
reactive current injection Iqinj determined in proportion to the voltage deviation
ΔV . The additional reactive current injection is based on the German grid code
for the high voltage level [7] and can be calculated as follows:
with the static gain Kqv . The voltage deviation ΔV is calculated from the filtered
terminal voltage Vt filt and the reference voltage Vref0 according to:
where Vref0 is determined from the pre-fault condition. The DVS capability is
implemented in the REEC B 1 part of Fig. 3.4 and activated only if a voltage dip
in the system is detected, i.e., Voltage dip = 1. In this control mode, the static
gain is set to Kqv = 4 pu, as shown in Fig. 3.9. This value gives a higher sensitivity
for the additional reactive current in terms of voltage deviations, compared to the
default value of 2 pu.
1.0
Additional reactive current [pu]
0.5
Inductive
0.0
Capacitive Kqv = 4 pu
−0.5
−1.0
−0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5
Voltage deviation [pu]
Figure 3.9: Additional reactive current injection for dynamic voltage support [7]
42 Chapter 3. Modelling and control of photovoltaic systems
Very slow recovery (10 %/s): This mode represents a recovery rate of rrpwr =
10 %/s to restore the active current of the PV plant. The value of the active current
recovery rate is widely used in the United States of America [14].
Slow recovery (20 %/s): This control mode describes an active current recovery
rate of rrpwr = 20 %/s. The value is the lower limit defined by the German grid
code for the high voltage level [7].
Medium recovery (100 %/s): This mode defines a recovery rate of rrpwr =
100 %/s to restore the active current of the PV system. The value is the lower limit
defined by the grid code in Great Britain for faults with a duration of more than
140 ms [90].
Fast recovery (1000 %/s): This control mode specifies a recovery rate of rrpwr =
1000 %/s to restore the active current. The value is the lower limit defined by the
grid code in Great Britain for faults with a duration of less than 140 ms [90].
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
Local reactive power control (Local Q ctr): This mode allows local constant
reactive power control using the feed-forward loop in the REEC B 2 part of Fig. 3.4
with QFlag = 0. The reactive power is kept at the value given by the initial power
flow computation.
Local voltage control (Local V ctr): This mode enables local constant voltage
control of the PV system using the second PI controller in the REEC B 2 part of
Fig. 3.4 with QFlag = 1 and VFlag = 0. The voltage is kept at its initial power
flow value.
Plant-level reactive power control (Plant Q ctr): This mode allows constant
reactive power control of the PV system using the lower path in the REPC A 1
part of Fig. 3.5 with RefFlag = 0 and the feed-forward loop in the REEC B 2 part
of Fig. 3.4 setting QFlag = 0.
Plant-level voltage control (Plant V ctr): This mode enables constant voltage
control of the PV system using the upper path in the REPC A 1 part of Fig. 3.5
with RefFlag = 1 and the feed-forward loop in the REEC B 2 part of Fig. 3.4 with
QFlag = 0.
Case
Local Q ctr
Plant Q ctr
Local V ctr
Plant V ctr
No LVRT
1000 %/s
100 %/s
10 %/s
20 %/s
A1 × × ×
A2 × × ×
A3 × × ×
B1 × × ×
B2 × × ×
B3 × × ×
C1 × × ×
C2 × × ×
D1 × × × ×
D2 × × × ×
D3 × × × ×
D4 × × × ×
46 Chapter 3. Modelling and control of photovoltaic systems
3.4 Parameters
In general, model parameters are crucial for power system dynamic studies. Be-
cause many grid codes do not define the detailed specification or characteristic of
the inverter control, the model parameters could be different depending on the
manufacturer of the inverter. Even if the model parameters of one inverter can be
identified through field validation, it is almost infeasible to identify the parameters
of all inverters connected to the power system. Therefore, one of the main contri-
butions of this thesis is to provide a set of typical model parameters. In that sense,
typical parameters represent the average behaviour of presently required control
modes and settings according to international guidelines and grid codes. In par-
ticular, the WECC generic PV system model consists of 75 model parameters in
total. The identification process and the specific parameter values for the REGC A,
REEC B and REPC A model are presented in the sequel.
PV
3.5.3 Contingency
The contingency that is considered for the model validation, can be characterised
by the fault location, fault impedance, fault type, and fault duration [92]. The fault
is emulated using a reference voltage dip, as depicted in Fig. 3.12. This voltage
dip is created by the voltage source, which represents the external grid in Fig. 3.11.
The dynamic behaviour of the PV system model is tested in response to this voltage
dip. The details to the reference voltage dip are outlined in the sequel.
Fault location: The reference voltage dip represents a fault located in the external
grid, as shown in Fig. 3.11.
Fault duration: For the validation of the PV system model, a fault duration of
0.1 s is considered, as depicted in Fig. 3.12. The fault occurs at t = 0.0 s and is
cleared at t = 0.1 s. According to the LVRT characteristic shown in Fig. 3.8, the
PV system needs to stay connected to the grid during the short circuit and ride
through the fault.
1.0
0.8
Voltage [pu]
0.6
0.4
0.2
0.0
3.6 Validation
The implemented WECC generic PV system model, as presented in Section 3.2, is
validated against the EPRI written REMV tool [93] using the test system described
in Section 3.5. The REMV tool represents the WECC specifications of the generic
PV system model. In general, this tool is applied to validate simulations against
field measurements of a PV park. In this work, the tool is used to validate the
PV system model implemented in DIgSILENT PowerFactory. It is important to
mention that the REMV tool is successfully used for the validation of real measure-
ments by numerous manufacturers [35, 94, 95]. The REMV tool provides active and
reactive power reference signals. The focus of this work are short-term dynamics in
the order of a few seconds up to several seconds. Therefore, current signals of the
PV system are more appropriate instead of power signals. Due to this fact, current
signals are calculated out of power signals using the corresponding voltage. This
conversion could lead to minor deviations between the REMV tool and PowerFac-
tory. In addition, deviations are likely because of different numerical integration
methods of the software simulation tools, as presented in Section 2.3.3. Moreover,
the modifications made within the model implementation, as outlined in Section 3.2,
could also lead to minor deviations. It should be noted that model validation is
generally difficult to quantify [86]. Mathematical norms, such as the sum of squared
error, the Euclidean norm, or the root mean square error, can serve as useful metrics
that are also used, to some extent, for model reduction techniques [96]. However,
the application of these norms are out of scope in this work. The validation of the
WECC generic PV system model, considering the control modes given in Table 3.4,
is presented in detail in the following sections.
Future work: Future work includes the validation of additional control modes
and parameters of the WECC generic PV system model. In this context, negative-
sequence control for unbalanced system conditions is of interest. Moreover, the
WECC generic PV system model is highly non-linear as it includes freeze functions
and deadbands. Therefore, the model is not necessarily appropriate for small-signal
stability analysis. However, if linearisation and eigenvalue analysis is of interest,
closed-form equations of the system should be developed based on strict linearisation
of the non-linear model. Furthermore, in weak systems the model may not be
adequate anymore to capture all relevant power system phenomena. To overcome
this problem, the inner-current control loops as well as detailed PLL modelling
should be considered. Moreover, the generator/converter model implementation as
a current source poses numerical challenges for state-of-the-art software simulation
tools, which can be solved using a voltage source model. Grid-forming control of
PV systems is also of interest, as it seems a promising method to improve power
system stability.
53
4.2.1 Overview
The single-line diagram of the one-load infinite-bus test system is depicted in Fig. 4.1.
The term “one-load” refers to the one composite load equivalent connected to bus 5,
which represents the distribution system. In order to keep the test system as sim-
ple as possible, while retaining the essential of voltage instability phenomena, the
following modelling assumptions are made, as presented in the sequel. It should be
noted that the load and the PV system model are of particular interest. Therefore,
these models are detailed in Sections 4.2.2 and 4.2.3, respectively.
Exponential
load
External Transmission Step-down
grid line 1 transformer
Shunt
Bus 6 capacitor Large induction
220 kV IM
motor load
Bus 7
20 kV
Interconnection
transmission line
Single-converter equivalent of large-
scale photovoltaic system (Fig. 4.4)
Bus 5 (PCC)
220 kV
External grid: The external system corresponds to an infinite bus, which is rep-
resented by a voltage source of constant voltage magnitude and constant frequency,
i.e., 220 kV and 50 Hz, respectively. It should be noted that the distinction between
short-term voltage instability, which is related to induction motors, and transient
instability, which is related to synchronous generators, is not always clear, and as-
pects of both phenomena may exist, as outlined in Section 2.2.5. Therefore, the
detailed dynamic characteristics of the synchronous generators that are included in
the external grid, are neglected in order to focus on the basic mechanism of short-
term voltage stability. This assumption is usually less critical for short-term voltage
stability.
Shunt capacitor: The shunt capacitor is connected to the load bus in order to
compensate the aggregated load. The value of the shunt capacitor is fix and adjusted
to achieve a voltage value of 1.0 pu at the load bus (bus 5) considering normal
operating conditions.
56 Chapter 4. Stability analysis of a one-load infinite-bus system
Static part: The static part of the load is represented by the well-known exponen-
tial load model that is defined according to:
α
Vbus
Pexp = P0 (4.1)
V0
β
Vbus
Qexp = Q0 (4.2)
V0
where Pexp and Qexp are the active and reactive power, respectively, consumed by
the load at the bus voltage Vbus . P0 and Q0 are the active and reactive power,
respectively, under the reference voltage V0 , obtained from the initial operating
conditions. Three particular cases for the load exponents are noteworthy:
Other exponents can be used to represent the aggregate effect of different load
components. Since the dynamic part is represented explicitly, a constant admittance
is assumed. The parameter values of the static load part are shown in Table C.2 of
Appendix C. It should be noted that the performed power system dynamic studies
consider rather small frequency deviations. Therefore, the frequency dependency of
loads is not addressed in this work.
4.2. Test system 57
Dynamic part: The dynamic part consists of two equivalent induction motor
loads. The induction motor is a crucial component in short-term voltage stability
and FIDVR studies [52], because:
• it is a fast restoring load in the time frame of a second;
• it is a low power factor load with a high demand of reactive power; and
• it is prone to stalling, when voltage is significantly depressed.
In power system dynamic studies, aggregated induction motor models are usually
considered, i.e., one motor representing a large number of similar motors fed by
the same substation [52]. Because of significantly different characteristics, it may
be desirable to model equivalents for both small and large induction motors [101].
Therefore, the dynamic part of the load model consists of two equivalent, single-cage
induction motors. Each equivalent represents 15 % of the load consumption, i.e.,
15 % of small and 15 % of large industrial motors, which is in total 30 % of load
consumption. Typical parameter values of the small and large induction motors
are considered and shown in Tables C.3 and C.4 of Appendix C, respectively. The
values are based on the well-known report [101]. Moreover, a compensation shunt
capacitor is connected in parallel to the induction motor in order to match the
reactive power consumption of the entire load in steady state. Each equivalent
induction motor is represented by a third-order model. The induction motor model
uses the rotor winding fluxes and the rotor speed as state variables. Assuming a
three-phase induction motor with constant resistance, the well-known equivalent
circuit is shown in Fig. 4.2. The induction motor rotates at a speed ωimr = ωims
determined by the induction motor slip, according to:
ωims − ωimr
sim = (4.3)
ωims
with the speed of the stator and rotor ωims and ωimr , respectively. In Fig. 4.2, the
induction motor stator and rotor leakage reactances are denoted as Xims and Ximr ,
respectively, whereas the magnetising reactance is denoted as Ximm . The induc-
tion motor stator and rotor resistances are denoted as Rims and Rimr , respectively.
The terminal voltage of the induction motor is referred to as Vim . Considering the
The differential equation of the rotor motion dynamics has the form of:
dωimr
2Him = Te (Vim , ωimr ) − Tm (ωimr ) (4.6)
dt
with the induction motor inertia constant Him and the electrical and mechanical
torques Te and Tm , respectively. The mechanical torque is assumed constant with:
Tm (ωimr ) = T0 (4.7)
where T0 is the torque value determined from the initial power flow computation, as-
suming the motor is operating in steady state. The characteristic given in Eq. (4.5),
is illustrated in Fig. 4.3 for different terminal voltages Vim . The stable equilibria
are the intersection points close to sim = 0 pu, where the mechanical torque Tm is
3.0
Te (Vim = 1.0 pu)
Te (Vim = 0.75 pu)
2.5 Te (Vim = 0.5 pu)
Tm
2.0
Torque [pu]
Unstable
1.5
1.0
0.5
Stable
0.0
1.0 0.8 0.6 0.4 0.2 0.0
Slip [pu]
equal to the electrical torque Te . If a voltage dip occurs, e.g., to Vim = 0.75 pu or
Vim = 0.5 pu, the electrical torque is reduced and the induction motor decelerates.
As long as the mechanical torque is less or equal than the electrical torque, the
motor is attracted by its stable equilibrium and able to reaccelerate. However,
if the mechanical torque is larger than the electrical torque, the induction motor
decelerates beyond the unstable equilibrium and stalls. The lower the voltage at
the terminals of the induction motor or the longer the fault duration, the higher
the risk of motor stalling.
PV
4.2.5 Contingency
The contingency that is considered to study the impact of PV system control on
power system stability, specifically short-term voltage stability and delayed voltage
recovery, can be characterised by the fault location, fault impedance, fault type,
and fault duration [92], as detailed in the sequel.
Fault location: The fault is located on transmission line 2, as shown in Fig. 4.1.
It takes place exactly in the middle of the line between bus 5 and bus 6, i.e., 50 %
of the line length. Therefore, the fault location highly influences the power transfer
from the external grid to the load center.
Fault type: The fault type is a three-phase short circuit that is considered for the
analysis. This type of fault leads to a severe voltage sag that, again, highly impacts
the operation of the large-scale PV system as well as the small and large induction
motor load.
Fault duration: The last characteristic of the contingency is the fault duration.
For the stability analysis of the one-load infinite-bus system, a fault duration of 0.1 s
is considered. The fault is cleared by opening the transmission line, which remains
open. The fault duration is typical for protection systems according to international
grid codes. Moreover, the large-scale PV system needs to stay connected to the grid
during the short circuit and ride through the fault. This requirement is defined by
the LVRT characteristic shown in Fig. 3.8.
4.3. Case studies 63
1.0
0.8
Voltage [pu]
0.6
0.2
A1
A2
0.0 A3
Figure 4.5: Impact of low voltage ride-through and dynamic voltage support
Voltage at bus 5 (point of common coupling)
1.0
0.8
WECC criterion
Voltage [pu]
0.6
0.4
0.2
A1
A2
0.0 A3
Figure 4.6: Impact of low voltage ride-through and dynamic voltage support
Voltage at bus 7 (load bus)
4.3. Case studies 65
0.0
−0.2
Active current [pu]
−0.4
A1
−0.6 A2
A3
−0.8
−1.0
Figure 4.7: Impact of low voltage ride-through and dynamic voltage support
Active current of photovoltaic system connected to bus 5
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
A1
A2
−1.0 A3
Figure 4.8: Impact of low voltage ride-through and dynamic voltage support
Reactive current of photovoltaic system connected to bus 5
66 Chapter 4. Stability analysis of a one-load infinite-bus system
1.0
0.8
0.6
Slip [pu]
A1
0.4 A2
A3
0.2
0.0
Figure 4.9: Impact of low voltage ride-through and dynamic voltage support
Slip of small induction motor load connected to bus 7
1.0
0.8
0.6
Slip [pu]
A1
0.4 A2
A3
0.2
0.0
Figure 4.10: Impact of low voltage ride-through and dynamic voltage support
Slip of large induction motor load connected to bus 7
4.3. Case studies 67
The stability margin is determined in terms of the CCT and the results are
shown in Table 4.2. The best dynamic performance is highlighted in grey. The
CCT is significantly increased from 0.07 s to 0.16 s comparing the disconnection of
the PV system (A1) with the LVRT and DVS capability of the PV system (A3).
Considering the blocking mode of the PV system (A2), the same CCT is calculated
as for the disconnection case (A1), i.e., 0.07 s.
The KL divergence measure is applied to analyse the voltage recovery and give
information about the effectiveness of short-term voltage controls. The results are
shown in Table 4.2. In the same manner, the KL divergence measure K7 that is
calculated from the voltage at the load bus (bus 7), shows the best performance
using the LVRT and DVS capability of the PV system (A3), which yields a value
of 3.49, as highlighted in grey. For the disconnection (A1) and the blocking mode
(A2) of the PV system, the same K7 value of 102.84 each, is obtained. Therefore,
no distinction can be made between these two short-term control modes of the PV
system.
The last metric that is used to evaluate the impact of LVRT and DVS, is the
VRI and the results are shown in Table 4.2. The best dynamic performance is
highlighted in grey. The same tendency as for the other metrics is observed. The
disconnection of the PV system (A1) results in an unacceptable V7 value of −0.13,
considering the voltage recovery at the load bus (bus 7). On the other hand, the best
performance is achieved using the LVRT and DVS capability of the PV system (A3),
with a VRI of 0.58. If only the LVRT capability of the PV system is incorporated
(A2), the calculation of the VRI yields −0.12. Compared to the KL divergence
measure, the disconnection of the PV system (A1) and the blocking mode (A2) can
be distinguished using the VRI, which shows the superiority of this metric.
Table 4.2: Metrics of low voltage ride-through and dynamic voltage support
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6
1.00
0.95
0.4
0.90
0.85
0.2 0.80 B1
B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
1.0
0.6 1.00
0.95
0.4 0.90
0.85
0.80
0.2 B1
0.75
B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0 B1
B2
A3
−0.2 B3
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4 0.05
0.00
−0.6
−0.05
−0.10
−0.8 B1
−0.15 B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
−1.0 B3
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
The CCT is calculated in order to determine the stability margin of the system.
The results are presented in Table 4.3. The best dynamic performance is highlighted
in grey. In general, the same tendency is observed as for the time-domain results.
The best performance is achieved with a medium and fast active current recovery
rate of 100 %/s (A3) and 1000 %/s (B3), respectively, which leads to a CCT of
0.16 s. Slower ramp rates, such as 10 %/s (B1) and 20 %/s (B2), lead to a smaller
CCT of 0.15 s, respectively.
The delayed voltage recovery is analysed using the KL divergence measure and
the results are given in Table 4.3. Again, the same trend is observed as for the CCT
and the time-domain results. The fastest active current ramp rate of 1000 %/s
(B3) yields the best KL divergence measure K7 that is calculated from the load bus
voltage (bus 7), with a value of 3.13, as highlighted in grey. On the other hand, the
lowest performance is reached if the PV system injects the active current with a
ramp rate of 10 %/s (B1), which results in a K7 value of 6.94. Furthermore, the KL
divergence measure for the slow active current recovery of 20 %/s (B2) is higher,
with a value of 4.82, compared to the medium active current recovery of 100 %/s
(A3), with a value of 3.49. Hence, the latter results in a better quality of voltage
recovery.
Finally, the impact of the active current recovery rate on the phenomenon of
FIDVR is assessed using the newly defined VRI and the results are shown in Ta-
ble 4.3. The best dynamic performance is highlighted in grey. Generally speaking,
a similar tendency is observed as for the aforementioned metrics. In the same man-
ner, the fastest active current recovery rate of 1000 %/s (B3) results in a high V7
value of 0.60, considering the voltage recovery at the load bus (bus 7). The worst
VRI is calculated considering the very slow active current recovery rate of 10 %/s,
which yields a V7 value of 0.15. If only a slow (B2) or medium (A3) active current
recovery rate of the PV system is incorporated, the calculation of the VRI yields
0.26 or 0.58, respectively.
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6
1.00
0.95
0.4
0.90
0.85
0.2 0.80
A3
0.0 0.5 1.0 1.5 2.0 2.5 3.0 C1
0.0 C2
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
1.0
0.6 1.00
0.95
0.4 0.90
0.85
0.80
0.2
0.75 A3
0.0 0.5 1.0 1.5 2.0 2.5 3.0 C1
0.0 C2
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0 A3
C1
C2
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4
0.00
−0.05
−0.6
−0.10
−0.15
−0.8 −0.20
A3
0.0 0.5 1.0 1.5 2.0 2.5 3.0 C1
−1.0 C2
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
The stability margin of the system is determined using the CCT and the results
are shown in Table 4.4. The best dynamic performance is highlighted in grey.
Compared to the time-domain simulations discussed before, the same trend for the
CCT can be observed. The best dynamic performance is achieved using either local
constant voltage control (C1) or local coordinated reactive power/voltage control
(C2), leading to CCT values of 0.17 s each. Considering local constant reactive
power control (A3), leads to the worst performance with a CCT value of 0.16 s.
The KL divergence measure is applied in order to assess the impact of the local
voltage control methods of the PV system on the delayed voltage recovery. The
results are given in Table 4.4. Similar to the analysis methods applied before, the
same trend is observed. Local constant voltage control (C1) of the PV system
results in the best KL divergence measure K7 that is calculated from the load bus
voltage (bus 7), with a value of 2.32, as highlighted in grey. On the other hand,
the lowest performance is reached if the PV system applies local constant reactive
power control (A3), which results in a K7 value of 3.49. The combination of both
local controls, i.e., coordinated reactive power/voltage control (C2), leads to a KL
divergence measure K7 of 2.41.
The last metric that is used to analyse the impact of local voltage control strate-
gies is the VRI. The results are given in Table 4.4 and the best performance is
highlighted in grey. In the same manner, the local constant voltage control (C1)
results in the best VRI with a V7 value of 0.95, considering the voltage recovery
at the load bus (bus 7). The worst VRI is calculated considering local constant
reactive power control (A3) of the PV system, which yields a V7 value of 0.58. If
coordinated reactive power/voltage control (C2) is applied, the calculation of the
VRI yields 0.75.
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6
1.00
0.95
0.4
0.90
0.85
0.2 0.80 D1
D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
1.0
0.6 1.00
0.95
0.4 0.90
0.85
0.80
0.2 D1
0.75 D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
0.0 D1
D2
D3
−0.2 D4
−0.90
Active current [pu]
−0.4
−0.95
−0.6
−1.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4 0.05
0.00
−0.6
−0.05
−0.10
−0.8 D1
−0.15 D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
−1.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
The calculation of the CCT determines the stability margin of the system. The
results are given in Table 4.5 and the best dynamic performance is highlighted in
grey. Compared to the time-domain simulations discussed before, a similar trend
for the CCT can be observed. The best CCT is achieved considering either plant-
level constant reactive power and local coordinated reactive power/voltage control
(D3) or plant-level constant voltage and local coordinated reactive power/voltage
control (D4), leading to values of 0.17 s, respectively. In contrast, using plant-level
constant reactive power control (D1) or plant-level constant voltage control (D2) of
the PV system, the CCT is worsened with values of 0.16 s, respectively.
The impact on FIDVR is analysed using the KL divergence measure. The re-
sults are shown in Table 4.5 and the best performance is highlighted in grey. Plant-
level constant reactive power and local coordinated reactive power/voltage control
(D3) or plant-level constant voltage and local coordinated reactive power/voltage
control (D4), lead to the same K7 value of 2.41 each. Hence, no distinction can
be made between these two control modes, which is a disadvantage of the KL
divergence measure. The lowest performance is reached if plant-level constant re-
active power control (D1) is used, leading to a value of 3.49. It should be noted
that local constant reactive power control (A3), as presented in Section 4.3.3, and
plant-level constant reactive power control (D1), lead to the same KL divergence
measures. The same correlation exists between coordinated reactive power/voltage
control (C2) and plant-level constant reactive power and local coordinated reactive
power/voltage control (D3). The similarities are caused due to the same behaviour
of short-term controls of the PV system.
Finally, the results of the VRI are given in Table 4.5. The best VRI is achieved
considering plant-level constant voltage and local coordinated reactive power/voltage
control (D4), which results in a V7 value of 0.76, as highlighted in grey. Compared
to the KL divergence measure, the former control mode (D4) and plant-level con-
stant reactive power and local coordinated reactive power/voltage control (D3) can
be distinguished using the VRI, which reveals the superiority of this metric. The
worst performance value is calculated if plant-level constant reactive power control
(D1) is applied, leading to a V7 value of 0.58. Note that the same relations between
local and plant-level voltage control are observed for the VRI, as described for the
KL divergence measure in the paragraph before.
Future work: Future work considers variations of the load model that is used
in the one-load infinite-bus test system. In particular, the mechanical torque of
the induction motor load can be modified to, e.g., quadratic, or a combination of
different torque characteristics to form composite mechanical loads. Another aspect
that is of interest for future research, is the investigation of additional operating
points in the test system considering a higher PV penetration level. Moreover, the
studied contingency could be varied in terms of fault location, fault impedance, fault
type or fault duration. Other stability criteria, such as long-term voltage stability
or small-signal stability, are also of interest.
81
5.2.1 Overview
The single-line diagram of the Nordic test system is depicted in Fig. 5.1. The system
consists of four areas, namely:
• “North” with hydro generation and some load,
• “Central” with thermal generation and the load center,
• “Equiv” connected to “North”, which includes a simple equivalent of an ex-
ternal system, and
• “South” with thermal generation, rather loosely connected to the rest of the
system.
The test network is composed by 32 transmission and subtransmission buses with
nominal voltages of 400 kV, 220 kV and 130 kV. In addition, 22 buses are at distri-
bution level with a voltage of 20 kV. The distribution systems are represented by
aggregated loads, as described in detail in Section 5.2.2. It should be noted that
the distribution system can be represented explicitly, if, for instance, interactions
between the transmission and distribution system are studied, and dispersed gen-
eration is considered [109, 110]. The nominal frequency of the system is 50 Hz.
The frequency is controlled through the speed governors of the hydro generators
in the “North” and “Equiv” areas only. In order to keep the test system as sim-
ple as possible, while retaining the essential of voltage instability phenomena, the
following modelling assumptions are made. It should be noted that the details of
the load and the PV system model are of particular interest, and therefore given
in Sections 5.2.2 and 5.2.4, respectively. Additional information about the general
system description can be found in report [98].
5.2. Test system 83
g19 g9 g2
71 11 13
12
EQUIV. 4012 1012 1014
4072
g10 g1 g3
g5 g11
72 g20 NORTH
400 kV
1021 1022 4022 4021
220 kV
g12 130 kV
22
g4
2032 2031 4031 4032
32 31
g8
42
PV
4042
4041
CENTRAL
g14
6&
41 g7 PV PV
g13 4044 4043 4046
4
3
4061 43
1043 g6 46
PV
PV PV
1044
6161
1042
g17 PV
1041 1045 2 g15 g15b
62 PV PV
47
PV 1 5 g16 g16b
g18
g18 g18b 63
63 SOUTH
Figure 5.1: Nordic test system [98]
84 Chapter 5. Stability analysis of a multi-machine system
Shunt capacitors and inductors: Shunt capacitors and inductors are connected
to the system in order to compensate the reactive power and control the voltage.
Compared to report [98], the values of the shunt capacitors and inductors need to
be adjusted due to the substitution of synchronous generators, which is presented
in Section 5.2.3.
5.2. Test system 85
PV
Further details to operating point B and the necessary modifications can be found
in report [98]. Considering this operating point, the demand of the load center in
the “Central” area is mainly covered by the generation in the “North” area, which
leads to large transfers. An overview of the load and generation in the Nordic
test system with respect to the area, is shown in Table 5.3. Please note that
the consumer-oriented sign convention is used in this work, i.e., consumed active
power is positive and generated active power is negative. In particular, the load
consumption is detailed in Table D.6 of Appendix D. Furthermore, all PV systems
are operated at unity power factor and their generation is detailed in Table D.7
of Appendix D. In order to match operating point B as described in report [98],
a new shunt is connected to bus 1042 and the other existing shunts are adjusted,
as shown in Table D.8 of Appendix D. Considering the load and generation shown
in Table 5.3, the PV penetration level is calculated according to Eq. (4.8), which
leads to 15 % for the entire system. For the “Central” area in particular, a PV
penetration level of 41 % is reached.
Generation
[MW] Load
[MW]
Area PV pen [%]
PSG PPV Pexp Psim + Plim
North −4628.5 0 826 354 0
Central −1680 −1170 4333 1857 41
South −1050 −540 973 417 34
Equiv −2437.4 0 2300 0 0
Total −9795.9 −1710 8432 2628 15
88 Chapter 5. Stability analysis of a multi-machine system
5.2.6 Contingencies
Two contingencies, named fault 1 and fault 2, are considered to investigate the im-
pact of PV system control on power system stability, specifically short-term voltage
stability and delayed voltage recovery. The contingencies can be characterised by
the fault location, fault impedance, fault type, and fault duration [92], as detailed
in the sequel.
Fault impedance: Fault 1 and fault 2 have a fault impedance of zero, which
corresponds to a solid short circuit. This impedance leads to a severe voltage
dip that highly influences the dynamic behaviour of PV systems. Moreover, the
resulting voltage dip affects the operation of small and large induction motor loads
as well as synchronous generators.
Fault type: Fault 1 and fault 2 are three-phase short circuits. This type of fault
leads to a severe voltage sag that, again, highly impacts the operation of large-
scale PV systems, small and large induction motor loads as well as synchronous
generators.
Fault duration: Fault 1 and fault 2 have a fault duration of 0.1 s, respectively. The
faults are cleared by opening the corresponding transmission line, which remains
open. The fault duration is typical for protection systems according to international
grid codes. Moreover, the large-scale PV systems need to stay connected to the grid
during the short circuit and ride through the fault. This requirement is defined by
the LVRT characteristic shown in Fig. 3.8.
5.3. Case studies 89
1.0
0.8
Voltage [pu]
0.6
0.4
WECC criterion
0.2
V4 (fault 1)
0.0 V2 (fault 2)
Fault 1 Fault 2
Case
tCCT1 [s] K4 [–] V4 [–] tCCT2 [s] K2 [–] V2 [–]
Base 0.15 1.47 0.74 0.12 0.24 0.87
90 Chapter 5. Stability analysis of a multi-machine system
1.0
0.8
Voltage [pu]
0.6
0.4
Figure 5.4: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Voltage at bus 1042 (point of common coupling)
1.0
0.8
Voltage [pu]
0.6
0.4
WECC criterion
0.2
A1
A2
0.0 A3
Figure 5.5: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Voltage at bus 2 (load bus)
92 Chapter 5. Stability analysis of a multi-machine system
0.0
A1
−0.2 A2
A3
Active current [pu]
−0.4
−0.6
−0.8
−1.0
Figure 5.6: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Active current of photovoltaic system connected to bus 1042
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
A1
A2
−1.0 A3
Figure 5.7: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Reactive current of photovoltaic system connected to bus 1042
5.3. Case studies 93
1.0
0.8
0.6
Slip [pu]
A1
0.4 A2
A3
0.2
0.0
Figure 5.8: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Slip of small induction motor load connected to bus 2
1.0
0.8
0.6
Slip [pu]
A1
0.4 A2
A3
0.2
0.0
Figure 5.9: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Slip of large induction motor load connected to bus 2
94 Chapter 5. Stability analysis of a multi-machine system
200.0
100.0
Rotor angle [◦ ]
0.0
−100.0
A1
A2
A3
−200.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time [s]
Figure 5.10: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Rotor angle of synchronous generator 16
1.4 A1
A2
1.2 A3
t = 0.5 s
Active power [pu]
1.0
0.8
0.6
0.4
0.2 t=0s
t = 0.1 s
0.0
−40.0 −20.0 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0
Rotor angle [◦ ]
Figure 5.11: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Active power as function of rotor angle of synchronous generator 16
5.3. Case studies 95
The stability margin is determined using the CCT and the results are shown in
Table 5.5. The best dynamic performance is highlighted in grey. Considering fault 2,
for instance, the CCT is significantly increased from 0.05 s to 0.12 s comparing
the disconnection of PV systems (A1) with the LVRT and DVS capability (A3).
The blocking mode of PV systems (A2) leads to a CCT of 0.09 s, which is higher
compared to the disconnection case (A1). The base case without PV systems leads
to the same stability margin, i.e., a CCT of 0.12 s. The calculation of the CCT for
fault 1 reveals the same tendency. The best dynamic performance is achieved using
the LVRT and DVS capability of PV plants (A3), which leads to a CCT of 0.12 s.
The worst CCT is calculated if PV systems disconnect (A1), which yields a value
of 0.02 s.
The KL divergence measure is applied to analyse the voltage recovery and give
information about the effectiveness of short-term voltage controls of PV systems.
The results are shown in Table 5.5 and the best dynamic performance is highlighted
in grey. In the same manner, the KL divergence measure K2 that is calculated
from the voltage at the load bus (bus 2), shows the best performance using the
LVRT and DVS capability of PV systems (A3), which yields a value of 0.84. The
lowest performance is reached if PV systems do not have any LVRT capability (A1),
with a K2 value of 159.03. A better voltage recovery is obtained considering the
blocking mode (A2), with a KL divergence measure of 36.12. The base case without
PV systems yields a value of 0.24 due to the higher voltage support capability of
synchronous generators during the fault-on period. The same trend, as described
before, can be observed for fault 1.
Finally, the impact of the LVRT and DVS capability on the phenomenon of
FIDVR is analysed using the VRI and the results are presented in Table 5.5. The
best dynamic performance is highlighted in grey. The same tendency as for the
other metrics is observed. With respect to fault 2, for instance, the disconnection
of PV systems (A1) results in an unacceptable V2 value of −0.28, considering the
voltage recovery at bus 2. On the other hand, the best performance is achieved
using the LVRT and DVS capability of PV systems (A3), with a VRI of 0.64. If
only the LVRT capability of PV systems is incorporated (A2), the calculation of
the VRI yields −0.02. The base case leads to a VRI of 0.87 due to the higher
voltage support capability of synchronous generators. Compared to fault 2, the
same tendency of the calculated metrics can be observed for fault 1.
Table 5.5: Metrics of low voltage ride-through and dynamic voltage support
Fault 1 Fault 2
Case
tCCT1 [s] K4 [–] V4 [–] tCCT2 [s] K2 [–] V2 [–]
A1 0.02 55.71 −0.04 0.05 159.03 −0.28
A2 0.05 31.53 −0.01 0.09 36.12 −0.02
A3 0.12 2.82 0.66 0.12 0.84 0.64
Base 0.15 1.47 0.74 0.12 0.24 0.87
96 Chapter 5. Stability analysis of a multi-machine system
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6
1.05
1.00
0.4
0.95
0.90 B1
0.2
0.85 B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
1.0
0.8
WECC criterion
Voltage [pu]
0.6 1.05
1.00
0.4 0.95
0.90
0.2 0.85 B1
B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
0.0 B1
B2
A3
−0.2 B3
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
0.00
−0.05
−0.6
−0.10
−0.15
−0.8 −0.20 B1
B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
−1.0 B3
Apart from the voltage dynamics discussed before, the active current recovery
rate of PV systems also influences frequency dynamics. Therefore, the frequency of
the center of inertia of the Nordic test system is determined. This frequency is a
weighted sum of all synchronous machine speeds in the system. The frequency of
the center of inertia is depicted in Fig. 5.16. At 0.6 s, the frequency rises due to the
load sensitivity, i.e., a lower voltage leads to a lower active power consumption, as
shown in Eq. (4.1). However, at 2.2 s, the active current ramp rates of 10 %/s (B1)
and 20 %/s (B2) lead to a frequency nadir of 49.8 Hz and 49.85 Hz, respectively.
On the other hand, considering an active current ramp rate of 1000 %/s (B3),
the frequency rises to a value higher than 50.2 Hz at 0.8 s and 1.8 s, respectively.
According to old German grid code requirements, PV systems would disconnect
at this frequency and cause a further frequency drop in the system, which is well
known as the 50.2 Hz problem [12]. Therefore, the medium active current recovery
rate of 100 %/s (A3) shows the best frequency performance. If both dynamics are
taken into account, i.e., voltage and frequency dynamics, as depicted in Figs. 5.13
and 5.16, respectively, the best performance is achieved with the medium active
current recovery rate of 100 %/s (A3).
50.2
50.1
Frequency [Hz]
50.0
49.9 B1
B2
A3
B3
49.8
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Time [s]
The CCT is calculated in order to determine the stability margin of the system.
The results are presented in Table 5.6 and the best dynamic performance is high-
lighted in grey. With respect to fault 1, for instance, the CCT is not influenced
by the different ramp rates, leading to 0.12 s for all cases. The base case leads to
a CCT of 0.15 s. Considering fault 2, the same tendency is observed as for the
time-domain results. The best performance is achieved with a medium and fast
active current recovery rate of 100 %/s (A3) and 1000 %/s (B3), which lead to a
CCT of 0.12 s, respectively. Slower ramp rates, such as 10 %/s (B1) and 20 %/s
(B2), lead to a smaller CCT of 0.11 s each.
The voltage recovery is analysed using the KL divergence measure and the results
are given in Table 5.6. The best performance is highlighted in grey. Again, the same
trend is observed as for the CCT and the time-domain results. The fastest active
current ramp rate of 1000 %/s (B3) yields the best KL divergence measure K4 that
is calculated from the load bus voltage (bus 4), with a value of 2.10 for fault 1. The
lowest performance is reached if PV systems inject active current with a ramp rate
of 10 %/s (B1), which results in a K4 value of 7.19. Furthermore, the KL divergence
measure for the slow active current recovery of 20 %/s (B2) is higher, with a value
of 5.65, compared to the medium active current recovery of 100 %/s (A3), with
a value of 2.82. Hence, the latter results in a better quality of voltage recovery.
The base case yields a value of 1.47 due to the higher voltage support capability of
synchronous generators. The same trend is observed for fault 2.
The last metric that is used to analyse the impact of the active current recovery
rate on FIDVR, is the VRI and the results are shown in Table 5.6. The best
dynamic performance is highlighted in grey. Generally speaking, a similar tendency
is observed as for the aforementioned metrics. In the same manner, the fastest
active current recovery rate of 1000 %/s (B3) results in a high V4 value of 0.68,
considering the voltage recovery at bus 4 for fault 1. The worst VRI is calculated
considering the very slow active current recovery rate of 10 %/s (B1), which yields
a V4 value of 0.22. If only a slow (B2) or medium (A3) active current recovery rate
of the PV system is incorporated, the calculation of the VRI yields 0.29 or 0.66,
respectively. Again, the same tendency is observed for fault 2.
Fault 1 Fault 2
Case
tCCT1 [s] K4 [–] V4 [–] tCCT2 [s] K2 [–] V2 [–]
B1 0.12 7.19 0.22 0.11 1.22 0.58
B2 0.12 5.65 0.29 0.11 1.07 0.60
A3 0.12 2.82 0.66 0.12 0.84 0.64
B3 0.12 2.10 0.68 0.12 0.82 0.66
Base 0.15 1.47 0.74 0.12 0.24 0.87
5.3. Case studies 101
1.0
0.8
Voltage [pu]
0.6
0.4
1.0
0.8
Voltage [pu]
0.6
0.4
WECC criterion
0.2
A3
C1
0.0 C2
0.0
A3
−0.2 C1
C2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.4 A3
C1
0.2 C2
Reactive current [pu]
0.0
−0.2
−0.4
−0.6
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Time [s]
1.0
0.8
0.6
Slip [pu]
A3
0.4 C1
C2
0.2
0.0
1.0
0.8
0.6
Slip [pu]
A3
0.4 C1
C2
0.2
0.0
The stability margin of the system is determined using the CCT and the results
are shown in Table 5.7. The best dynamic performance is highlighted in grey.
Compared to the time-domain simulations discussed before, a different trend for
the CCT can be observed owing to the relative effects of active and reactive current
injections of PV systems. Using local constant reactive power control (A3), leads to
the best performance with a CCT of 0.12 s for fault 1. If, for the same fault, local
constant voltage control (C1) or local coordinated reactive power/voltage control
(C2) is applied, the CCT results in a value of 0.11 s, respectively. The base case
leads to a CCT of 0.15 s due to the higher voltage support capability of synchronous
generators. The same tendency is observed for fault 2.
The KL divergence measure is applied in order to assess the delayed voltage
recovery. The results are given in Table 5.7 and the best performance is highlighted
in grey. The same trend is observed as for the CCT. Local constant reactive power
control (A3) results in the best KL divergence measure K4 that is calculated from
the load bus voltage (bus 4), with a value of 2.82 for fault 1. The lowest performance
is reached, if local constant voltage control (C1) is applied, which results in a K4
value of 71.99 due to the system collapse. The combination of both local controls,
i.e., local coordinated reactive power/voltage control (C2), leads to a KL divergence
measure K4 of 3.78. Compared to the time-domain results, the KL divergence mea-
sure of local constant reactive power control (A3) is better than local coordinated
reactive power/voltage control (C2) due to the higher amount of voltage samples
that are closer to the pre-fault voltage. The same trend is observed for fault 2.
Finally, the VRI is used to analyse the impact of local voltage control strategies.
The results are given in Table 5.7 and the best performance is highlighted in grey.
The same trend is observed as for the CCT and the KL divergence measure. Local
constant reactive power control (A3) results in the best VRI with a V4 value of
0.66 for fault 1. The worst VRI is calculated considering local constant voltage
control (C1), which yields 0.02 due to the system collapse. If coordinated reactive
power/voltage control (C2) is applied, the calculation of the VRI results in 0.49,
which is worse than local constant reactive power control (A3) due to the reason
described in the paragraph before. The base case leads to a VRI of 0.74. Again,
the same trend is observed for fault 2.
Table 5.7: Metrics of local voltage control
Fault 1 Fault 2
Case
tCCT1 [s] K4 [–] V4 [–] tCCT2 [s] K2 [–] V2 [–]
A3 0.12 2.82 0.66 0.12 0.84 0.64
C1 0.11 71.99 0.02 0.09 62.44 0.04
C2 0.11 3.78 0.49 0.10 2.66 0.55
Base 0.15 1.47 0.74 0.12 0.24 0.87
106 Chapter 5. Stability analysis of a multi-machine system
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6
1.05
0.4
1.00
0.2 0.95 D1
D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
1.0
0.8
WECC criterion
Voltage [pu]
0.6 1.05
0.4 1.00
0.95
0.2 D1
D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
0.0 D1
D2
D3
−0.2 D4
−0.85
Active current [pu]
−0.4
−0.90
−0.6
−0.95
0.0 0.5 1.0 1.5 2.0 2.5 3.0
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
0.4 D1
D2
0.2 D3
D4
Reactive current [pu]
0.0
−0.2
−0.4
−0.6
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
The calculation of the CCT determines the stability margin of the system. The
results are given in Table 5.8 and the best dynamic performance is highlighted in
grey. The best CCT is achieved considering either plant-level constant reactive
power control (D1) or plant-level constant voltage control (D2), leading to values
of 0.12 s for fault 1, respectively. In contrast, using plant-level constant reactive
power and local coordinated reactive power/voltage control (D3) or plant-level con-
stant voltage and local coordinated reactive power/voltage control (D4), the CCT is
worsened with values of 0.11 s, respectively. The base case leads to 0.15 s due to the
higher voltage support capability of synchronous generators. The same tendency is
observed for fault 2.
The impact on FIDVR is analysed using the KL divergence measure. The results
are shown in Table 5.8 and the best performance is highlighted in grey. Plant-level
constant voltage control (D2) leads to the best voltage recovery, with a K4 value of
2.77 for fault 1. The lowest performance is reached if plant-level constant reactive
power and local coordinated reactive power/voltage control (D3) is used, leading to
a value of 3.78. It should be noted that local constant reactive power control (A3),
as presented in Section 5.3.4, and plant-level constant reactive power control (D1),
lead to the same KL divergence measures. The same correlation exists between local
coordinated reactive power/voltage control (C2) and plant-level constant reactive
power and local coordinated reactive power/voltage control (D3). The similarities
are caused due to the same behaviour of the short-term controls of PV systems.
The same trend is observed for fault 2.
Finally, the VRI is used to analyse FIDVR. The results are given in Table 5.8
and the best performance is highlighted in grey. Plant-level constant voltage control
(D2) leads to the best VRI, with a V4 value of 0.67 for fault 1. On the contrary, the
worst VRI is calculated if plant-level constant reactive power and local coordinated
reactive power/voltage control (D3) is applied, leading to a V4 value of 0.49. The
same trend is observed for fault 2. It should be noted that similar correlations
between local and plant-level voltage control are observed for the VRI, as described
for the KL divergence measure in the paragraph before.
Fault 1 Fault 2
Case
tCCT1 [s] K4 [–] V4 [–] tCCT2 [s] K2 [–] V2 [–]
D1 0.12 2.82 0.66 0.12 0.84 0.64
D2 0.12 2.77 0.67 0.12 0.81 0.66
D3 0.11 3.78 0.49 0.10 2.66 0.55
D4 0.11 3.67 0.50 0.10 2.19 0.60
Base 0.15 1.47 0.74 0.12 0.24 0.87
110 Chapter 5. Stability analysis of a multi-machine system
Future work: Future work should consider variations of the load model. In par-
ticular, the mechanical torque of the induction motor load can be modified to, e.g.,
quadratic, or a combination of different torque characteristics to form composite
mechanical loads. Another aspect that is of interest for future research is the inves-
tigation of additional operating points in the test system considering a higher PV
penetration level. Moreover, the studied contingencies could be varied in terms of
fault location, fault impedance, fault type or fault duration. Furthermore, the im-
pact of the plant control of PV systems on long-term voltage stability is of interest.
111
• Control: The relevant PV system control modes that impact the dynamic
performance of the power system, specifically short-term dynamics, are de-
termined. The considered control modes are the LVRT and DVS capability,
the active current recovery rate, the local voltage control and the plant-level
voltage control.
• Open source: To serve the need of academia and industry, the DIgSILENT
PowerFactory model of the WECC generic PV system is open-source software
and can be downloaded free of charge and used for any purpose.
• LVRT and DVS capability: The time-domain results show that without the
LVRT capability of PV systems, the system is unstable. Even the blocking
mode of PV systems is not sufficient to prevent short-term voltage instability.
Only the LVRT and DVS capability of PV systems helps avoiding instability
and improves FIDVR. The CCT that determines the stability margin of the
system, shows the same tendency. The CCT is significantly improved using
the LVRT and DVS capability of PV systems. In the same manner, the KL
divergence measure as well as the VRI indicate the enhancement of the voltage
recovery applying the latter control mode.
6.1. Conclusions 113
• Active current recovery rate: The time-domain results show that the best
voltage recovery is achieved using the fastest active current recovery rate of
PV systems. However, this recovery rate also leads to overfrequencies in the
system that could cause disconnection of generation. Therefore, the active
current recovery rate of PV systems should be tuned considering both, volt-
age and frequency dynamics. The CCT shows the best performance with
the fastest active current recovery rate. The same trend is observed for the
analysis of FIDVR using the KL divergence measure and the VRI.
• Local voltage control: The results for the multi-machine system with a medium
PV penetration level are different compared to the one-load infinite-bus sys-
tem with a high PV penetration level. The time-domain simulations for the
multi-machine system show that in case of local constant voltage control, PV
systems try to restore their pre-fault voltage by increasing their reactive cur-
rent. However, due to the rather stiff grid behaviour, restoring grid voltage
is impossible. The active current is reduced to zero due to the missing re-
active power limitation, and the system collapses owing to the lack of active
power produced by the PV systems. Although counterintuitive, the results
show that a high reactive power injection is not always the best to counteract
a voltage collapse. Therefore, adequate reactive power limitation is required
or a different current limit logic of the PV system needs to be selected. The
stability margin in terms of the CCT as well as the KL divergence measure
and the VRI that quantify FIDVR, show the best performance considering
local constant reactive power control. The time-domain simulations of the
one-load infinite-bus system yield the best results in case of local constant
voltage control of PV systems. The same holds true for the CCT, the KL
divergence measure and the VRI, considering this test system.
• Plant-level voltage control: If plant control of PV systems is used, the time-
domain results show the best performance using plant-level constant volt-
age and local coordinated reactive power/voltage control. This control mode
combines a fast voltage control that injects a fast reactive current to improve
FIDVR, together with a slow voltage control that slowly injects reactive power
to restore the grid voltage. It is shown that the fast voltage control leads to
a less oscillatory voltage evolution during the recovery period. The enhanced
short-term dynamics are also confirmed by the CCT, the KL divergence mea-
sure and the VRI for the one-load infinite-bus system. On the contrary, for
the multi-machine system, the latter metrics are worsened using plant-level
constant voltage and local coordinated reactive power/voltage control owing
to the rather stiff grid behaviour. This leads to a higher deviation to the
pre-fault voltage, which deteriorates the KL divergence measure and the VRI.
Finally, it is important to mention that the plant control of the PV system
model needs to be considered for short-term voltage stability and FIDVR
studies because it can significantly influence short-term dynamics.
114 Chapter 6. Conclusions and future work
• Time-domain simulation: For very high PV penetration levels and weak sys-
tem conditions, RMS time-domain simulations might not be accurate or suit-
able anymore. Instead, combined RMS and EMT, or pure EMT analysis is
required.
• CCT: The calculation of the CCT is very helpful to determine the stability
margin of the power system. However, a smaller step size for the simulations
would lead to somewhat more accurate values of the CCT.
• VRI: Different weighting functions for overvoltages, i.e., voltages higher than
the normal operating voltage, could be introduced for the calculation of the
VRI.
• Transient P–V curves: Transient P–V curves could be used to determine the
stability boundary of the load bus including PV generation for the analysis of
short-term voltage stability. It is an analytical method and the P–V curves
represent the power transfer capability to the load at a given time.
6.2. Future work 115
Stability analysis: The results of the stability analysis of the one-load infinite-
bus system and the multi-machine system, namely the Nordic test system, indicate
further directions for future research.
• Load modelling: Additional variations of the load model could be considered.
In particular, the mechanical torque of the induction motor load can be mod-
ified to, e.g., quadratic, or a combination of different torque characteristics
to form composite mechanical loads, which would affect short-term voltage
stability and FIDVR.
• Operating point: Additional operating points in the test system should be con-
sidered, e.g., PV systems that do not operate at unity power factor. Moreover,
higher PV penetration levels are also of interest.
• Contingency: The studied contingencies could be varied in terms of fault loca-
tion, fault impedance, fault type or fault duration in order to affect different
PV systems, induction motors and synchronous generators in the test system.
• Long-term voltage stability: The impact of the plant control of PV systems
on long-term voltage stability is of interest. Therefore, the test system should
include the associated controls of synchronous generators as well as the load
tap changer control of step-down transformers.
• Small-signal stability: The impact of PV system control on small-signal sta-
bility is of interest. The WECC generic PV system model is highly non-linear
due to freeze functions and deadbands. Therefore, it is not necessarily ap-
propriate for linearisation and eigenvalue analysis. However, if small-signal
stability analysis is of interest, closed-form equations of the system should be
developed based on strict linearisation of the non-linear model.
117
Appendices
119
A.2
A.2.1
xmax xmax
y 1 x y 1 x̃ x y 1 x
s s s
according to: ⎧
⎪
⎨xmax if x̃ ≥ xmax
x = x̃ if xmin < x̃ < xmax (B.3)
⎪
⎩
xmin if x̃ ≤ xmin
The third integrator shown in Fig. B.1 (c) includes anti-windup limits. The system
equation has the form of:
dx
=y (B.4)
dt
and the anti-windup limits are defined as:
dx dx
if x ≥ xmax and ≥ 0 → x = xmax and =0
dt dt
dx dx
if x ≤ xmin and ≤ 0 → x = xmin and =0 (B.5)
dt dt
dx
otherwise → =y
dt
The windup effect of the three configurations is illustrated in Fig. B.2 with y(t) =
sin(t), xmax = 0.5 and xmin = −0.5. For x̃ > xmax at t = 1.05 s the output x is
locked and remains locked until x̃ < xmax at t = 5.23 s. This is the typical windup
effect of the integrator. The anti-windup function solves this problem, i.e., x starts
decreasing at t = 3.14 s, which is the instant for which dx/dt becomes negative.
The anti-windup limiter of the PI controller in DIgSILENT PowerFactory uses a
semi-implicit formulation [111].
2.0
1.5
1.0
Signals
0.5
0.0
y(t)
−0.5 x(t) (without limits)
x(t) (windup)
x(t) (anti-windup)
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Time [s]
Freeze function: If a fault in the system occurs, the function freezes the state
variables of the control blocks that are used for normal operating conditions. The
function is activated only if the PV system turns from the normal operating condi-
tion into the fault condition, defined by the Voltage dip signal shown in Eq. (3.4).
The freeze function is implemented as follows:
A simple integrator model with anti-windup limits and the Voltage dip signal is
shown in Fig. B.3. A numerical example of the freeze function with the input signal
y(t) = sin(t) and the limits xmax = 0.5 and xmin = −0.5 is shown in Fig. B.4.
The freeze function is activated from 3 s to 5 s. During the fault-on period, i.e.,
Voltage dip = 1, the state variable x is frozen and remains at the value 0.5. At
t = 5 s, the model returns to the normal operating condition and x starts decreasing.
Voltage dip
xmax
y 1 x
s
xmin
Figure B.3: Block diagram of integrator model with anti-windup limits and freeze func-
tion
1.0
0.5
Signals
0.0
−0.5 y(t)
Voltage dip
x(t) (without freeze function)
x(t) (with freeze function)
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Time [s]
Figure B.4: Responses of integrator models with and without freeze function
124
B.2
Table B.1: Parameters of generator/converter model —————— B.2.1
Model Description Symbol Value Unit
REGC A 0 High voltage clamp logic acceleration factor Khv 2 [–]
REGC A 0 Voltage limit for high voltage clamp logic Volim 1.2 [pu]
Parameters
REGC A 0 Inverter reactive current regulator lag time constant Tgq 0.02 [s]
REGC A 0 Current limit for high voltage clamp logic Iolim −1 [pu]
REGC A 0 Maximum rate-of-change of reactive current Iqrmax 20 [pu/s]
REGC A 0 Minimum rate-of-change of reactive current Iqrmin −20 [pu/s]
REGC A 1 Terminal voltage filter time constant Tfltr 0.01 [s]
REGC A 1 Inverter active current regulator lag time constant Tgp 0.02 [s]
REGC A 1 Enable (1) or disable (0) low voltage power logic Lvplsw [0 1]∗ [–]
Generator/converter model
REGC A 1 Maximum rate-of-change of active current rrpwr [0.1 0.2 1 10]∗ [pu/s]
REGC A 1 Low voltage power logic zero crossing zerox 0.7 [pu]
REGC A 1 Low voltage power logic breakpoint brkpt 0.9 [pu]
REGC A 1 Low voltage power logic gain breakpoint Lvpl1 1 [pu/pu]
REGC A 1 Low voltage active current management breakpoint 0 lvpnt0 0.01 [pu]
REGC A 1 Low voltage active current management breakpoint 1 lvpnt1 0.05 [pu]
∗ The value of this parameter varies according to the selected control mode.
Appendix B. Modelling and control of photovoltaic systems
B.2.2
Table B.2: Parameters of electrical control model ————————-
REEC B 2 Bypass (0) or engage (1) inner voltage regulator loop QFlag [0 1]∗ [–]
REEC B 2 Local voltage regulator proportional gain Kvp 12 [pu/pu]
REEC B 2 Local voltage regulator integral gain Kvi 300 [pu/pu-s]
REEC B 2 Reactive current regulator lag time constant Tiq 0.01 [s]
REEC B 2 Local reactive power regulator proportional gain Kqp 0.05 [pu/pu]
REEC B 2 Local reactive power regulator integral gain Kqi 0.1 [pu/pu-s]
REEC B 2 Local reactive power (0) or voltage (1) control VFlag [0 1]∗ [–]
REEC B 2 Maximum voltage at inverter terminal bus Vmax 1.1 [pu]
REEC B 2 Minimum voltage at inverter terminal bus Vmin 0.9 [pu]
REEC B 2 Maximum reactive power if VFlag = 1 Qmax 0.4 [pu]
REEC B 2 Minimum reactive power if VFlag = 1 Qmin −0.4 [pu]
∗ The value of this parameter varies according to the selected control mode.
125
126
REPC A 1 Plant-level reactive power (0) or voltage (1) control RefFlag [0 1]∗ [–]
Reactive power deadband if RefFlag = 0 or
REPC A 1 dbd 0.005 [pu]
voltage deadband if RefFlag = 1
REPC A 1 Volt/var regulator proportional gain Kp 0.36 [pu/pu]
REPC A 1 Volt/var regulator integral gain Ki 0.18 [pu/pu-s]
REPC A 1 Voltage for freezing volt/var regulator integrator Vfrz 0.9 [pu]
REPC A 1 Plant controller reactive power output lead time constant Tft 0 [s]
REPC A 1 Plant controller reactive power output lag time constant Tfv 0.1 [s]
REPC A 1 Maximum volt/var error emax 0.1 [pu]
REPC A 1 Minimum volt/var error emin −0.1 [pu]
REPC A 1 Maximum plant reactive power Qmax plant 0.4 [pu]
REPC A 1 Minimum plant reactive power Qmin plant −0.4 [pu]
∗ The value of this parameter varies according to the selected control mode.
127
128
B.4 Validation
B.4.1 Low voltage ride-through and dynamic voltage
support
0.0 A2 (REMV)
A2 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
A2 (REMV)
−1.0 A2 (PowerFactory)
0.0 A3 (REMV)
A3 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
Figure B.7: Validation of low voltage ride-through and dynamic voltage support
Active current of photovoltaic system
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
A3 (REMV)
−1.0 A3 (PowerFactory)
Figure B.8: Validation of low voltage ride-through and dynamic voltage support
Reactive current of photovoltaic system
132 Appendix B. Modelling and control of photovoltaic systems
0.0 B1 (REMV)
B1 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
B1 (REMV)
−1.0 B1 (PowerFactory)
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0 B2 (REMV)
B2 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
B2 (REMV)
−1.0 B2 (PowerFactory)
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0 B3 (REMV)
B3 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
B3 (REMV)
−1.0 B3 (PowerFactory)
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
0.0
−0.2
C1 (REMV)
Active current [pu]
C1 (PowerFactory–Qgen)
−0.4 C1 (PowerFactory–Vt filt)
−0.6
−0.8
−1.0
1.0
0.5 C1 (REMV)
Reactive current [pu]
C1 (PowerFactory–Qgen)
C1 (PowerFactory–Vt filt)
0.0
−0.5
−1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time [s]
0.0 C2 (REMV)
C2 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
1.0
C2 (REMV)
C2 (PowerFactory)
0.5
Reactive current [pu]
0.0
−0.5
−1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time [s]
0.0 D1 (REMV)
D1 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
D1 (REMV)
−1.0 D1 (PowerFactory)
0.0 D2 (REMV)
D2 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
D2 (REMV)
−1.0 D2 (PowerFactory)
0.0 D3 (REMV)
D3 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
D3 (REMV)
−1.0 D3 (PowerFactory)
0.0 D4 (REMV)
D4 (PowerFactory)
−0.2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
D4 (REMV)
−1.0 D4 (PowerFactory)
·10−2
B1
8 ·10−2 B2
1.8 A3
B3
6 1.6
1.4
Slip [pu]
4 1.2
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
Figure C.1: Impact of active current recovery rate
Slip of small induction motor load connected to bus 7
·10−2
B1
8 ·10−2 B2
1.5 A3
B3
6
1
Slip [pu]
4 0.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
Figure C.2: Impact of active current recovery rate
Slip of large induction motor load connected to bus 7
C.2. Case studies 145
·10−2
A3
8 ·10−2 C1
1.6 C2
1.4
6
1.2
Slip [pu]
1
4 0.8
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
·10−2
A3
8 ·10−2 C1
C2
1
6
0.5
Slip [pu]
4
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
·10−2
D1
8 ·10−2 D2
1.6 D3
1.4 D4
6
1.2
Slip [pu]
1
4 0.8
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
·10−2
D1
8 ·10−2 D2
D3
1 D4
6
0.5
Slip [pu]
4
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
Bus Ssim [MVA] Rsims [pu] Xsims [pu] Xsimm [pu] Rsimr [pu] Xsimr [pu] Hsim [s] PF sim [–]
D.1. Test system
Bus Slim [MVA] Rlims [pu] Xlims [pu] Xlimm [pu] Rlimr [pu] Xlimr [pu] Hlim [s] PF lim [–]
1 113 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
2 62 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
3 49 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
4 158 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
5 135 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
11 38 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
12 56 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
13 19 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
22 53 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
31 19 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
32 38 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
41 101 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
42 75 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
43 169 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
46 131 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
47 19 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
51 150 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
61 94 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
62 56 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
63 111 0.013 0.067 3.8 0.009 0.17 1.5 0.8ind
Appendix D. Stability analysis of a multi-machine system
D.1. Test system 151
Bus Streq [MVA] Vtreqh [kV] Vtreql [kV] Rtreq [pu] Xtreq [pu] ntreq [pu/pu]
1041 130 20 0.6 0.0088 0.0593 1
1042 130 20 0.6 0.0088 0.0593 1
1043 130 20 0.6 0.0088 0.0593 1
1044 130 20 0.6 0.0088 0.0593 1
1045 130 20 0.6 0.0088 0.0593 1
4042 240 20 0.6 0.0088 0.0593 1
4043 240 20 0.6 0.0088 0.0593 1
4046 240 20 0.6 0.0088 0.0593 1
4061 300 20 0.6 0.0088 0.0593 1
4062 300 20 0.6 0.0088 0.0593 1
Bus Ssubtr [MVA] Vsubtrh [kV] Vsubtrl [kV] Rsubtr [pu] Xsubtr [pu] nsubtr [pu/pu]
1041 130 130 20 0.003 0.15 1.05
1042 130 130 20 0.003 0.15 1.05
1043 130 130 20 0.003 0.15 1.05
1044 130 130 20 0.003 0.15 1.05
1045 130 130 20 0.003 0.15 1.05
4042 250 400 20 0 0.15 1.05
4043 250 400 20 0 0.15 1.05
4046 250 400 20 0 0.15 1.05
4061 300 400 20 0 0.15 1.05
4062 300 400 20 0 0.15 1.05
Bus Pexp [MW] Qexp [Mvar] Psim [MW] Qsim [Mvar] Plim [MW] Qlim [Mvar]
1 420 103.7 90 22.2 90 22.2
2 231 49.7 49 10.6 49 10.6
3 182 58.6 39 12.5 39 12.5
4 588 176.4 126 37.8 126 37.8
5 504 133.2 108 28.5 108 28.5
11 140 48.1 30 10.3 30 10.3
12 210 58.6 45 12.5 45 12.5
13 70 24.0 15 5.1 15 5.1
22 196 55.9 42 11.9 42 11.9
31 70 17.2 15 3.7 15 3.7
32 140 27.7 30 5.9 30 5.9
41 378 91.9 81 19.7 81 19.7
42 280 89.1 60 19.1 60 19.1
43 630 178.2 135 38.1 135 38.1
46 490 148.2 105 31.7 105 31.7
47 70 30.8 15 6.6 15 6.6
51 560 180.7 120 38.7 120 38.7
61 350 85.7 75 18.3 75 18.3
62 210 58.6 45 12.5 45 12.5
63 413 185.2 88 39.6 88 39.6
71 300 83.8 – – – –
72 2000 396.1 – – – –
1.2
1.0
0.8
Voltage [pu]
0.6
0.4
LVRT characteristic
0.2
A1
A2
0.0 A3
1.2
1.0
0.8
Voltage [pu]
0.6
0.4
WECC criterion
0.2
A1
A2
0.0 A3
0.0
A1
−0.2 A2
A3
Active current [pu]
−0.4
−0.6
−0.8
−1.0
Figure D.3: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Active current of all photovoltaic systems
0.0
−0.2
Reactive current [pu]
−0.4
−0.6
−0.8
A1
A2
−1.0 A3
Figure D.4: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Reactive current of all photovoltaic systems
D.2. Case studies 157
1.0 A1
A2
A3
0.8
0.6
Slip [pu]
0.4
0.2
0.0
Figure D.5: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Slip of all small induction motor loads
1.0 A1
A2
A3
0.8
0.6
Slip [pu]
0.4
0.2
0.0
Figure D.6: Impact of low voltage ride-through and dynamic voltage support (fault 2)
Slip of all large induction motor loads
158 Appendix D. Stability analysis of a multi-machine system
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6 1.15
1.10
1.05
0.4 1.00
0.95
0.90 B1
0.2
0.85 B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
1.0
0.8
WECC criterion
Voltage [pu]
0.6 1.10
1.05
1.00
0.4
0.95
0.90
0.2 B1
0.85
B2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 A3
0.0 B3
0.0 B1
B2
A3
−0.2 B3
Active current [pu]
−0.4
−0.6
−0.8
−1.0
0.0
−0.2
Reactive current [pu]
0.00
−0.4 −0.05
−0.10
−0.15
−0.6 −0.20
−0.25
−0.30
−0.8 −0.35 B1
0.0 0.5 1.0 1.5 2.0 2.5 3.0 B2
A3
−1.0 B3
1.2
1.0
0.8
Voltage [pu]
0.6
0.4
LVRT characteristic
0.2
A3
C1
0.0 C2
1.2
1.0
0.8
Voltage [pu]
0.6
0.4
0.0
A3
−0.2 C1
C2
Active current [pu]
−0.4
−0.6
−0.8
−1.0
1.0 A3
C1
C2
0.5
Reactive current [pu]
0.0
−0.5
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Time [s]
1.0
0.8
0.6
Slip [pu]
A3
0.4 C1
C2
0.2
0.0
1.0
0.8
0.6
Slip [pu]
A3
0.4 C1
C2
0.2
0.0
1.0
0.8
LVRT characteristic
Voltage [pu]
0.6 1.15
1.10
0.4 1.05
1.00
D1
0.2 0.95
D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
1.0
0.8
WECC criterion
Voltage [pu]
0.6 1.10
1.05
0.4 1.00
0.95
0.2 0.90 D1
D2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 D3
0.0 D4
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
0.0 D1
D2
D3
−0.2 D4
−0.80
Active current [pu]
−0.85
−0.4
−0.90
−0.95
−0.6
−1.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
1.0 D1
0.8 D2
D3
0.6
D4
Reactive current [pu]
0.4
0.2
0.0
−0.2
−0.4
−0.6
−0.8
−1.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Time [s]
Nomenclature
Abbreviations
CCT Critical Clearing Time
DAE Differential Algebraic Equation
DVS Dynamic Voltage Support
EMT Electro-Magnetic Transient
EPRI Electric Power Research Institute
FACTS Flexible Alternating Current Transmission System
FIDVR Fault-Induced Delayed Voltage Recovery
FRT Fault Ride-Through
HVDC High Voltage Direct Current
IM Induction Motor
KL Kullback-Leibler
LVRT Low Voltage Ride-Through
PCC Point of Common Coupling
PDF Probability Density Function
PI Proportional Integral
PLL Phase-Locked Loop
PV PhotoVoltaic
REEC Renewable Energy Electrical Control
REGC Renewable Energy Generator/Converter
REMV Renewable Energy Model Validation
REPC Renewable Energy Plant Control
RMS Root Mean Square
SC Synchronous Condenser
SG Synchronous Generator
VRI Voltage Recovery Index
WECC Western Electricity Coordinating Council
166 NOMENCLATURE
λ−
1 Penalising weighting coefficient for voltage limit 1 of Western [–]
Electricity Coordinating Council violation criterion
λ−
2 Penalising weighting coefficient for voltage limit 2 of Western [–]
Electricity Coordinating Council violation criterion
λ−
3 Penalising weighting coefficient for voltage limit 3 of Western [–]
Electricity Coordinating Council violation criterion
λ−j Penalising weighting coefficient for j-th voltage limit [–]
σ Threshold for accuracy of weighting function [–]
ϕbus Bus angle [◦ ]
ωimr Induction motor rotor speed [pu]
ωims Induction motor stator speed [pu]
K Kullback-Leibler divergence measure [–]
K2 Kullback-Leibler divergence measure of voltage at bus 2 [–]
K4 Kullback-Leibler divergence measure of voltage at bus 4 [–]
K7 Kullback-Leibler divergence measure of voltage at bus 7 [–]
Kbad Kullback-Leibler divergence measure of bad voltage recovery [–]
Kgood Kullback-Leibler divergence measure of good voltage recovery [–]
Kideal Kullback-Leibler divergence measure of ideal voltage recovery [–]
Kosc Kullback-Leibler divergence measure of oscillatory voltage [–]
recovery
KWECC Kullback-Leibler divergence measure of Western Electricity [–]
Coordinating Council voltage violation criterion
Number of time-domain simulation [–]
P KL Probability density function for Kullback-Leibler divergence [–]
measure
Pbad
KL
Probability density function for Kullback-Leibler divergence [–]
measure of bad voltage recovery
Pgood
KL
Probability density function for Kullback-Leibler divergence [–]
measure of good voltage recovery
PiKL Number of voltage samples in i-th sub-interval for Kullback- [–]
Leibler divergence measure
Pideal
KL
Probability density function for Kullback-Leibler divergence [–]
measure of ideal voltage recovery
Posc
KL
Probability density function for Kullback-Leibler divergence [–]
measure of oscillatory voltage recovery
Pref
KL
Probability density function for Kullback-Leibler divergence [–]
measure of reference voltage recovery
Pref,i
KL
Number of voltage samples in i-th sub-interval of reference voltage [–]
recovery for Kullback-Leibler divergence measure
PWECC
KL
Probability density function for Kullback-Leibler divergence [–]
measure of Western Electricity Coordinating Council voltage
violation criterion
P VRI Probability density function for voltage recovery index [–]
168 NOMENCLATURE
Pbad
VRI
Probability density function for voltage recovery index of bad [–]
voltage recovery
Pgood
VRI
Probability density function for voltage recovery index of good [–]
voltage recovery
PiVRI Number of voltage samples in i-th sub-interval for voltage [–]
recovery index
P VRI+ Rewarding probability density function for voltage recovery [–]
index
P1i
VRI+
Number of rewarding voltage samples for voltage limit 1 in [–]
i-th sub-interval for voltage recovery index
P2i
VRI+
Number of rewarding voltage samples for voltage limit 2 in [–]
i-th sub-interval for voltage recovery index
P3i
VRI+
Number of rewarding voltage samples for voltage limit 3 in [–]
i-th sub-interval for voltage recovery index
PiVRI+ Number of rewarding voltage samples in i-th sub-interval for [–]
voltage recovery index
PjiVRI+ Number of rewarding voltage samples for j-th voltage limit in [–]
i-th sub-interval for voltage recovery index
P VRI− Penalising probability density function for voltage recovery [–]
index
P1i
VRI−
Number of penalising voltage samples for voltage limit 1 in [–]
i-th sub-interval for voltage recovery index
P2i
VRI−
Number of penalising voltage samples for voltage limit 2 in [–]
i-th sub-interval for voltage recovery index
P3i
VRI−
Number of penalising voltage samples for voltage limit 3 in [–]
i-th sub-interval for voltage recovery index
PiVRI− Number of penalising voltage samples in i-th sub-interval for [–]
voltage recovery index
PjiVRI− Number of penalising voltage samples for j-th voltage limit in [–]
i-th sub-interval for voltage recovery index
V Voltage recovery index [–]
V2 Voltage recovery index of voltage at bus 2 [–]
V4 Voltage recovery index of voltage at bus 4 [–]
V7 Voltage recovery index of voltage at bus 7 [–]
Vbad Voltage recovery index of bad voltage recovery [–]
Vgood Voltage recovery index of good voltage recovery [–]
A Total number of voltage samples [–]
brkpt Low voltage power logic breakpoint [pu]
Bcoleq Collector system equivalent susceptance [pu]
Bitl Interconnection transmission line susceptance [pu]
Btl1 Transmission line 1 susceptance [pu]
Btl2 Transmission line 2 susceptance [pu]
NOMENCLATURE 169
Bibliography
[1] SMA Solar Technology AG, SMA Supplies System Technology for One of
Australia’s Largest PV Power Plants, Press Release, 2018.
[2] T. Ackermann, G. Andersson, and L. Söder, “Distributed Generation: A
Definition”, Electric Power Systems Research, vol. 57, no. 3, pp. 195 –204,
2001.
[3] Ecofys Germany GmbH, “Klimaschutz durch Reduzierung des Energiebe-
darfs für Gebäudekühlung”, Tech. Rep., 2011.
[4] Bundesverband Wärmepumpe (bwp) e. V., Marktzahlen 2017: Wärmepumpen-
Absatz wächst deutlich, Press Release, 2018.
[5] IEEE Standards Association, Standard for Interconnecting Distributed Re-
sources with Electric Power Systems, IEEE Std. 1547, 2014.
[6] European Network of Transmission System Operators for Electricity (ENTSO-
E), Network Code on Requirements for Grid Connection Applicable to All
Generators (NC-RfG), EU Regulation 2016/631, 2016.
[7] Verband der Elektrotechnik, Elektronik und Informationstechnik (VDE),
Guideline VDE-AR-N 4120, “Technical Requirements for the Connection
and Operation of Customer Installations to the High Voltage Network”, Tech.
Rep., 2015.
[8] E. Quitmann and E. Erdmann, “Power System Needs – How Grid Codes
Should Look Ahead”, IET Renewable Power Generation, vol. 9, no. 1, pp. 3–
9, 2015.
[9] J. C. Boemer, “On Stability of Sustainable Power Systems: Network Fault
Response of Transmission Systems with Very High Penetration of Distributed
Generation”, PhD thesis, Delft University of Technology, 2016.
176 BIBLIOGRAPHY
[46] J. Hossain and H. R. Pota, Robust Control for Grid Voltage Stability: High
Penetration of Renewable Energy. Springer, 2014.
[47] I. Green, “CAISO Experience with Impact of High Penetration of Renew-
able Resources on Short-Term Voltage Stability”, in IEEE Power & Energy
Society General Meeting, Denver, 2015.
[48] J. N. Notholt Vergara, “Fault Ride-Through Capabilities of Inverter-Based
Distributed Generation Connected to Low and Medium Voltage Distribution
Networks”, PhD thesis, University of Kassel, 2009.
[49] P. Kundur, J. Paserba, V. Ajjarapu, G. Andersson, A. Bose, C. Cañizares, N.
Hatziargyriou, D. Hill, A. Stankovic, C. Taylor, T. Van Cutsem, and V. Vit-
tal, “Definition and Classification of Power System Stability IEEE/CIGRE
Joint Task Force on Stability Terms and Definitions”, IEEE Transactions on
Power Systems, vol. 19, no. 3, pp. 1387–1401, 2004.
[50] IEEE Task Force on Stability Definitions and Characterization of Dynamic
Behavior in Systems with High Penetration of Power Electronic Interfaced
Technologies, Ongoing Activities.
[51] C. W. Taylor, Power System Voltage Stability. McGraw Hill, 1994.
[52] T. Van Cutsem and C. Vournas, Voltage Stability of Electric Power Systems.
Springer, 2008.
[53] K. Kawabe and K. Tanaka, “Analytical Method for Short-Term Voltage Sta-
bility Using the Stability Boundary in the P–V Plane”, IEEE Transactions
on Power Systems, vol. 29, no. 6, pp. 3041–3047, 2014.
[54] S. Dasgupta, M. Paramasivam, U. Vaidya, and V. Ajjarapu, “Real-Time
Monitoring of Short-Term Voltage Stability Using PMU Data”, IEEE Trans-
actions on Power Systems, vol. 28, no. 4, pp. 3702–3711, 2013.
[55] A. A. Fouad and V. Vittal, Power System Transient Stability Analysis Using
the Transient Energy Function Method. Prentice Hall, 1992.
[56] M. Pavella, D. Ernst, and D. Ruiz-Vega, Transient Stability of Power Systems
– A Unified Approach to Assessment and Control. Springer, 2000.
[57] North American Electric Reliability Corporation (NERC), “Assessment of
Fault-Induced Delayed Voltage Recovery (FIDVR) Simulations”, Tech. Rep.,
2017.
[58] S. M. Halpin, K. A. Harley, R. A. Jones, and L. Y. Taylor, “Slope-Permissive
Under-Voltage Load Shed Relay for Delayed Voltage Recovery Mitigation”,
IEEE Transactions on Power Systems, vol. 23, no. 3, pp. 1211–1216, 2008.
[59] H. Bai and V. Ajjarapu, “A Novel Online Load Shedding Strategy for Mit-
igating Fault-Induced Delayed Voltage Recovery”, IEEE Transactions on
Power Systems, vol. 26, no. 1, pp. 294–304, 2011.
180 BIBLIOGRAPHY
[75] IEEE Task Force on Terms and Definitions, “Proposed Terms and Definitions
for Power System Stability”, IEEE Transactions on Power Apparatus and
Systems, vol. PAS-101, no. 7, pp. 1894–1898, 1982.
[76] Z. Wang, X. Wang, and C. Y. Chung, “An Analytical Method for Calculating
Critical Voltage Sag Clearance Time of Induction Motors”, IEEE Transac-
tions on Power Delivery, vol. 27, no. 4, pp. 2412–2414, 2012.
[77] S. Kullback and R. A. Leibler, “On Information and Sufficiency”, Annals of
Mathematical Statistics, vol. 22, no. 1, pp. 79–86, 1951.
[78] L. K. Nash, Elements of Classical and Statistical Thermodynamics. Addison-
Wesley, 1968.
[79] T. M. Cover and J. A. Thomas, Elements of Information Theory. Springer,
1991.
[80] D. J. Shoup, J. J. Paserba, and C. W. Taylor, “A Survey of Current Practices
for Transient Voltage Dip/Sag Criteria Related to Power System Stability”,
in IEEE PES Power Systems Conference and Exposition, New York, 2004.
[81] M. Paramasivam, A. Salloum, V. Ajjarapu, V. Vittal, N. B. Bhatt, and S.
Liu, “Dynamic Optimization Based Reactive Power Planning to Mitigate
Slow Voltage Recovery and Short-Term Voltage Instability”, IEEE Transac-
tions on Power Systems, vol. 28, no. 4, pp. 3865–3873, 2013.
[82] Western Electricity Coordinating Council (WECC) Renewable Energy Mod-
eling Task Force, “WECC Solar PV Dynamic Model Specification”, Tech.
Rep., 2012.
[83] Western Electricity Coordinating Council (WECC) Renewable Energy Mod-
eling Task Force, “WECC Solar Plant Dynamic Modeling Guidelines”, Tech.
Rep., 2014.
[84] Western Electricity Coordinating Council (WECC) Renewable Energy Mod-
eling Task Force, “WECC Second Generation Wind Turbine Models”, Tech.
Rep., 2014.
[85] National Renewable Energy Laboratory (NREL), “User Guide for PV Dy-
namic Model Simulation Written on PSCAD Platform”, Tech. Rep., 2014.
[86] Western Electricity Coordinating Council (WECC) Renewable Energy Mod-
eling Task Force, “WECC Central Station Photovoltaic Power Plant Model
Validation Guideline”, Tech. Rep., 2015.
[87] Vereinigung Deutscher Elektrizitätswerke (VDEW), “Eigenerzeugungsanla-
gen am Mittelspannungsnetz – Richtlinie für Anschluss und Parallelbetrieb
von Eigenerzeugungsanlagen am Mittelspannungsnetz”, Tech. Rep., 1998.
[88] Vereinigung Deutscher Elektrizitätswerke (VDEW), “Erzeugungsanlagen am
Niederspannungsnetz – Richtlinie für Anschluss und Parallelbetrieb von Eigen-
erzeugungsanlagen am Niederspannungsnetz”, Tech. Rep., 2001.
182 BIBLIOGRAPHY
Vol 9
ISBN 978-3-7376-0716-2
9 783737 607162
Gustav Lammert
were determined in collaboration with a manufacturer. The stability
analysis is performed in DIgSILENT PowerFactory using: i) a one-load
infinite-bus system; and ii) an IEEE multi-machine voltage stability
test system, namely the Nordic test system. The results show that
with the adequate control of photovoltaic systems, power system
dynamic performance can be significantly improved.