Operational strategies for HVDC transmission

in smart grids: the security versus markets

dilemma

Master Thesis

Chanpreet Kaur Talwar

Technische Universiteit Delft
O PERATIONAL STRATEGIES FOR HVDC TRANSMISSION IN SMART
GRIDS : THE SECURITY VERSUS MARKETS DILEMMA

M ASTER T HESIS

by

Chanpreet Kaur Talwar

in partial fulfillment of the requirements for the degree of

Master of Science

in Electrical Engineering and Computer Science

(Intelligent Electrical Power Grids)

at the Delft University of Technology,

to be defended publicly on Monday August 28, 2017 at 10:00 AM.

Supervisors: Prof. dr. Peter Palensky, TU Delft

Dr. ir. Georgios Papaefthymiou, Elia Grid International, Germany
Ir. Martijn de Jong, TU Delft

Thesis committee: Prof. dr. Peter Palensky, TU Delft

Dr. ir. Jose Luis Rueda Torres, TU Delft
Dr. Domenico Lahaye, TU Delft
Ir. Martijn de Jong, TU Delft

An electronic version of this thesis is available at http://repository.tudelft.nl/.

Preface

First of all, I wish to thank my responsible supervisor, prof. Peter Palensky for guiding me in pursuing
my thesis under his kind patronage, and allowing me to be a part of the Intelligent Electrical Power Grid
(IEPG) research group in the Netherlands. Second and foremost, I am highly thankful to my daily su-
pervisor Martijn De Jong for his monetary and moral support during the course of thesis studies. Words
cannot express my sincere appreciation, but all I can say is that I shall always remain highly obliged
and grateful to you for supervising my work, and finding time for me from your busy schedule to clarify
all my queries and doubts in the best possible way. I want to thank my supervisor in Germany, George
Papaefthimiou for guiding me in pursuing my thesis to this level, and extending valuable suggestions
and clarification as and when required.

During the process of study at TU Delft since 2015, I have been receiving constant feedback on my
work from Jose Luis Rueda Torres. I wish to thank you for your kind co-operation in helping and
guiding me at every stage of the programme with patience and motivation. I would like to extend my
gratitude to Domenico Lahaye for being a part of my thesis committee, and for providing me valuable
comments on my work .

I wish to thank my dearest parents, Mandeep Singh and Kawaljit Kaur for encouraging me to pur-
sue my quest for further advanced studies out of my mother country. It is because of their support that
I could see my dream come true. I would like to thank my best friend, Avik Basu for being with me
always. I want to extend my heartiest thanks to my most special friend Bheeshma Chatrath for being
my constant pillar of support all through this phase. With his constant encouragement and positivity I
have been able to do my best in the thesis. Lastly, I thank my Babaji (my Almighty) for all his beautiful
blessings that all my sufferings and disappointments got converted into abundance of happiness and
achievements.

Chanpreet Kaur Talwar

Delft, August 2017

iii
Contents

Abstract vii

List of acronyms ix

List of Figures xi

List of Tables xiii

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem analysis and definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Study goals and structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Smart grid operation and power flow modeling - a review 7

2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Power system security assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Deterministic security assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 Probabilistic security assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Review on smart components and smart technologies . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Flexible AC transmission system devices . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 HVDC technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.3 Dynamic line rating (DLR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.4 Demand side management (DSM) . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Power system modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Power flow equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.5.1 DC power flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.5.2 AC power flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Optimal power flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6.1 Need for optimal power flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6.2 Extensions of OPF. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3 Risk-based power system security assessment 17

3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Modeling of stochastic dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1 Spatial correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.2 Rank correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

v
 3.2.3 Copula Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 An overview of state-of-the-art-research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 General overview: Proposed RBSA methodology . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.1 Monte-Carlo framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5 Two risk identifying tools. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 High-Voltage Direct Current (HVDC) technology 27

4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 HVDC grid operation and power flow controllability . . . . . . . . . . . . . . . . . . . . . 27
4.2.1 HVDC Topology: LCC vs. VSC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 HVDC transmission line modeling concepts . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3.1 Steady state VSC-HVDC transmission line model . . . . . . . . . . . . . . . . . . . 30
4.3.2 VSC-HVDC transmission line model in MATPOWER . . . . . . . . . . . . . . . . . 30

5 Experiments and results 33

5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.1 Implementation and hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.2 Test system and its adjustments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.3 Result visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3 Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3.1 Case study 1: Sensitivity analysis on HVDC set-points . . . . . . . . . . . . . . . . 35
5.3.2 Case study 2: Adaptability of HVDC set-point by TSO as curative remedial action . 44
5.3.3 Case study 3: 2D study with two HVDC transmission lines . . . . . . . . . . . . . . 46

6 Conclusions and future research 50

6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.3 Future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

A System parameters 52
A.1 Bus parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
A.2 Generator parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.3 Transmission line parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
A.4 Generation cost parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
A.5 HVDC line specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

B Visualization tool 57
B.1 Detailed analysis when HVDC set-point is set to 325 MW. . . . . . . . . . . . . . . . . . . 57
B.2 Detailed analysis HVDC set-point is set to 500 MW . . . . . . . . . . . . . . . . . . . . . . 58

Bibliography 59

vi
Abstract

Over the last decades, the European transmission system has made many profound changes in the net-
work and has focused on three main concepts: i) flexibility, ii) integration, and iii) sustainability to
increase the technological innovations, and to improve the market design. The currently used method
by transmission system operators (TSOs) is trying to accomplish these requirements, but it is important
to realize that each TSO has its own grid protocols and standards. Consequently, all major TSOs in
the interconnected meshed European transmission system are facing a huge difficulty in maintaining
a strong operational coordination to work together as a one single European technical market model.
In order to guarantee the highest security of electricity supply, it is necessary to structure a stable, reli-
able and secure analytical AC framework that takes into consideration the stochastic nature of system
in-feeds in the daily operational planning. In this thesis it is analyzed how incorporation of smart tech-
nologies such as HVDC transmission can be used as a smart grid solution to improve the power system
security and lower the risk in different adjacent areas/zones. The proposed risk-based security assess-
ment (RBSA) methodology based on Monte-Carlo sampling is employed to investigate the security of
the system and to quantify the expected system risk. It is shown that the market optimal HVDC power
set-points may result in unnecessarily high risk when subjected to the unavoidable uncertainty of inputs
(fluctuations in load and RES) inherent to day-ahead forecasting. A detailed comparison of market
optimal versus security optimal HVDC power set-point is presented. It is proposed to properly adapt
the HVDC set-points with respect to the actual operating situation, which can be quite different from
the day-ahead point forecast. Moreover, it is shown that by being able to adapt HVDC set-points in real-
time operation, further more serious and more costly remedial actions such as active re-dispatch and
load shedding, can be avoided. Furthermore, a study with two HVDC transmission lines is performed
to show the necessity of coordinated control of the HVDC lines, and how this can reduce the stress in
the network by acting as a tool to shift generation.

Keywords: HVDC, set-points, remedial actions, smart grids, RBSA, Monte-Carlo

vii
List of acronyms

AC Alternating Current
AGC Automatic Generation Control
CDF Cumulative Distribution Function
CWE Central Western Europe
DC Direct Current
DLF Deterministic Load Flow
DLR Dynamic Line Rating
DSM Demand Side management
DSA Deterministic Security Assessment
ENTSO-E European Network of Transmission System Operators for Electricity
EOPF Enhanced Optimal Power Flow
EU European Union
FACTS Flexible Alternating Current Transmission System
FMBC Flow-based Market Coupling
GISO Geographically Separated Independent System Operator
GUI Graphical User Interface
HV High Voltage
HVAC High Voltage Alternating Current
HVDC High Voltage Direct Current
IGBT Insulated-Gate Bipolar Transistor
LCC Line Commutated Converter
LV Low Voltage
MC Monte-Carlo
MTDC Multi-terminal Direct Current
MV Medium Voltage
OHL Overhead Line
OL-RBSA Online Risk-based Security assessment
OPF Optimal Power Flow
PDF Probability Density Function
PLF Probabilistic Load Flow
PSA Probabilistic Security Assessment
PST Phase Shifting Transformer
PWM Pulse Width Modulation
RB-DCSA Risk-based DC Security Assessment

ix
RBSA Risk-based Security Assessment
RES Renewable Energy Sources
RSC Regional Security Coordinator
SOA System Operation Agreement
SCOPF Security Constraint Optimal Power Flow
SISO Super Independent System Operator
SSSC Static Synchronous Series Compensator
STATCOM Static Synchronous Compensator
TCSC Thyristor-Controlled Series Capacitor
TCSR Thyristor-Controlled Series Reactor
TDP Time-Dependent Phenomena
TISO Technology Separated Independent System Operator
TSO Transmission System Operator
VaS Value of Security
VRES Variable Renewable Energy Source
VSC Voltage Source Converter

x
List of Figures

1.1 Governing factors in the existing transmission grid infrastructure. . . . . . . . . . . . . . . 2

1.2 Thesis outline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1 Classification of decision drivers of power system security analysis [3]. . . . . . . . . . . . 8

2.2 Five services to TSOs by RSCs in operational planning for security purposes [4]. . . . . . . 8

3.1 Flowchart for McCalley’s power system security methodology. . . . . . . . . . . . . . . . . . 19

3.2 Flowchart for Kirschen’s probabilistic framework for one MC sample. . . . . . . . . . . . . 20
3.3 Layout of proposed RBSA methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 Outline of RBSA methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5 Flowchart of RBSA methodology for one MC sample. . . . . . . . . . . . . . . . . . . . . . . 23

4.1 Cost comparison of HVAC and HVDC transmission technology. . . . . . . . . . . . . . . . . . 28

4.2 LCC-HVDC converter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 VSC-HVDC converter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.4 MATPOWER model of VSC-HVDC transmission line. . . . . . . . . . . . . . . . . . . . . . . . 31

5.1 The two-zone IEEE 24-bus reliability test system with HVDC line between buses 23 and
12 (blue). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.2 Flowchart of proposed HVDC-RBSA methodology.1 . . . . . . . . . . . . . . . . . . . . . . . 36
5.3 Comparison of dispatch costs and risk (active re-dispatch and lost active load). . . . . . . 39
5.4 Re-dispatched active power for every generator. . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.5 Lost active load for every bus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.6 Number of overloaded circuits at every bus. Total number of MC-samples is 5000. . . . . . 40
5.7 Number of voltage violations at every bus. Total number of MC-samples is 5000. . . . . . 40
5.8 Improved voltage profiles at bus 3 with RBSA check. . . . . . . . . . . . . . . . . . . . . . . 41
5.9 Probabilities for causes of problems in the system. . . . . . . . . . . . . . . . . . . . . . . . 41
5.10 Probabilities of risk levels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.11 The best compromise between cost and risk is indicated by the blue dot. . . . . . . . . . . 42
5.12 Impact of forecast uncertainty: Risk in terms of re-dispatch active load and lost active load. 43
5.13 Impact of correlation: Risk in terms of re-dispatch active load and lost active load. . . . . 43
5.14 Average operating state when HVDC set-point is set to 325 MW. . . . . . . . . . . . . . . . 45
5.15 Average operating state when HVDC set-point is set to 500 MW. . . . . . . . . . . . . . . . 45
5.16 Comparison of “dumb” (blue) and “smart” approach (green). . . . . . . . . . . . . . . . . . 47
5.17 Adaptability of HVDC set-point 325 MW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.18 Risk in terms of re-dispatch active load and lost active load. . . . . . . . . . . . . . . . . . . 48

xi
5.19 Ratio of cost and risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

B.1 Network parameters when HVDC set-point is set to 325 MW. . . . . . . . . . . . . . . . . . 57

B.2 Network parameters when HVDC set-point is set to 500 MW. . . . . . . . . . . . . . . . . . 58

xii
List of Tables

2.1 Security related-decisions for TSOs based on different time-periods [3]. . . . . . . . . . . . 9

5.1 Comparison of “security optimal” and “market optimal” HVDC power set-points. . . . . . . 37
5.2 Zone-wise re-dispatched active generation and lost active load. . . . . . . . . . . . . . . . . 38

A.1 Bus parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

A.2 Generator parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.3 Transmission line parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
A.4 Generation cost parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
A.5 HVDC line specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

xiii
 To my parents,
Mandeep Singh and Kawaljit Kaur

xv
Chapter 1

Introduction

1.1 Background
The radical transformation of the electrical power industry, along with the growing interest towards
smart grid technologies such as power flow controllable components are having a strong impact on the
operational security of the European interconnected transmission system. Transition to non-conventional
energy sources, decommissioning of conventional power plants, difficulties in constructing new over-
head transmission lines (OHLs), incorporation of high voltage direct current (HVDC) transmission sys-
tems, implementation of new European network code, and regional market coupling are the several
governing factors involved in changing the dynamics of the existing transmission grid infrastructure as
illustrated in Figure 1.1 [1, 2, 5, 6]. These radical changes in the existing transmission grid infrastruc-
tures moderate the ability of the transmission system operators (TSOs) to assure the highest security of
the power system, while maintaining equitably low operating costs. Safeguarding and maintaining the
power system security in the existing European transmission system is necessary for successful integra-
tion of variable renewable energy sources (VRES) such as wind and solar generation.

The tremendous growth in the share of VRES in the interconnected European transmission system
over the last decade has been driven by significant policies set by the European Union (EU). These poli-
cies are focused on attaining the EU Energy policy objectives: i) sustainability, ii) security of electricity
supply, and iii) competitiveness [1, 2, 7, 8]. However, it must be noted that the successful implementa-
tion of the security measures is crucial to support the EU Energy Policy such as, the 20-20-20 targets.
The goal is to increase the share of VRES in the transmission system by 20%, mitigate the greenhouse
gas emission by 20%, and enhance the energy efficiency by 20% by the year 2020 [1]. Moreover, the
attention is deviating towards EU Energy Roadmap 2050, which focuses on reducing the emission of
greenhouse gases by 80% by the year 2050 [8]. In May 2015 the establishment of flow-based market
coupling (FMBC) has marked a major milestone in the European target electricity model [9] in Central
Western Europe (CWE), i.e., the Netherlands, Germany, France, Luxembourg, and Belgium [10]. These
emerging policies have a very real impact on the interconnected transmission system whole across Eu-
rope. As a result, the European Network of Transmission System Operators for Electricity (ENTSO-E),
are exploring solutions on how to overcome the challenges in the changing European interconnected
transmission system while maintaining security of power supply.

Large RES penetrating into the meshed network require a power grid upgradability, or an extension to

1
maintain large energy inflows. In order to overcome the transformation in the existing infrastructure,
the VSC-HVDC technology is supposed to be the most suitable solution for the power grid reinforcement
and upgradability. It uses pulse width modulation (PWM) control scheme, and provides fast controlla-
bility of active and reactive power flows [11]. Hence, the incorporation of VSC-HVDC connections in
the European transmission system can facilitate the bulk transmission of electrical power across cross-
border electricity market.

As a consequence of these new trends in the European capacity market, the individual TSOs need
to assure the security of electricity supply in their domain to facilitate a smooth functioning across
cross-border markets. Hence, it is essential to conduct a more comprehensive research on the estima-
tion of the system parameters by incorporating a probabilistic risk-based security assessment (RBSA)
framework in the operational day-ahead planning and system operation.

Contribution to
EU energy policies

Implementation
of new European
Concept of network codes
supergrid

Governing Factors
Higher
Difficulties in intermittent RES
building new system in-feeds
overhead
transmission lines

Development of
European integrated
market

Figure 1.1: Governing factors in the existing transmission grid infrastructure.

1.2 Problem analysis and definition

Many projects based on HVDC technology in the European transmission system are providing eye-
catching solutions for exporting bulk amount of electrical power over long transmission distances. The
projects based on multi-terminal direct current grids (MTDC) such as DESERTEC and Medgrid, are sug-
gested to integrate massive intermittent renewable energy sources (RES) across Europe, the Middle East
and North Africa [11, 12]. These projects are supposed to be the milestones for the European super grid
structure. The HVDC installations named Baltic cable, Kontek and SwePol link are established across
the Nordic and Central Eastern Europe to relieve the phenomenon of loop flows between Germany,
Sweden, Denmark and Poland [13]. The first onshore VSC-HVDC interconnection in Europe is installed
between Spain(Vic) and France(Baixas), with a rated transmission capacity of 2GW, and a DC voltage

2
level of ±320 kilovolts (kV). This onshore VSC-HVDC interconnection is placed in the existing meshed
AC synchronous system, and is parallel with the 400 kV Spain(Vic)-France(Baixas) transmission line.
In order to manage the flow of DC active power, various control strategies such as i) predefined active
power set-points which enables TSOs to fix the active power transfer in the VSC-HVDC connections,
and ii) procedures that would emulate the characteristics of the AC transmission line are incorporated
[14]. As a result, this transformation in the existing transmission grid infrastructure necessitates the
need to analyze how incorporation of HVDC transmission can be used as a smart grid solution to im-
prove power system security and lower the risk in different adjacent areas. Related ideas are discussed
under EC FP7 UMBRELLA and iTesla projects, where a risk-based toolbox based on stochastic methods
is developed to adequately capture the complex grid operation and help TSOs to access the system risk
in the interconnected European transmission system [15, 16].

1.3 Study goals and structure

The integration of stochastic generation and the liberalization of the European energy market poses
several operational challenges in a highly meshed European transmission system. In order to guarantee
the highest security of electricity supply, it is necessary to structure a stable, reliable and secure ana-
lytical AC framework that takes into consideration the stochastic nature of system in-feeds in the daily
operational planning. In this thesis it is analyzed how incorporation of an HVDC transmission line can
effect the overall system risk, or the risk in different areas/zones. It is shown that HVDC set-points com-
puted by the market may result in unnecessarily high risk. A RBSA methodology using a combination
of a Monte-Carlo sampling and the copula theory is employed to quantify the expected system risk in
the operational day-ahead planning, taking into account forecast uncertainty and dependence of system
in-feeds. It is shown that HVDC set-points should be adjusted in response to the actual operating situ-
ation, which can be quite different from the day-ahead point forecast as a result of uncertainty in load
and RES in-feeds. It is shown that TSOs cannot always use the cost optimal HVDC set-point computed
by the market as there may be cases where the system can run into serious problems. It is shown that
by being able to adapt HVDC set-points in real-time operation, further more serious and more costly
remedial actions such as active re-dispatch and load shedding, can be avoided. Additionally, a study
with two HVDC transmission lines is performed to check how HVDC lines can serve as a tool to shift
generation and reduce the stress in the network.

1.4 Research questions

The main research questions for this thesis, addressing how HVDC transmission can be used as smart grid
technology to improve the operational security, can be formulated as follows:

• Investigate how the proposed RBSA methodology can be used in determining the best HVDC
set-point when it is subjected to the unavoidable uncertainty of inputs (fluctuations in system
in-feeds) inherent to day-ahead forecasting?

• Investigate how the proper adaptability of the HVDC set-points can be ensured in improving
the operational security of the interconnected transmission system?

• Investigate how two HVDC transmission lines can work together to find the operating point
that provides the best compromise between risk and operating costs?

3
1.5 Outline of the thesis
The outline of the thesis is enlisted as follows:

Chapter 2 presents the literature review on power system security, deterministic and probabilistic ap-
proaches and smart technologies including HVDC technology, FACTS, DLR and DSM. It gives an account
on the power flow modeling, power flow equations, and optimal power flow and its extension.

Chapter 3 discusses the literature on stochastic load and generation. It presents the key aspects on
the proposed RBSA methodology used to quantify system risk. It gives an account on how it can be
used to assess the system risk in the day-ahead operational planning.

Chapter 4 presents the theory and modeling principles behind HVDC technology. It discusses the cur-
rent practice concerning day-ahead HVDC operation that are used by TSOs in determining the HVDC
set-points.

Chapter 5 present the experimental setup and the several case studies used in analyzing how HVDC
transmission line can be used as a smart grid solution to improve power system security. It discusses
the results obtained from the study in determining the best HVDC set-point that can achieve the highest
security level under all possible uncertainties. It is shown how multiple (two) HVDC lines should be
coordinated to provide the best operating point in terms of both security and costs.

Chapter 6 summarizes the main findings of this research and provides the recommendation for the
future work.

Figure 1.2 presents the schematic representation of the thesis outline.

4
 Motivation

Problem analysis and

Background Study goals and structure Research questions
definition

Literature review

Power system security Smart components and smart

Power system modeling Optimal power flow
assessment technologies

Proposed RBSA methodology

Overview of state-of-the-
art-research
HVDC grid operation and
power flow controllability

Modeling of stochastic
dependence

HVDC transmission line

modeling concepts
Monte-Carlo framework
HVDC operation

Two-risk identifying tools

Tool

Case studies

Sensitivity Analysis on HVDC set- Adaptability of HVDC set-point by TSO 2D study with two HVDC
points as curative remedial action transmission lines

Conclusions and future research

Figure 1.2: Thesis outline.

5
Chapter 2

Smart grid operation and power flow

modeling - a review

2.1 Overview
In this chapter, an overview on the smart grid operation and power flow modeling is presented. A description
of power system security assessment is provided in order to understand the need to investigate the expected
system risk inherent to day-ahead forecasting. The standard AC (and DC) power flow equations are given,
and the need to use optimal power flow to solve the power flow computations is discussed.

2.2 Power system security assessment

Nowaday’s power systems are predominantly non-linear and work continuously in varying operating
environments. As a result, the operating criteria for a system is to ensure resilience against the most
critical as well as non-critical contingencies. Maintaining the security of the system under varying op-
erating environments, and returning the system to an adequate state of security after a disturbance
are the main measures taken by TSOs during the power system security assessment. In particular, the
decision drivers differentiates between three levels of system security: i) overload security, ii) voltage
security, and iii) dynamic security, as illustrated in Figure 2.1 [3]. Many definitions of power system
security have been stated over many years. In [17], the security of a power system is stated as:

Security of a power system refers to the degree of risk in its ability to survive imminent disturbances (con-
tingencies) without interruption of customer service. It relates to robustness of the system to imminent
disturbances and, hence, depends on the system operating condition as well as the contingent probability of
disturbances.

Several operational challenges faced by TSOs are reported in [18], and it states that the high installation
of wind in Germany escalates the unscheduled loop flows through German-Czech and German-Polish
borders. These unscheduled power flows could exceed the thermal capacities, and can cause severe
problems to the entire system. However, in December 2015, a multilateral agreement has been signed
between TSOs and ENTSO-e which stated that each TSOs would obtain services by regional security co-
ordinators (RSCs) [4] to improve the power system security. These services are illustrated in Figure 2.2.

7
 Power system security

Overload security Voltage security Dynamic security

Transformer overload Low voltage Transient instablity

Transmission line overload Unstable voltage Oscillatory instablity

Figure 2.1: Classification of decision drivers of power system security analysis [3].

Hence, the transformation in the existing grid infrastructure necessitates the need to investigate the ex-
pected system risk in day-ahead operational planning.

Security Capacity Outage Adequacy Common grid

analysis calculation coordination forecast model

Figure 2.2: Five services to TSOs by RSCs in operational planning for security
purposes [4].

To guarantee the overall system security, a balance of economy and quality of power supply should be
maintained within certain permissible margins. The security related-decisions are based on different
time-periods, which are drawn by system analysts and TSOs to operate and control the power system.
Table 2.1 [3] summarizes the time frames on which the security related-decisions are examined. More-
over, the ENTSO-E provides a detailed information on the power system operating states in the Policy 5
of the Emergency Operations handbook [19] in order to assess the bottlenecks in the European trans-
mission system. The classification of the operational security in a power system is described as follows
[19]:
• Normal state: System state is N-Situation secure and it is within the permissible margin of the
allowed operational security. The system has low vulnerability to the critical events such as power
balance mismatch, catastrophic failures and power outages;

• Alert state: System state is within the permissible margin of the allowed operational security.
The system has high vulnerability to the critical events. Remedial curative actions such as re-

8
 Table 2.1: Security related-decisions for TSOs based on different time-periods [3].

Time-period Decision framer Security related-decision Criteria for security related-decision

How to confine the most economic
Operating criteria and cost,
Real-Time assessment Operator operation to real-time assessment,
real-time rules
maintain the secure operating state?

How the operating criteria Minimal operating criteria and cost,

Operational planning Analyst
should be defined? reliability

How to upgrade/support the

Reliability criteria for
Seasonal Planning Analyst reliability criteria for the
network plan and cost
network?

dispatching of active power, or re-scheduling of the transmission lines are needed to take the
system back to a secure state;

• Emergency state: System state has violated the technical operational security limits due to the
occurrence of severe disturbances. Emergency control actions are needed to bring the system to
the alert state;

• Blackout state: System state has violated all the technical operational security limits, and can
cause the whole network to collapse;

• Restorative state: System state reconnects all the damaged power equipment and re-establishes
the connection either to the normal or alert state, and as a result the permissible margin of the
allowed operational security is maintained.
The security assessment is associated with the investigation of all possible set of credible as well as non-
credible contingencies occurring in the system. The assessment of power system security in day-ahead
operation is categorized as follows: i) deterministic security assessment, and ii) probabilistic security
assessment [20].

2.2.1 Deterministic security assessment

Deterministic security assessment (DSA) analysis is based on a pre-defined list of the most probable
contingencies that may occur in a given system. On occurrence of any probable contingency, the system
should operate in such a way that the new operating state is within the post-contingency limits. Thus,
the deterministic approach depends on the following two assumptions [20]:
• Credibility: the topology of a network, outage of a circuit and operating condition are probable,
i.e., they are most likely to happen during the power system security assessment;

• Severity: the extent to which the threshold limits of a system are violated. There should not be
any probable contingency, wherein the combination of network topology, loss of any component
and operation conditions have a more severe effect on the system performance.
The methodology to evaluate for the DSA comprises of the following six steps [20]:
(i) The base cases to represent varied time horizons (years, seasons) and loading levels are selected.
The unit commitment is performed for each case based on the selected system topology and time
frame;

9
 (ii) The set of credible contingencies based on the (N-1) criterion are considered;

(iii) The study parameters to be investigated are selected, and their ranges of the operating conditions
are chosen. This range of operating conditions is called as study range;

(iv) The limiting contingencies, i.e., the events which are first to violate the performance assessment
criteria are examined. If there are no violations within the range of the operating conditions, the
evaluation is said to be done;

(v) The security boundary is defined by identifying the set of operating condition within the range
of operating conditions, i.e., the study range, where a limiting contingency is first to violate the
performance assessment criteria;

(vi) The security boundary is then visualized in a form of a sorted table, or plots such that it can be
well understood by the operators during the system operation.

2.2.2 Probabilistic security assessment

The DSA methodology is extensively used by TSOs to evaluate the security of the interconnected trans-
mission system. However, there are several weaknesses associated with this methodology, when used
to assess the expected system risk: i) only probable contingencies are considered, ii) operating cost
is relatively high, and iii) trivial information is obtained about the unacceptable region [21]. Thus, a
probabilistic framework is needed in order take into account the contingencies which are rare to occur
yet significant in terms of determining the operational security of the system. The probabilistic secu-
rity assessment (PSA) analysis is based on the set of probable as well as non-probable contingencies
to assess the operational risk in a given power system. It takes into account the following aspects: i)
uncertainty in load, ii) uncertainty in generation, and iii) uncertainty in contingency. This assessment
evaluates the security level by considering, the probability of the event and its severity in one index,
named the risk index [20, 22].

X
Ri sk(S) = P (E m ) ∗ Sev(E m , S), (2.1)
m

where,

• S is the consequence of the event (e.g., outage of a circuit, load mismatch);

• E m is the m-th event;

• P (E m ) is the probability of the m-th event;

• Sev( E m , S) is the severity associated with m-th event.

The methodology to evaluate for the PSA comprises of the similar six steps which are discussed in the
DSA, however steps 2, 4 and 5 are changed [20]:

(i) The base cases to represent varied time horizons (years, seasons) and loading levels are selected.
The unit commitment is performed for each case based on the selected system topology and time
frame;

(ii) The set of probable as well as non-probable contingencies are considered, which are based on the
procedure of event enumeration. This process is limited by taking into account only those events
whose predefined value has been exceeded;

10
(iii) The study parameters to be investigated are selected, and their ranges of the operating conditions
are chosen. This range of operating conditions is termed as study range;

(iv) The risk index is calculated all through the study range, and a threshold level is defined beyond
the unacceptable operating condition;

(v) The security boundary is defined by identifying the set of operating condition within the range of
operating conditions, i.e., the study range, where the index value is equivalent to the threshold
level;

(vi) The security boundary is then visualized in a form of a sorted table, or iso-risk plots such that it
can be well understood by the operators during the system operation.

2.3 Review on smart components and smart technologies

The inclusion of smart components such as flexible AC transmission system devices, and high capacity
DC interconnectors in the existing energy infrastructure, and the impact of smart technologies such as
demand side management and dynamic line rating add capabilities to the power system to handle a
broad range of operating points in order to ensure security and reliability. Below an overview is given
of the advantages of smart components and smart technologies in the existing grid.

2.3.1 Flexible AC transmission system devices

Flexible AC transmission systems (FACTS) are power electronics-based components, which are used
to control power flows or to provide reactive power compensation at weak network points in the AC
system [23]. These devices supply the desired amount of reactive power by regulating the bus voltages
of the system. Other types of devices such as phase shifting transformers (PSTs) can also be incorpo-
rated in the extended range of the FACTS family. These devices can provide active power flow control
on the transmission lines by varying the power angle in interconnected electrical systems. FACTS de-
vices are categorized into two groups: i) series compensating elements and, ii) parallel compensating
elements. On the one hand, series FACTS compensating elements such as thyristor-controlled series re-
actor (TCSR), thyristor-controlled series capacitor (TCSC), and static synchronous series compensator
(SSSC) can increase the active power flows and henceforth help in controlling the transient stability
problems. The TCSC offers a more effective solution for controlling the power oscillation damping of
the network, in comparison to parallel FACTS devices. On the other hand, parallel FACTS compensat-
ing devices such as static VAR compensator (SVC) and static synchronous compensator (STATCOM) are
mainly used to provide voltage support, and regulate the power factor of the network. Such devices
provide fast voltage regulation during steady-state and transient instability conditions. They provide
fault recovery and can aid in reducing the blackout risk [23].

2.3.2 HVDC technology

Below an overview is given of the advantages of integration of HVDC transmission lines in existing AC
grids [24]:

• HVDC connections enable the transmission of bulk power over long distances. For example, the
NorNed link is the long distance high-voltage undersea cable of length 580 km connected between
Norway and the Netherlands;

11
 • HVDC connections are economical because the line cost is reduced beyond the break-even point.
More power per conductor can be transmitted over long distances in a HVDC transmission system.
Since only two conductors are required for a bipolar HVDC line, the costs per conductor are lower
than for an equivalent AC line;

• HVDC can interconnect two AC asynchronous grids that operate with different nominal frequen-
cies. The HVDC connections can easily adjust to the rated voltage and rated frequency of the
system connected at the other end of the network;

• There is no skin effect, and comparatively less line and corona losses in an HVDC transmission
system;

• HVDC converters enable bidirectional power flow as they have the ability to operate either in
rectification mode or in inversion mode.

The details on HVDC technology are discussed in Chapter 4.

2.3.3 Dynamic line rating (DLR)

The ENTSO-E members are exploring for solutions on how to exploit the existing transmission system
infrastructure. The possible solution identified towards the control and planning of such infrastructures,
is the implementation of dynamic line rating (DLR) on real-time system operations [25]. The transmis-
sion line rating determines the highest current capacity that the transmission line can transfer without
violating the static constraints, or jeopardizing system security. However, there exist a crucial difference
on how the static and dynamic line rating alleviate congestions and improve the system flexibility. The
static line ratings are bounded by the static capacities and computes the static current based on standard
atmospheric conditions, while the dynamic line rating computes the dynamic current based on different
ambient atmospheric conditions, wind speeds, solar irradiation and precipitation. Therefore, incorpo-
rating DLR in the existing system may significantly impact the thermal capacity of the transmission lines
which can be used to increase transmission system efficiency of a highly meshed network.

2.3.4 Demand side management (DSM)

Demand side management (DSM) is based on a broad concept where consumers have the capability
to modify their usage from their nominal or actual energy consumption patterns by employing various
measures such as shifting the consumption to lower tariff periods [26]. Implementing DSM can provide
a greater controllability over the electrical power that has been used and produced. Moreover, smart
loads such as electric vehicles, energy storage, residential heat pumps, dynamic demand, and other
smart applications can further provide flexibility at the consumer side of the network. DSM provides
several advantages to the TSOs and further allows for higher incorporation of RES infeed, improved
security of power supply and increased network capacity. DSM incentives are divided into four cate-
gories: 1) wholesale and balancing markets, 2) grid and retail tariffs, 3) system services and 4) capacity
markets. The first two DSM incentives are price triggered, whereas the last two are system integrity
triggered. Therefore, DSM is expected to play a vital role in increasing the operational flexibility of the
interconnected transmission system.

2.4 Power system modeling

Power flow analysis plays a crucial role in the operational day-ahead planning of the interconnected
transmission system [27]. The flow of active power is computed by calculating the angular differ-

12
ences between the voltages at the buses, whereas the flow of reactive power is computed by calculating
the magnitude differences of the voltages at the buses. Moreover, the total active and reactive power
injected remains equal to the total active and reactive power extracted, plus the losses along the trans-
mission system. In power flow modeling, the buses are characterized into three types [27]:

• Load buses (PQ buses);

• Generator buses (PV, voltage controlled buses);

• Slack buses (swing buses).

2.5 Power flow equations

2.5.1 DC power flow

The DC power flow problems are formulated by taking into account the following two quantities: i)
active power injections and, ii) bus-bar voltage phase angles. In DC power flow modeling, the standard
power flow equations are simplified to make the system linear. This linearization of the system reduces
the computational time for obtaining the solution. The voltage profiles are approximated to 1 p.u.
and all the losses in the transmission system are ignored. The amount of reactive output and bus-bar
voltage magnitudes are eliminated, and active power injections result in a linear function of voltage
phase angles, and is given by :
N
B i j θi j ,
X
Pi = (2.2)
j =1

where B i j is the susceptance between the bus i and bus j.

The DC power flow computations are fast, and estimate approximate real power flows in the transmis-
sion system. Moreover, the assumptions made in the DC power flow equations provide a less accurate
power flow solution.

2.5.2 AC power flow

The AC power flow problems are formulated by taking into account the following four quantities: i)
active power injections, ii) reactive power injections , iii) bus-bar voltage magnitudes, and iv) bus-bar
voltage phase angles. These variables constitute a set of non-linear equations which are linearized and
solved iteratively in steps. Following are the standard AC power flow equations for real and reactive
power injections:
N
|Vi ||V j |(G i j cos θi j + B i j sin θi j ),
X
Pi = (2.3)
j =1

N
|Vi ||V j |(G i j sin θi j − B i j cos θi j ),
X
Qi = (2.4)
j =1

where P i and Q i is the amount of active power and reactive power entering bus-bar i, respectively, and
G i j is the conductance between the bus-bar i and bus-bar j.

13
2.6 Optimal power flow
Mathematically, the AC (and DC) OPF problem formulation is formulated as follows:

Minimize:
f (u), (2.5)

subject to
m(u) = 0, (2.6)

n(u) ≤ 0, (2.7)

and
u mi n ≤ u ≤ u max . (2.8)

where f(u) is the objective function, and m(u), n(u) represent the equality constraints and the inequality
constraints respectively. In case of an AC OPF, the optimization vector u represents the active (P) and
reactive power injections (Q), and bus-bar voltage magnitudes (V) and angles (θ ) [28]. Usually, there
are two main objectives in AC OPF which are investigated:

• Total operating cost minimization of active and reactive power output. The summation of cost
function can either be a piecewise linear, or a polynomial function:
ng
X
min f cost j (P j ) + f cost j (Q j ), (2.9)
j =1

where n g represents total number of generating units;

• Active power losses minimization of all the branches in the transmission system. The summation
of active power losses is given by:
nl
X
min f l oss j (P j ), (2.10)
j =1

where n l represents total number of branches.

In OPF problem formulation, the constraints are categorized into: i) equality constraints, and ii) in-
equality constraints. The equality constraints represent the set of power balance equations for active
and reactive power output respectively, given by:

m P (θ,V, P ) = 0, (2.11)

mQ (θ,V,Q) = 0. (2.12)

The inequality constraints represent the set of operational limits on the branch flows which are ex-
pressed in terms of non-linear function of the voltage magnitudes and angles, given by:

n f (θ,V ) 6= 0, (2.13)

n t (θ,V ) 6= 0. (2.14)

The variable constraints represent the permissible operating ranges for active and reactive injections,
and bus voltage magnitudes and angles:

P mi n, j ≤ P j ≤ P max, j , where j = 1 t o n g , (2.15)

14
 Q mi n, j ≤ Q j ≤ Q max, j , where j = 1 t o n g , (2.16)

Vmi n, j ≤ V j ≤ Vmax, j , where j = 1 t o n b , (2.17)

θr e f ≤ θ j ≤ θr e f where j = j r e f . (2.18)

2.6.1 Need for optimal power flow

The standard power flow problem does not take into account constraints while performing the power
flow computation. These constraints include: i) voltage level limits, ii) generation limits, and iii) circuit
overloadings. For these reasons, OPF has been developed to properly compute a dispatch that meets
all physical constraints, and various objective functions are used commonly, such as: i) reduction of
generation costs, and ii) reduction of system losses to obtain a more precise and accurate solution.
The proposed RBSA methodology uses AC OPF to perform: i) redispatch active generation, ii) load
curtailment, and iii) HVDC remedial action. More details on the RBSA methodology are discussed in
Chapter 3.

2.6.2 Extensions of OPF

The standard OPF does not guarantee on how secure the system is against unexpected outages of net-
work elements. In order to be able to compute a feasible generation dispatch that is both cost efficient
and also does guarantee that the system is N-1 secure, an extension in OPF is used. This extension
in OPF which is used to perform the N-1 contingency analysis is termed as security constraint OPF
(SC OPF). The objective is to find cost efficient dispatch, i.e., minimization of active power generation
costs such that the dispatch is N-1 secure for all operating conditions. Incorporating SC OPF guaran-
tees a secure grid operation, and makes the system much more reliable [29]. Moreover, an enhanced
OPF (EOPF) algorithm with additional constraints for transformer tap ratios, phase shift angles, HVDC
transmission lines and shunt elements is proposed under European commission FP7 research project
Umbrella [30]. This multi-objective optimization algorithm is efficient enough to synchronize all the
possible remedial actions in an optimized way, whilst taking into account all constraints.

An appropriate control strategy is required to guarantee a secure operation of HVDC grids against
any unplanned disturbances in the transmission network. The primary control action is carried out by
a communication-free scheme such as DC voltage droop control. Secondary control actions are needed
for the adaptation of the voltage and active power set-points in response to the uncertainties in loads
and RES in-feeds [31, 32]. Usually, an OPF-based tool is used to formulate the secondary control ac-
tions for the HVDC grids with various objective functions such as [33]: i) reduction of generation costs,
ii) increase of grid security, iii) reduction of system losses, and iv) higher integration of RES. In [34],
Iggland and co-workers study the operation of HVAC-HVDC interconnected meshed power systems.
They introduce a distributed OPF formulation and propose three operational configurations for inter-
linked grids: i) the super independent system operator (SISO) where the entire system (both HVAC
and HVDC) is controlled centrally by a single entity, ii) the technology separated independent system
operator (TISO) where only the HVDC grid is controlled by a separate entity and iii) the geographically
separated independent system operator (GISO) where the HVDC and HVAC grids are controlled based
on their geographical separation. Thus, it was emphasized that the controllability of HVDC transmission
should be incorporated in the OPF formulation in order to compute the cost efficient dispatch in each
zone/area.

15
Chapter 3

Risk-based power system security assessment

3.1 Overview
In this chapter, an overview on modeling of stochastic dependence is presented. A detailed description of
the proposed RBSA methodology which is used in [35, 36] is given. It is acknowledged that a risk-based
approach should be used by TSOs to quantify the expected system risk in day-ahead security assessment.
Moreover, the foundation for the case studies proposed in Chapter 5 on the impact of HVDC power set-points
is laid.

3.2 Modeling of stochastic dependence

The massive penetration of stochastic RES, together with the ongoing liberalization of electricity mar-
kets have significantly led to higher uncertainties in the operational planning of the European intercon-
nected transmission system. Consequently, this radical change in the existing transmission system has
reduced the ability of the TSOs to guarantee the overall security of power supply, while maintaining
equitably low operating costs. Uncertainties due to RES in-feeds, and fluctuating loads must be incorpo-
rated in the day-ahead planning of the power system. In the next sections, the mathematical concepts
of modeling the load forecast uncertainty, renewable generation uncertainty, dependency of the system
buses are discussed.

3.2.1 Spatial correlation

The spatial correlation of the snapshot data (point forecasts) should be considered in order to under-
stand the complete representation of the dependence structure of the random variables [37]. The spa-
tial dependency of the snapshot data X m (m = 1, ......., n) which are assumed to be normally distributed,
is described by the product moment correlation matrix C X .
 
ρ 2X 1 ρ X1 X2 ... ρ X1 Xn
ρ X2 X1 ρ 2X 2 ρ X2 Xn 
 
...
CX =  . ,
 
 .. .. .. .. 
 . . . 
ρ Xn X1 ρ Xn X2 ... ρ 2X n

17
where ρ stands for Pearson product moment correlation coefficient. By applying Cholesky decomposi-
tion, the matrix C X can be written as [37]:

C X = A × AT , (3.1)

where A is a lower triangular matrix. The correlated random variables Y are then constructed from the
multivariate normal distributions Zm (m = 1, ......., n) using the matrix C X [37]:

Y = A × Zm . (3.2)

Lastly, the correlated standard normals are transformed to uniforms by cumulative distribution function
(cdf), and by applying the inverse cdf transformation, the correlated uniforms are transformed into the
marginals [37].
U = φ (Y ), (3.3)

F = F −1 (U ). (3.4)

3.2.2 Rank correlation

Pearson product moment correlation measures the linear relationship between the random variables. It
captures the linear dependence, i.e., when the fluctuation in one random variable is perfectly correlated
with the fluctuation in other random variable. This coefficient does not evaluate for the relationships
which are not necessarily linear. Therefore, Spearman’s rank correlation is used alternatively which
allows to measure for all the monotonic relationship between the random variables. It assign ranks
to the samples, and then calculates the product moment correlation for the corresponding ranks. The
Spearman’s rank correlation of random variables X m (m = 1, ......, n) and Ym (m = 1, ......, n) with cdfs F X m
and FYm is given by [38]:
ρ r (X m , Ym ) = ρ(F X m (X m ), F Ym (Ym )), (3.5)

where ρ r stands for Spearman’s rank correlation.

3.2.3 Copula Theory

In statistical simulation, copula-based function generate multivariate distributions which allow to cap-
ture the spatial dependency of the system buses, such as RES and load forecast errors. In particular,
copulas can model the non-normal marginals to represent complex stochastic dependency. Abe Sklar in
1959 introduced copulas by describing them as functions that join together one-dimensional distribution
functions to form multivariate distribution functions [39]. The Sklar’s theorem states for every cdf F X m
and FYm of random variable X m (m = 1, ......, n) and Ym (m = 1, ......, n), there exist a copula such that [38]:

F X m Ym (x m , y m ) = C T (F X m (x m ), F Ym (x m )), (3.6)

If all F X m and FYm are continuous, then C T is unique.

3.3 An overview of state-of-the-art-research

In the recent years, extensive research has been conducted to develop risk-based tools, whilst taking into
account forecast uncertainty and dependence of system in-feeds. In [40] Ni, McCalley and co-workers
proposed the online risk-based security assessment (OL-RBSA) methodology based on the risk indices
to access the security levels of a transmission system with consideration of the possible contingency

18
 · Contingencies (generator/circuits)
· Topology
Construct Case
· Load data
· Initial dispatch

Selection of contingency

· Low voltage
· Voltage instability
Evaluation of risk indices · Cascading
· Circuit overload

Contingences left?

Summation of risk indices over the

set of contingencies

Figure 3.1: Flowchart for McCalley’s power system security methodology.

conditions in the near future state. Severity functions are employed to quantify the severity associated
with the following: i) low voltage, ii) voltage instability, iii) cascading, and iv) circuit overload. The
flowchart for the McCalley’s OL-RBSA methodology is illustrated in Figure 3.1, which summarizes the
procedure to evaluate for the risk indices using the severity functions.

In [41], Kirschen and co-workers proposed a probabilistic methodology based on MC framework to

assess the risk of time-dependent phenomena (TDP), such as transient instability and weather condi-
tion, and cascade and sympathetic tripping. This methodology uses MC sampling to evaluate the Value
of Security (VaS), which is defined to be the summation of the expected cost of the unscheduled outage,
and the cost of the curative measure. The flowchart for the Kirschen’s probabilistic methodology for
one MC sample is illustrated in Figure 3.2, which summarizes the procedure which is used to calculate
VaS in the power system. As can be understood, the McCalley’s methodology is based on a pre-defined
list of credible contingencies, whereas the Kirschen’s probabilistic methodology is based on the set of
credible as well as non-credible contingencies to assess the operational risk of the power system.

In [38],[42],[43] Papaefthymiou and co-workers emphasized the importance of copula-function based

MC sampling to capture the forecast uncertainty of the system RES inputs inherent to day-ahead fore-
casting. It has been proposed to use the copula theory which allows for sampling of spatial dependency
of the system inputs, and non-Gaussian distributions. It is analysed that the system can run into serious
problems when the spatial dependencies of the RES inputs are not taken into consideration. In [44],
Longatt and co-workers proposed a risk-based DC security assessment (RB-DCSA) methodology to en-
sure the security of supply in a highly stressed DC transmission system. Severity functions are employed
to quantify the severity associated with the following: i) DC- voltage excursion, ii) cable overload, and
iii) converter overload. The methodology takes into account the systemic and individual component
risk through a combination of MC sampling and fuzzy interface system. In [45], Li and fellow workers
proposed a methodology to study the transmission line overload in wind-power systems using the math-
ematical concept of severity functions, whilst taking into consideration wind and load-power generation
correlation. Moreover, commercial probabilistic planning tools such as GridView, REMARK, PROCOSE,

19
 · Topology
· Load data Construct Case
· Initial dispatch

· Transmission lines
· Generators
· Transformers
Simulate arbitrary outages
· Protection equipments

Restoration of load
generation equilibrium

Simulate power flow

No
Converged? Load shed

Yes

Any limit violations? Yes

(voltage/overloads)
Identify remedial measures

No

Calculate costs (VaS)

Figure 3.2: Flowchart for Kirschen’s probabilistic framework for one MC sample.

CREAM, and CORAL are based on DC PF computations, and may indeed perform poorly if voltage prob-
lems are a prime contributor to system risk [46].

In contrast with the other existing methods, the proposed RBSA methodology uses AC and AC OPF
combinations to perform a detailed risk assessment in day-ahead operation. The methodology employs
a more comprehensive analysis based on AC OPF for alert and emergency situations, which can al-
low TSOs to steer the system to secure states by providing valuable information on expected curative
measures. These key features of the proposed methodology allow for a accurate and complete risk
assessment by TSOs. More details on the proposed RBSA methodology are discussed in Section 3.4.

3.4 General overview: Proposed RBSA methodology

In this work the proposed RBSA methodology as presented in [35, 36] is employed to quantify the
expected system risk in day-ahead security assessment. The methodology allows to forecast the risk
that comes with a provided market dispatch for a day-ahead time horizon. The methodology involves
modeling of the random behavior of fluctuating loads and RES in-feeds. The methodology takes into
account load and RES forecast errors and spatial dependency of the system in-feeds through a com-
bination of a Monte-Carlo (MC) framework and copula theory. Moreover, the methodology uses full
AC PF and AC OPF combinations to obtain a very high level of accuracy in the day-ahead operational
planning. The methodology keeps track of the violation of acceptable system limits under a variety of
circumstances such as shortage of active generation, voltage violations and thermal branch overloads.
Also, the tool can simulate cascading events and it can be used to perform N-1 contingency analysis.

The risk for overloaded circuits, bus voltage deviations, lost loads and cascading events can be eval-
uated. For this, the methodology employs two tools: i) a fast-screening tool using the concept of

20
severity functions, and ii) a comprehensive analysis tool using extended AC OPF which provides the
valuable information on remedial actions. The proposed remedial actions can help TSOs to steer the
system to secure states under all possible uncertainties. There are several graphical display functions
included which can be used to visualize the results in a transparent manner, i.e., a GUI to generate
color risk-snapshots of the system displaying the regions of highest system stress on a nodal level, iso-
risk plots for 2D study parameter studies, and empirical pdfs/cdfs. The layout of the proposed RBSA
methodology is illustrated in Figure 3.3.

Outputs
Inputs
- Expected lost load
- MATPOWER test case
Risk-based security - Expected number of circuit overloads and
- Technical violation limits such as upper and
voltage violations
lower voltage limits and thermal ratings assessment methodology - AGC control for shift in generator
- Generation dispatch
power
- Fixed set of contingencies - Probabilities of risk related events such as
cascading events

Figure 3.3: Layout of proposed RBSA methodology.

3.4.1 Monte-Carlo framework

The outline of the proposed RBSA methodology is presented in Figure 3.4. Sample generation in MC
framework captures the complex stochasticity of probabilistic forecast of the random variables. The
methodology weighs the severity of the risk-based events that are related to outages, blackouts and
cascadings with the probability that it would occur in the near future. On the one hand, the severity is
related to the overall reliability parameters and violation in the threshold technical limits of the power
system. On the other hand, the probability of an event (e.g. voltage exceeding 1.05 p.u.) is evaluated by
simple enumeration, i.e., it equals the number of times the event happens divided by the total number
of MC samples.

· Contingencies (generators/
circuits)
Network data · Network topology
· Initial dispatch

m:=0
Monte-Carlo Framework

PLF
(Simulate for load uncertainty) Generate case M m:=m+1

No

Collect simulation data DSA tool m=N?

Yes

End

Figure 3.4: Outline of RBSA methodology.

21
3.4.1.1 Probabilistic load flow (PLF)

The proposed methodology uses copula-function based MC sampling to create a probabilistic load flow
(PLF) which captures the forecast uncertainty of the system inputs. For the numerical experiments, the
snapshot data i.e., point forecasts are in this work assumed to be normally distributed, but it must be
noted that the RBSA methodology can work with modeling of correlated non-normal distributions as
well, and empirical distributions by sampling from actual measurements. Here, the point forecast at
every bus m = 1, .........., n b is replaced by a normal marginal, given by:

P m ∼ N (µm , σm ), (3.7)

where P m is the active load at bus m , µm is mean of the active load and σm is the magnitude of the
standard deviation chosen to capture the load and RES forecast uncertainty. The reactive load Q m is
computed consequently, by assuming a constant power factor. It must be noted that if the ratio Q m /P m
is chosen to be constant, then the power factor remains constant. The stochastic dependence between
the system buses is described by computing the correlation matrix C . Zone-wise uniform correlation of
the active loads is considered to understand the complete representation of the dependence structure
of the random variables. Thus, the correlation matrix C of the power system having total number of m
buses is given by:  
C1 0 0 ... 0
 
0 C2 0 ... 0
C = . ,
 
 .. .. .. .. .. 
 . . . .
0 0 0 ... Cz

where the sub-matrix R z is the uniform correlation matrix for all the buses that are clustered in one
zone and is given by:  
1 ρr ρr ... ρr
ρ r ρr ρr 
 
1 ...
Cz =  . ,
 
 .. .. .. .. .. 
 . . . .
ρr ρr ρr ... 1

where ρ r represents the rank correlation.

The MC sampling of the random variables are performed based on the copula-based model, where
the obtained sample values correspond to the marginal distributions and the correlation matrix C . The
ranks of the correlated random variables Um (i = 1, ....., n), are then constructed from the multivariate
normal distributions using the matrix C and the Gaussion copula model. Further, by applying the
inverse cdf transformation, the correlated ranks Um are transformed in order to obtain the marginal
distributions of the random variables.

3.4.1.2 Deterministic security assessment

The flowchart of DSA algorithm for each MC simulation is presented in Figure 3.5. The methodology
involves performing a detailed deterministic load flow (DLF) for each MC sample. Many DLF computa-
tions are performed to have a high level of accuracy, and is thus computationally expensive. The DSA
algorithm consists of: i) full AC PF computations, ii) automatic generation control (AGC) for restoration
of power balance, iii) a cascading mechanism, and iv) two risk identifying tools. These risk identifying
tools are explained in the next section.

22
 Scenario
(One MC sample)

· Contingencies (generators/
circuits)
Check islanding Remove overloaded circuits
· Network topology
· Initial dispatch

Yes
Automatic generation control

Run AC PF

Yes
Converged? Any overloads?

No
No

Automatic generation control

Yes Overloads and

Re-dispatch generator power
voltage violations?

Yes Compute re-dispatch generator

Feasible solution? power and load shedding
System stablized

Re-dispatch generator No
power with load shedding
Feasible solution? System collapse

Figure 3.5: Flowchart of RBSA methodology for one MC sample.

23
3.5 Two risk identifying tools
The proposed RBSA methodology employs two risk identifying tools: i) Tool I: a fast-screening tool
using the concept of severity functions, and ii) Tool II: a detailed analysis tool using extended AC OPF
to provide valuable information to TSOs on expected curative measures. The details are as follows:

• Tool I uses McCalley’s concept of severity functions [40]. These functions are employed to quantify
the severity associated with the following: i) circuit overloads, and ii) voltage violations. These
functions are defined such that they start measuring severity if:

• the power flows exceeds the thermal rating by a factor 0.9, i.e., the power flows approach
their thermal limits. It is calculated by:

Pl
x= , (3.8)
P l ,max

where x is load factor, P l represents the actual power flowing through the transmission line
l , and P l ,max represents the thermal rating of the transmission line l .
• the bus voltage is close to either maximum voltage (Vmax ), or close to minimum voltage
(Vmi n ), i.e., the bus voltage magnitudes approach their maximum, or minimum limits. It is
calculated by:
Vb − Vb,nom
y= , (3.9)
sb
where Vb represents the true voltage at the bus b , Vb,nom represents the nominal voltage at
the bus b , and s b is the permissible range of the voltages.

The tool can be used to compute the probability of cascading events as well as the probability of
islanding.

• Tool II is based on Kirschen’s concept of risk indicators. To access the risk levels, the tool distin-
guishes the curative remedial actions into four categories:

(i) No remedial measures;

(ii) Remedial active re-dispatch;
(iii) Remedial active re-dispatch plus lost active load;
(iv) System collapse, or blackout.

These risk indicators are implemented within the AC OPF framework, and provide valuable infor-
mation on how to access the expected system risk in day-ahead operation. The expected amount
of remedial re-dispatch of generator’s active power is calculated by:

1 XN
E [ac t i ve r e − d i spat ch] = AR i ,
N i =1

where

N g en
1 X post pr e
AR i = (P j − Pj ), (3.10)
N g en j =1

pr e post
is the expected amount of active re-dispatch for a MC sample i . In Equation (3.10) P j and P j
represent the amount of generator’s active power before and after re-dispatching, respectively, N
represents the total number of MC samples, and N g en the total number of generators. Moreover,

24
this corresponds to the MC samples with risk levels 2, and 3, and is implemented within the
detailed analysis of Tool II. The expected amount of curative load shedding is calculated by:

1 XN
E [Load shed ] = LS i ,
N i =1

where

NX
1 bus
post pr e
LS i = (P k − Pk ), (3.11)
Nbus k=1
pr e post
represents the expected load shed for a MC sample i . In Equation (3.11) P k and P k represent
the amount of load’s active power before and after load shedding, respectively, and Nbus the
total number of buses. Moreover, this corresponds to the MC samples with risk levels 3, and is
implemented within the detailed analysis of Tool II. For samples with risk levels 4, the lost active
load for MC sample i is calculated by:

NX
1 bus
pr e
LS i = Pk .
Nbus k=1

Moreover, in this work it has been proposed to adapt the HVDC set-point as a part of curative
(cost free) remedial action to avoid active re-dispatch and shedding of load.

25
Chapter 4

High-Voltage Direct Current (HVDC)

technology

4.1 Overview
In this chapter, an overview of the HVDC technology, i.e., theory and modeling principles in order to under-
stand the operation of the HVDC transmission using RBSA methodology is presented. The current practice
concerning day-ahead HVDC grid operation used by TSOs in determining the HVDC set-points is discussed.

4.2 HVDC grid operation and power flow controllability

In contrast with the AC grids, the DC grids provide fast controllability of active and reactive power flow
which in turn improves the operational security of the grid by preventing congestions and risky events
such as cascading. Additionally, the HVDC technology provides ancillary services to the existing AC
infrastructures, and helps in preventing the congestions on parallel HVAC transmission lines [47]. The
HVDC technology supersedes the cost of expensive power electronic components such as converters and
is expected to provide the best economical solution beyond the break-even distance of about 600-800
km for overhead transmission lines and about 50 km for submarine cables, as illustrated in Figure 4.1.

Valuable research has been conducted to study the complex interaction of HVAC-HVDC interlinked
meshed power systems. In [48, 49], Beerten and fellow workers address the steady-state modeling of
hybrid AC and DC grids, discuss the open source software package MatACDC that they developed, and
it is emphasized that such tools are needed to be able to correctly analyze the steady-state interaction of
complex interconnected hybrid systems. The proposed tool computes iteratively the interrelated AC/DC
power flow solutions for multi-terminal HVDC (MTDC) systems embedded in existing AC networks. In
[50], Roald and co-workers proposed a methodology for the corrective control of HVDC transmission
lines and PSTs when subjected to both forecast uncertainty and contingencies. These corrective control
measures are based on affine control policies, where the set-points are adapted in response to the fore-
cast error. In this work, we are focused on the determining the best HVDC set-point using AC OPF when
the system is subjected to unavoidable uncertainty of inputs inherent to day-ahead forecasting.

27
 Cost

AC total costs
Break-even distance
AC losses
DC total costs
DC losses

AC investments costs
DC investments costs

DC terminal

AC terminal

Cable length

Figure 4.1: Cost comparison of HVAC and HVDC transmission technology.

Nowadays, the cooperation between the neighboring as well as non-neighboring TSOs is playing a vital
role in the transmission planning of the HVDC interconnections in meshed hybrid power systems [51].
Usually, the HVDC interconnections are operated between two energy markets based on the contractual
agreement or market regulations, and, depending on the operating region, the active power set-points
are regulated either by one or more TSOs. Market regulations limit the inherent controllability of the
HVDC technology to relieve the network from risky operating states during the occurrence of severe
contingencies. In [52] a coordination methodology is proposed, and it is shown more international
interaction among TSOs is needed for determining the optimal operating set-points for cross-border
HVDC interconnections.

ENTSO-E published a document to serve as a guide for software developers to build an IT platform
to exchange information relative to HVDC schedules between multiple TSOs. It identifies four main
steps that TSOs should take into account when determining the HVDC schedules based on day-ahead
operation: i) HVDC constraint determination, ii) Cross-border capacity computation, iii) HVDC config-
uration and iv) HVDC schedule calculation. Furthermore, it states that depending on TSOs the same
steps could be carried out again for purposes of re-matching and re-generation in intraday operation
[53].

Additionally, in Appendix 8 of the Nordic System Operation Agreement (SOA) [54], the trading ca-
pacity on the HVDC connections is limited to a maximum value of 600 MW per hour. This restriction in
the trading capacity has been imposed to deal with the flexible controllability in HVDC line flow with re-
spect to the AC networks, as this might introduce significant power deviations in HVDC power transfer.
This could make it for TSOs more difficult to handle the balance regulation of the meshed interlinked
systems. Therefore, ramping constraints of 600 MW per hour are imposed on the trading plans for
the cross-border Nordic HVDC interconnections. For example, there is a system operation agreement
for the SwePol HVDC connection installed between Sweden and Poland. This agreement contains the
information on the system operation regulations such as technical operational limits, emergency con-
trol functions and capacity allocation. This HVDC connection is regulated biyearly by the respective
TSO, and is isolated from the intraday capacity allocation. More details on the transmitted HVDC active
power and the ramping rate are given in Article 9 of the network code on HVDC, published by ENTSO-E
[55].

28
As can be understood, the current practice concerning HVDC operation does not allow TSOs to de-
termine which HVDC set-point gives the highest security of the power system when it is subjected to
the unavoidable uncertainty of inputs (fluctuations in load) inherent to day-ahead forecasting. For all
operating conditions, the power flow schedules between the HVAC and HVDC power systems are based
on the pre-defined converter set-points [56]. Moreover, the HVDC interconnections such as, EstLink
1, EstLink 2, SwePol and NordBalt are generally operated to its maximum NTC value to export power
across cross-border, further restricting TSOs to adapt the HVDC set-points in response to the forecast
uncertainty [57]. Therefore, the proposed RBSA methodology can be used in determining the overall
“best” HVDC set-point and help the power system to achieve the highest security level under all possible
uncertainties.

4.2.1 HVDC Topology: LCC vs. VSC

There are two main topologies for an HVDC transmission system:

• Line Commutated Convertor (LCC);

• Voltage Source Convertor (VSC).

LCC-HVDC is a mature technology with thyristor-based commutation [11]. It needs a strong AC grid
with adequate short circuit capacity, and reactive power compensating components to avoid the risk of
commutation failures. LCC-HVDC systems can allow the transmission of bulk power over long distances,
and exhibits relatively low losses. The DC current is kept constant, while the polarity of DC voltage is
altered to control the power flow. Moreover, it is inconvenient to operate a multi terminal LCC-HVDC
system with continuous change in power flows. Figure 4.2 illustrates the LCC-HVDC converter topology.

Thyristor bridge Constant current

AC

DC

Figure 4.2: LCC-HVDC converter topology.

VSC-HVDC technology is based on fully controlled insulated-gate bipolar transistors (IGBTs) [11]. It
uses high-frequency pulse width modulation (PWM) control and henceforth exhibits the flexibility in
the power flow. This flexibility in the power flow of the converter stations makes this technology more
suitable for a highly meshed interconnected electricity system. Moreover, it is also attractive from the
market perspective considering that VSC-HVDC technology allows to control the active power flow
faster than the LCC-HVDC technology. As active and reactive power can be controlled independently,
this technology can exhibit the operations on weaker networks. Figure 4.3 illustrates the VSC-HVDC
converter topology.

For these reasons, VSC-HVDC technology is preferred over LCC-HVDC and is most often used in new
HVDC transmission systems [11]. In this work we also consider VSC-HVDC technology only.

29
 IGBT bridge Constant voltage

AC

DC

Figure 4.3: VSC-HVDC converter topology.

4.3 HVDC transmission line modeling concepts

4.3.1 Steady state VSC-HVDC transmission line model
A VSC-HVDC transmission line model consists of two voltage-source converter stations connected to the
AC bus through a DC overhead line or cable. One converter in VSC-HVDC model operates as a rectifier,
whereas the other converter operates as an inverter. They allow the transfer of constant DC power
from rectifier to inverter station with high controllability. The VSC-HVDC converters are represented by
complex AC voltage sources connected to the AC bus of the network through a transformer impedance.
They are modeled as ordinary generators which extract or inject active and reactive power.

In VSC-HVDC converters, two independent control loops are employed for the regulation at the con-
verter stations, namely the active and the reactive control power loop [58]. In the active power control
loop, one converter controls the injected active power at the AC side of the grid whereas the other
converter controls the magnitude of DC voltage. In the reactive power control loop, both the converters
operate independently controlling either the injected reactive power or the voltage magnitude at the
AC side of the grid. Combining the aforementioned, each converter has an ability to operate either as
a PQ or PV node. As a PQ node, the converter is set to control the active and reactive power while as
a PV node, the converter is set to control the active power and the magnitude of AC voltage. In either
case, if one converter is set to control the active power flows, then the other will be set to regulate the
DC voltage of the point-to-point HVDC system.

The PQ capability chart determines the operating area of VSC-HVDC transmission line by controlling
the flow of the active and reactive power [59]. The maximum reactive power that either can be ab-
sorbed or injected is about 0.5 p.u. of the rated power of the VSC-HVDC light configuration [60]. The
upper and lower bound for the voltages should be set within the specific boundary limits.

In this work, we also set the maximum and minimum of reactive power to 0.5 p.u. of the HVDC
line rating. Furthermore, we set the converter voltage limits equal to 1 p.u. These limits were chosen
in [61] as well.

4.3.2 VSC-HVDC transmission line model in MATPOWER

There are several ways to model a steady state VSC-HVDC transmission line in the load flow analysis.
By leaving out details and assumptions for the converter data discussed in [62], the most precise model
for the VSC-HVDC transmission line can be reduced to the easiest model, as shown in Figure 4.4. A
simple mathematical representation of the VSC-HVDC connection has been used by MATPOWER [63].
This VSC-HVDC model provides us the desired accuracy in power flow simulations.

30
In MATPOWER, the steady-state VSC-HVDC transmission line is modeled as a pair of coupled gen-
erators (PV buses) with opposite sign of active power injection. One converter extracts active power
from one end of the line, whereas the other converter injects active power to the other end of the line.
Concerning the losses of the HVDC line: MATPOWER defines a linear relationship for the active power
loss in the VSC-HVDC transmission line, as given in Eq. (4.1) and Eq. (4.2).

P n = P m − P l oss , (4.1)
= (1 − l 1 )P m − l 0 , (4.2)

where P m is the active power extracted from bus m and P n is active power injected to bus n , and l 0 is
constant loss coefficient and l 1 is linear loss coefficient.

Bus m Bus n
VSC-HVDC Line
Pm Pn

P loss = (1 – l1) Pm - l0

Bus m Bus n

Pm
Pn

Figure 4.4: MATPOWER model of VSC-HVDC transmission line.

As the HVDC line is modeled as a pair of generators, the active power output can be controlled, as well
as the voltage set-points. If, in the load flow equations, the constraints for reactive power output of any
of the generators are reached, then the respective end of the VSC-HVDC transmission line operates as
a PQ bus. Hence, both active and reactive power output of each of the two generators are controlled.
Combining the aforementioned, the VSC HVDC transmission line can thus operate either in the PQ
control mode or in the PV control mode. In the AC OPF problem formulation the inclusion of a VSC-
HVDC transmission line is equivalent to the inclusion of a set of two ordinary generators in the network.
Hence, the constraints taken into consideration for the HVDC line during AC OPF are the exact same as
for ordinary generators:

(i) Constraints for maximum and minimum amount of active output;

(ii) Constraints for maximum and minimum amount of reactive output;

(iii) Voltage set-point of each of the two generators must lie within the same range as the bus they are
connected to;

31
(iv) Voltage angle of each of the two generators must be within the same range as the bus they are
connected to.

There are no additional costs for the VSC-HVDC dummy generators which are added to the objective
function during AC OPF. These generators can attain any value (within bounds), which results in mini-
mization of the objective function. This means that, for example, if the objective function is cost efficient
dispatch (minimization of active power generation costs), then the other generators, the loads and the
constraints on the elements determine how much power must be subtracted from the grid and injected
into the grid by the two dummy generators so that the cost function attains lowest overall costs. Thus,
the steady-state VSC-HVDC transmission line that has been modelled in MATPOWER [63] is employed
in the proposed RBSA methodology to quantify system risk inherent to day ahead forecasting.

32
Chapter 5

Experiments and results

5.1 Overview
In this chapter, the proposed RBSA methodology based on Monte-Carlo sampling is employed to investigate
the security of the system and to quantify the expected system risk in the operational day-ahead planning,
taking into account load and RES forecast uncertainty and dependence of system in-feeds. Several case
studies are performed to analyze how incorporation of HVDC transmission can be used as a smart grid
solution to improve power system security and lower the risk in different adjacent areas. Moreover, the
results obtained from the study reflect the need to properly adapt the HVDC set-points in response to the
actual operating situation, as it can be quite different from the day-ahead point forecast as a result of
uncertainty in load and RES in-feeds.

5.2 Experimental setup

In this section the setup of the experiment is analyzed. We first describe the implementation and hard-
ware used for the experiment, followed by a description of the test system and its adjustments, and
finally we present the experimental results of our analysis.

5.2.1 Implementation and hardware

The RBSA methodology has been modeled in MATLAB using the MATPOWER toolbox [63]. The nu-
merical experiments are performed on a desktop computer of a quad-core 3.50 Ghz Intel(R) Core(TM)
i5-4690 processor and 8 GB RAM.

5.2.2 Test system and its adjustments

For the case studies, the IEEE 24-bus reliability test system is used, see Figure 5.1. This system comprises
24 buses, 38 circuits (transmission lines and transformers), 32 thermal generators and one synchronous
condenser. The system is divided into two zones: Zone 1 and Zone 2. More details on the test system
can be found in the Appendix A. A single replacement of HVAC line 23-12 (right part of system) by the
HVDC transmission line is considered in order to facilitate more transfer of active power from zone 2
to zone 1. This line seems to be the best candidate of the 24-bus system to be replaced as: i) it has a

33
decent capacity (500 MW), ii) it has one of the longest line lengths (100 km) in the entire system, iii)
there is a big generation plant connected to it at what is in the base case the “sending end”, and iv) it is
an important line to transport energy from Zone 1 to Zone 2 and vice versa; it is directly connected to
the tie-station.

The following scenario is considered:

• The load in Zone I is increased by 20%, and the load in Zone II is decreased by 20%;

• The generation in Zone I is decreased by 20%, and the generation in Zone II is increased by 20%
of the original capacity;

• The generation cost in Zone I is increased by 20%.

Because of the high demand in Zone I, higher generation costs, and less generation capacity within
this zone, a scenario is created in which significant power flows are observed cross-border, i.e., the
HVDC transmission line will inject “cheap” power from Zone 2 into Zone 1, where generation is more
expensive.

Zone II

Zone I

Figure 5.1: The two-zone IEEE 24-bus reliability test system with HVDC line
between buses 23 and 12 (blue).

34
5.2.3 Result visualization
There are several graphical display functions included in the proposed RBSA methodology to provide
valuable information to TSOs on how to steer the system into a risk-averse state. The graphical inter-
pretation of the results are visualized using:

• GUI color risk-snapshots of the system displaying the regions of highest system stress on a nodal
level;

• Iso-risk plots displaying the risk for 2D study parameter studies;

• Empirical distribution functions of quantities of interest, e.g., line loading, bus voltage magnitude,
curative re-dispatch of active generation for each simulated grid point.

Moreover, for each simulated grid point, the results can be visualized in a transparent manner by giving
a detailed analysis using the following criteria:

• Overloaded circuits/branches (transformers and transmission lines) per zone;

• Participating generators in curative re-dispatch;

• Load curtailment.

5.3 Case studies

In this section the following three case studies are investigated:

(i) Sensitivity analysis on HVDC set-points;

(ii) Adjustment of HVDC set-points by TSOs as part of curative remedial actions;

(iii) A 2D study with two HVDC transmission lines.

It is shown how the proposed RBSA methodology can be used in determining the best HVDC set-
point when it is subjected to the unavoidable uncertainty of inputs (fluctuations in infeeds) inherent
to day-ahead forecasting. For all the case studies, we first compute a cost efficient dispatch assuming
a central dispatch market which is planning the system based on the expected average load at the
buses, i.e., point forecasts. Moreover, the default AC PF solver is replaced by a custom load flow solver
based on MATLAB’s fsolve() routine, and MATPOWER’s default AC OPF is replaced by the interior point
solver IPOPT from the COIN library [64]. The flowchart of the proposed HVDC-RBSA methodology is
presented in Figure 5.2.

5.3.1 Case study 1: Sensitivity analysis on HVDC set-points

In this case study we perform sensitivity analysis on HVDC set-points by investigating:

(i) Comparison of “security versus markets dilemma”: which set-point is the best compromise
between cost and risk;

(ii) Impact of forecast uncertainty and correlation: which set-point performs the best in different
scenarios (combination of HVDC set-points and, forecast uncertainty and correlation).

It must be noted that HVDC set-point must remain fixed during execution of the RBSA algorithm; in
this first study the set-point is not considered to be adaptable by TSOs.

35
 Base case

Initial dispatch

Conservative HVDC setpoints

1 2 3 …………….. n-1 n

RBSA methodology

Risk Indicators

Thermal overload
Active re-dispatch Lost active load Voltage severity
severity

Figure 5.2: Flowchart of proposed HVDC-RBSA methodology.1

5.3.1.1 Part 1: Comparison of “security versus markets dilemma”

In this part of the case study we investigate how incorporation of HVDC transmission can affect the
overall system risk, or risk in different areas/zones. We perform a Monte-Carlo simulation with the
proposed RBSA tool to assess the HVDC set-points in day-ahead operation. Here, we use the IEEE 24
bus reliability test system with the same adjustments as described in Section 5.2.2.

5.3.1.2 DSA setup

The key features of the DSA algorithm at the core of the RBSA tool are:

• Active re-dispatch is implemented within an extended AC OPF formulation with a cost penalty
term of 1 per MW;

• Lost active load is implemented within an extended AC OPF formulation with a cost penalty term
(the highest penalty) of 1000 per MW.

It must be noted that the cost penalty terms should be set to sufficiently large values. Arbitrary numbers
can be set by the operators, depending on the costs of the curative remedial actions. Here, the lost active
load has the highest penalty as we assumed that TSOs would use this alternative as a last resort to avoid
the risky system states.

5.3.1.3 Stochastic setup

Operational uncertainty is introduced to the system by considering uncertainty of in-feeds. For the
numerical experiments, the snapshot data, i.e., point forecasts are assumed to be normally distributed.

1 It
must be noted that in this thesis, the analysis is focused on two risk indicators: i) active re-dispatch, and ii) lost active load.
However, the proposed HVDC- RBSA methodology can be used to determine thermal overload severity, and voltage severity as
well.
The point forecast at every bus m = 1, ....., n b is replaced by a normal marginals to indicate the uncertainty
in the base case, given by:
P m ∼ N (µm , σm ), σ = f . µm (5.1)

where P m is the active load at bus m , µm is mean of the active load, and σm is the magnitude of the
standard deviation, i.e., f = 0.15 (15% of mean). Furthermore, we consider a zone-wise correlation
matrix with rank correlation ρ = 0.70, i.e., all buses within the same zone are equally dependent on
each other, as discussed in Section 3.4.1.1.

5.3.1.4 Case setup

Given the point forecast (the “best guess”) of the operational situation tomorrow, two scenarios are
investigated:

(i) “Market optimal”: The HVDC set-point is computed at the same time with the dispatch of active
generation using AC OPF with an economic objective function;

(ii) “Security optimal”: A fixed HVDC set-point is considered, and selected prior to computation of
the power dispatch. By doing so, a sensitivity analysis is obtained which can be used to determine
those HVDC power set-points that come with the lowest risk. It must be noted that as the OPF is
forced into a certain direction, probably the OPF cannot attain its economic best solution; it will
attain a sub-optimal dispatch, however, this may result in higher security of the system.

A 1D grid consisting of 25 grid points is obtained by varying the HVDC set-point in appropriate (small)
steps, i.e., 25 MW. For every grid point we generate a PLF consisting of 5.000 MC samples (P,Q).

5.3.1.5 Results and reflection

For the following evaluation, we perform 25 MC simulations (25 grid points) in order to determine the
system risk associated with every HVDC set-point. For every grid point we ran a MC simulation with
5.000 samples. Essentially, by doing so, for every HVDC set-point we obtain two values: i) risk (in terms
of expected active re-dispatch, and lost active load) and ii) generation dispatch costs. This allows for
a detailed comparison “risk vs. costs” using a 1D plot, as presented in Figure 5.3. It is seen that the
expected amount of active remedial re-dispatch gradually increases from point A to point B, and the
generation dispatch costs gradually decreases from point C to point D, see Figure 5.3 (left). A similar
trend can be identified for the expected amount of curative lost active load, see Figure 5.3 (right). Thus,
much higher risk is observed for the most optimal “cost set-point”, while lower risk is observed for the
most optimal “security set-point”. In Table 5.1 a comparison for the “security optimal” and “market
optimal” is presented2 .

Table 5.1: Comparison of “security optimal” and “market optimal” HVDC power
set-points.

Generation Active re-dispatch Lost active load

Grid point
dispatch cost[$/hr] [MW] [MW]
Security optimal set-point 18 59596 26.89 1.78

Market optimal set-point 25 59332 39.77 2.45

2 As it can be seen from Table 5.1, for our simulations, there is a slight difference in the generation dispatch costs. However, in
this thesis, it is emphasized, that there is a need for a security check to evaluate the market results.
The results obtained from the comprehensive analysis tool for: i) grid point 18, i.e., the optimal “se-
curity set-point”, and ii) grid point 25, i.e., the optimal “market set-point” are shown in the following
figures/tables:

• The zone-wise active re-dispatch and lost active load are presented in Table 5.2. It can be observed
that the generators in Zone 2 contribute more in re-dispatching the active power in order to
alleviate the problems associated with voltage violations in Zone 1. In addition, significant load
is shed in Zone 1 to attain lower system risk;

• The average re-dispatch of active power per generator for the grid point 18 is estimated around
26.88 MW, see Figure 5.4 (left), where the red marker G1 and G2 represents the generators
connected to bus 13, and bus 22, respectively. The average re-dispatch of power per generator for
the grid point 25 is around 39.76 MW, see Figure 5.4 (right), where the red marker G3 and G4
represent the generators connected to bus 13, and bus 22, respectively. It must be noted that: at
bus 13 and bus 22, the maximum generation is re-dispatched, and therefore these generators are
particularly important to restore the systems security;

• The average lost active load at every bus for market-optimal set-point is estimated around 1.77
MW, see Figure 5.5 (left), whereas for security optimal set-point is around 2.45 MW, see Figure
5.5 (right). It must be noted that: the red marker L1 and L2 represent the bus number (bus 6)
where the maximum load is shed to help the system to go to secure states;

• Figure 5.6 presents the number of overloaded circuits and Figure 5.7 presents the number of
voltage violations for every bus of the test system. As it can be seen from the figures, thermal
violations occur more often for the security optimal set-point than voltage violations. This means,
voltage violations have a significant contribution to the overall system risk, which could be ne-
glected if the risk assessment tools would have relied on DC approximations. The proposed RBSA
methodology successfully traced the problems related to voltage violations at the buses. As bus 3
turned out to be the weakest bus, the comprehensive analysis of voltage magnitudes after and
prior to RBSA check is illustrated in Fig 5.8. It must be noted that circuit 10, i.e., the line con-
necting bus 6 and bus 10 is one of the most critical branch in the entire network, which was also
highlighted by the RBSA check;

• Figure 5.9 presents the probabilities for the causes of problem, i.e., overloaded circuits, voltage
violations and non-convergence of power flows. In Figure 5.10 the probabilities of risk levels are
illustrated. These figures provide a great insight of where the risk comes from. The red marker
highlights the market optimal and security optimal set-points.

Table 5.2: Zone-wise re-dispatched active generation and lost active load.

Active re-dispatch Load active load

Grid point
[MW] [MW]
Zone 1 Zone 2 Zone 1 Zone 2
18 0.43 26.46 1.69 0.08

25 0.48 39.29 2.38 0.068

38
 ×10 4 ×10 4
8 100 8 10

Re-dispatched active generation [MW]

7.5

80
Cost of dispatch [$/hr]

Cost of dispatch [$/hr]

Lost active load [MW]

7

G (325, 59596)
6.5 60 6 5
H (500, 59332)

D (500, 59332)
6
C (325, 59596) F (500, 2.4509)
40
B (500, 39.7678)
5.5
E (325, 1.7734)

A (325, 26.8873)
5 20 4 0
-100 0 100 200 300 400 500 -100 0 100 200 300 400 500
HVDC set-points [MW] HVDC set-points [MW]

Figure 5.3: Comparison of dispatch costs and risk (active re-dispatch and lost active
load).

For grid point 18: HVDC set-point [325 MW] For grid point 25: HVDC set-point [500 MW]
4 4
Sum of active re-dispatch for every Sum of active re-dispatch for every
generator equals to 26.88 MW generator equals to 39.76 MW

3 3
Re-dispatched active generation for every generator

Re-dispatched active generation for every generator

G3

G1

2 2

1 1

0 0

-1 -1

G2
-2 -2

G4
-3 -3

-4 -4
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Generator Generator

Figure 5.4: Re-dispatched active power for every generator.

39
 For grid point 18: HVDC set-point [325 MW] For grid point 25: HVDC set-point [500 MW]
0.9 0.9

0.8 0.8
L2

0.7 0.7
Lost active load for every bus

Lost active load for every bus

0.6 0.6

0.5 0.5

0.4 0.4
L1
0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Bus Bus

Figure 5.5: Lost active load for every bus.

For grid point 18: HVDC set-point [325 MW] For grid point 25: HVDC set-point [500 MW]
1500 1500
Number of overloaded circuits

Number of overloaded circuits

1000 1000

500 500

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

Figure 5.6: Number of overloaded circuits at every bus. Total number of

MC-samples is 5000.

For grid point 18: HVDC set-point [325 MW] For grid point 25: HVDC set-point [500 MW]
800 800

700 700
Number of voltage violations

Number of voltage violations

600 600

500 500

400 400

300 300

200 200

100 100

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Bus Bus

Figure 5.7: Number of voltage violations at every bus. Total number of MC-samples
is 5000.

40
 No violation check RBSA check
900 900

800 800

700 700
Number MC samples

Number MC samples
600 600

500 500

400 400

300 300

200 200

100 100

0 0
0.8 0.85 0.9 0.95 1 1.05 0.94 0.95 0.96 0.97 0.98 0.99 1 1.01 1.02
Voltage (Bus 3) Voltage (Bus 3)

Figure 5.8: Improved voltage profiles at bus 3 with RBSA check.

0.9

0.8

0.7
Probability

0.6

0.5

0.4

0.3
No problems
0.2 Overloaded circuits
Voltage violations
0.1 Overloaded circuits plus voltage violations
Non-convergence of PF

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
HVDC set-points [MW]

Figure 5.9: Probabilities for causes of problems in the system.

0.9

0.8

0.7
Probability

0.6

0.5

0.4

0.3
No remedial action
0.2 Remedial re-dispatch
Remedial re-dispatch
plus lost active load
0.1 System collapse, or blackout

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
HVDC set-points [MW]

Figure 5.10: Probabilities of risk levels.

41
In order to provide more insight, we take the ratio of generation dispatch costs and risk to find the
most “optimal HVDC set-point” on overall, i.e., the best compromise between cost and risk. As shown
in Figure 5.11, the most “optimum set-point” that balances both cost and security is when the HVDC
set-point is set to 325 MW. Ratio of cost and risk may help TSOs to identify / determine the "optimal
set-point". By applying these set-points the operating point of the system moves to the one which comes
with lowest risk, and lowest cost on overall.

×10 -4 Ratio = Risk/cost [MW.hr/$]

14

13

12

11

10

5
R (325, 4.51 e-04)

4
-100 0 100 200 300 400 500
HVDC set-points [MW]

Figure 5.11: The best compromise between cost and risk is indicated by the blue dot.

5.3.1.6 Part 2: Impact of forecast uncertainty and correlation

In this part of the case study we assess the impact of forecast uncertainty and correlation on the expected
system risk in day-ahead operational planning. Here, we use the same test system with the same
adjustments, and the same DSA setup as described in part 1 of Case Study 1.

5.3.1.7 Case setup 1: Impact of forecast uncertainty

A 2D grid consisting of 25 × 11 grid points is obtained as follows:

• X axis: HVDC set-point is varied in appropriate (small) steps, i.e., 25 MW;

• Y axis: Forecast uncertainty is varied in appropriate steps, i.e., 0.02.

5.3.1.8 Results and reflection

For the following evaluation, we perform 25 × 11 = 275 MC simulations in order to determine the
sensitivity of the expected system risk with respect to HVDC set-points and forecast uncertainty. For
each grid a MC simulation is performed with 5.000 samples. Figure 5.12 presents the expected amount
of remedial re-dispatch (see left) and the expected amount of curative lost active load (see right). It can
be seen that when forecast uncertainty is high, the risk perhaps becomes higher. Moreover, the HVDC

42
set-points between 325 MW and 400 MW performs the best when the correlation is fixed to a value of
0.7.

Figure 5.12: Impact of forecast uncertainty: Risk in terms of re-dispatch active load
and lost active load.

5.3.1.9 Case setup 2: Impact of correlation

A 2D grid consisting of 25 × 11 grid points is obtained as follows:

• X axis: HVDC set-point is varied in appropriate (small) steps, i.e., 25 MW;

• Y axis: Correlation is varied in appropriate steps, i.e., 0.1.

Figure 5.13: Impact of correlation: Risk in terms of re-dispatch active load and lost
active load.

43
5.3.1.10 Results and reflection

In Figure 5.13, a similar trend for remedial expected active re-dispatch and curative lost load as de-
scribed in Section 5.3.1.8 is observed with respect to the combination of HVDC set-points and correla-
tion. It can be seen that the HVDC set-points ranging between 325 MW and 400 MW performs the best
when forecast uncertainty is set to a value of 0.15. Thus, the market optimal set-point where HVDC
link is operating to its full capacity does not necessarily provide the most secure set-point to steer the
system away from risky situations.

5.3.1.11 Visualization tool

The visualization tool has been developed to display the regions of highest system stress on a nodal
level. Figure 5.14 and Figure 5.15 present the GUI color average operating state for HVDC set-point
325 MW (= security optimal set-point) and HVDC set-point 500 MW (= market optimal set-point),
respectively. For security optimal set-point, the re-dispatching of the generators, and the load shedding
at the buses are highlighted using the color code, where the color brown corresponds to the generators
and loads where the re-dispatching, and load shedding is maximum, respectively. It can easily be
observed that bus 13 and 22 contribute maximum to re-dispatch, and maximum load is shed at bus 6,
see Figure 5.14 (left). The load factor is shown in Figure 5.14 (right), where the lines 6-10, 16-17, and
16-19 have the highest load factor (green corresponds to the lowest, and red corresponds to the highest
values for load factor). A similar trend can be observed for the market optimal set-point, where bus 13,
15, and 22 contribute maximum to re-dispatch, and again maximum load is shed at bus 6, see Figure
5.15 (left). Moreover, the critical lines remain almost the same as the security optimal set-point. A
detailed analysis on the outcome of the generation dispatches, loads, voltage magnitudes at the buses,
and power flowing through the lines are provided in Appendix B. Thus, by doing a sensitivity analysis
for HVDC set-points; it can be concluded that TSOs would set the HVDC set-point to 325 MW as being
the best compromise for both cost and security, and also with respect to the forecast uncertainty and
correlation.

5.3.2 Case study 2: Adaptability of HVDC set-point by TSO as curative remedial

action
In this case study we investigate how the adaptability of the HVDC set-points can enable TSOs to
improve the system security under possible set of uncertainties. We perform a Monte-Carlo simulation
with the proposed RBSA tool to assess the adaptability of HVDC set-point as a curative remedial action
in day-ahead operation. Here, we use the same test system with the same adjustments, and the same
stochastic and case setup as described in Case Study 1. Moreover, a level of risk is added to the four
levels defined previously. We now have:

(i) No remedial measures;

(ii) Remedial HVDC adaptation;

(iii) Remedial re-dispatch;

(iv) Remedial re-dispatch plus lost active load;

(v) System collapse, or blackout.

It must be noted that the HVDC adaptation is inserted as the new level 2, so adapting the HVDC set-point
is first tried, prior to curative remedial actions such as active re-dispatch and load shedding.

44
 IEEE RTS 24-bus IEEE RTS 24-bus

< <
<

<
<< <<
< < < <
<< <<
> << > <<
> > <

> > <

<

<
< <

<

<
<< < << < HVDC
HVDC
0.98 0.98

<

<
<

<
< < < <
<

<
<

<
>

>
<

<
< <
< <
<

<

<

<
< <
> > > >
<

<
<

<
<

<
< <
< <
<

<
>

>
>

>
< <
> <
< > <
<
<

<
>

>
>

>
< <
>

>
<

<
> > > >

Figure 5.14: Average operating state when HVDC set-point is set to 325 MW.

IEEE RTS 24-bus IEEE RTS 24-bus

< <
<

<

<< <<
< < < <
< <
< <
> >> > >>
> > <

> > <

<

< < <

<

<

<< < HVDC << <

0.99 HVDC
0.99
<

<
<

<

< < < <

<

<
<

<
>

>
<

<

< <
< <
<

<

<

<

< <
> > > >
<

<
<

<
<

<

< <
< <
<

<
>

>
>

>

< <
> <
< > <
<
<

<
>

>
>

>

< <
>

>
<

<

> > > >

Figure 5.15: Average operating state when HVDC set-point is set to 500 MW.

45
5.3.2.1 DSA setup

The key features of the DSA algorithm consists of:

• HVDC set-point adaptation is enforced within an extended AC OPF formulation with a cost penalty
term of 0 per MW. This corresponds to the MC samples with risk levels 2, implemented within the
detailed analysis of Tool II;

• Active re-dispatch is implemented within an extended AC OPF formulation with a cost penalty
term of 1 per MW. This corresponds to the MC samples with risk levels 3, and 4, implemented
within the detailed analysis of Tool II;

• Load active load is implemented within an extended AC OPF formulation with a cost penalty term
of 1000 per MW. This corresponds to the MC samples with risk levels 4 and 5, implemented within
the detailed analysis of Tool II.

5.3.2.2 Result and reflection

For the following evaluation, we perform 25 MC simulations (25 grid points) to quantify system risk in
the day-ahead security assessment when:

(i) The HVDC set-point is not able to adapt in curative remedial action, i.e., the “dumb” approach;

(ii) The HVDC set-point is able to adapt in curative remedial action, i.e., the “smart” approach.

In both the scenarios, for each grid point a MC simulation is performed with 5.000 samples. Essentially
by doing so, we obtain a set of HVDC set-points in case HVDC adaptation (= “smart” approach) to
resolve the problems. This allows for a detailed comparison between “dumb” and “smart” approach,
see Figure 5.16 (green represents the “dumb” approach, and blue represents the “smart” approach). It is
seen that the expected amount of remedial active re-dispatch is less for “smart” approach as compared to
the “dumb” approach. Figure 5.17 presents the empirical PDF using 5.000 samples for HVDC set-point
325 MW (= security optimal set-point) where the big spike indicates that most of the time, there is no
need for HVDC set-point to adapt; however, in few cases the HVDC power set-point must be lowered
and occasionally, the full capacity of the HVDC link must be used to steer away from risky situations. By
being able to adapt the HVDC set-point, we show that more serious and more costly remedial actions
can be avoided (such as active re-dispatch, shedding of load). Therefore, the HVDC set-point should
not be fixed to only a single value, rather it should be able to adapt to help the system go to secure
states in response to all possible uncertainties in real-time situation.

5.3.3 Case study 3: 2D study with two HVDC transmission lines

The two-zone IEEE 24-bus reliability test system with HVDC line between buses 23 and 12 is stressed
on right part of the system, as shown in Figure 5.14 and Figure 5.15. In order to reduce stress on one
side of the system, another HVDC replacement on the left part of the system between buses 15 and
24 is considered. In this case study we investigate how the set-points of two HVDC transmission lines
can work together to find the most optimum risk vs. cost operating set-points. We perform a Monte-
Carlo simulation with the proposed RBSA tool to check how two HVDC set-points coordinate to steer
the system away from risk. Here, we use the test system and the same DSA and stochastic set-up as
described in Case Study 1.

46
 Re-dispatched active generation [MW] Lost active load [MW]
100 100 10 10

HVDC as not a part of remedial measure

HVDC as not a part of remedial measure

HVDC as a part of remedial measure

HVDC as a part of remedial measure

80 80 8 8

60 60 6 6

40 40 4 4

20 20 2 2

0 0 0
-100 0 100 200 300 400 500 -100 0 100 200 300 400 500
HVDC set-points [MW] HVDC set-points [MW]

Figure 5.16: Comparison of “dumb” (blue) and “smart” approach (green).

4000

3500

3000
Number of MC samples

2500

2000

1500

1000

500

0
-400 -300 -200 -100 0 100 200 300 400 500
HVDC set-points [MW]

Figure 5.17: Adaptability of HVDC set-point 325 MW.

5.3.3.1 Case setup

A 2D grid consisting of 25 × 8 grid points is obtained as follows:

• Line 1 (Bus 23 to bus 12): HVDC set-point [-100, 500] is incremented in steps of 25 MW;

• Line 2 (Bus 15 to bus 24): HVDC set-point [200, 375] is incremented in steps of 25 MW.

HVDC line 1 can be controlled to its full capacity between -100 MW and 500 MW, whereas HVDC line 2
can be controlled partially maybe due to market restrictions, so therefore, HVDC line 1 can offer more
options to the TSOs to withstand the uncertainties in load.

47
5.3.3.2 Results and reflection

For the following evaluation, we perform 25 × 8 = 200 MC simulations to quantify the system risk
associated with the coordination of two HVDC set-points. If the HVAC transfer from buses 15 to 24 was
275 MW in the single HVDC case, i.e., between buses 23 and 12; replacing the HVAC by HVDC line on
left part of the system did the system good as TSOs can now steer to more green areas, and have more
options to stress the system away from risky situations, see Figure 5.18. Much lower risk is observed
when the set-point for HVDC line 1 are set in the range [375, 475 MW], and the set-point for HVDC line
2 are set in the range [300, 375 MW]. That means, HVDC lines can serve as a tool to shift generation.

Figure 5.18: Risk in terms of re-dispatch active load and lost active load.

Figure 5.19 illustrates that the ratio of cost and risk can perhaps help TSOs to to identify / determine the
“optimal” set-point. As it can be seen that the best “optimal” set-points that balances both generation
dispatch costs and risk are indicated by green areas. By applying these set-points the operating point
of the system moves to the most “optimal” set-point which comes with the best compromise between
cost and risk. The coordination of two HVDC set-points offers more options to TSOs to steer the system
away from serious problems, and if not coordinated, we may end up in risky system states. Thus, smart
transmission can also have a negative impact affect on the system operational security.

Figure 5.19: Ratio of cost and risk.

48
49
Chapter 6

Conclusions and future research

In this chapter, the main findings and conclusions are presented, followed by recommendations for follow
up research work.

6.1 Summary
This thesis proposed the operational strategies for HVDC transmission in smart grid by addressing the
security versus markets dilemma. Chapter 2 laid the foundation to understand the smart grid operation,
and power flow modeling, necessary to investigate the power system security in the daily operational
planning. Chapter 3 described the proposed RBSA methodology used to access the stochastic nature of
system in-feeds. Chapter 4 discussed the current practices used by TSOs concerning day-ahead HVDC
grid operation. The steady-state VSC- HVDC transmission that has been modelled in MATPOWER is
described in detail. Chapter 5 proposed the need to forecast the risk associated with HVDC set-point
that comes with a provided market dispatch inherent to day-ahead operation. It is observed that there
is a perceived trade-off between costs and security. It was found that the coordination of the smart
transmission plays an important role to lower risky system states.

6.2 Conclusions
The main conclusions can be summarized as follows:
• Proposed RBSA methodology can be used to determine “best” HVDC set-point: For assessing
the power system security, it is proposed that TSOs should employ the proposed RBSA method-
ology to determine the best HVDC set-point when it is subjected to the unavoidable uncertainty
of inputs (fluctuations in load and RES) inherent to day-ahead forecasting. It was shown that: i)
without proper adaptability of the HVDC set-point, i.e., HVDC operated at a single, precomputed
set-point, not taking into account forecast uncertainty, we could run into severe problems that
can result in unnecessary high risk, and ii) with proper adaptability of the HVDC set-point, i.e.,
HVDC operated using a varying set of set-points, we could help to improve system security under
possible uncertainties. Given a specific scenario, it was shown that HVDC set-point may improve
capacity, security and efficiency of the grid (in terms of money). Moreover, TSOs cannot always
use the cost optimal HVDC set-point computed by the market as there may be cases where the
system can run into serious problems, e.g., cascading events or even blackouts;

50
 • HVDC as a part of remedial curative measure: In order to alleviate the expected risk in day-
ahead security assessment, the adaptability of HVDC set-points as a part of remedial curative
measure was proposed. As HVDC have the lowest overall cost, the HVDC set-point should be able
to adapt, and further more serious and more costly remedial actions such as active re-dispatch
and load shedding can be avoided.

• Coordination of HVDC set-points: A study with two HVDC transmission lines shows how the
set-points of the two HVDC transmission lines must be set by TSOs in order to find the most
optimum risk vs. cost operating set-points. Two HVDC lines allow to reduce stress on one side
of the system as the HVDC lines serve as a tool to shift generation. It more or less has the same
effect as using a different generation shift key (GSK). Varying two HVDC set-points at the same
time gives more options (degrees of freedom) to TSOs to steer the system away from risk.

6.3 Future research

The recommendations for the follow-up research can be summarized as follows:

• The proposed RBSA methodology based on the combination of MC sampling and full AC PF de-
mand significantly complex computational resources. In order to save the computational time, we
ran a MC simulation with 5.000 samples for all the case studies. However, in certain cases higher
accuracy may be demanded, which could require in the order of 100.000 or even 1 million MC
samples. In order to be still able to perform the risk analysis on time, one has several options: i)
more hardware resources: as MC is easy to parallelize, one can simply divide the entire set of MC
samples as well as the different grid points used for 1D and 2D studies over multiple computers.
TSOs typically own a big cluster of computing nodes, so the proposed RBSA method can read-
ily be used by them, ii) speed up the computations by more efficient algorithms (faster solvers),
more efficient implementation and different programming language (e.g., C instead of Matlab),
and iii) variance reduction techniques can be used to reduce the required number of MC samples
to obtain a predefined level of accuracy;

• HVDC set-point in RBSA algorithm is set to a single value, and does not adapt during the real-time
framework. However, it would be worthwhile to implement HVDC set-point adaptability during
load balancing in order to steer the system away from cascading events. In this way, the HVDC
set-point can automatically be adjusted by the RBSA algorithm, and TSO can adapt the set-point
on a real-time framework;

• Market designs such as central dispatch, and self-dispatch markets can be evaluated based on the
curative remedial actions. As the HVDC set-point in this study can be adjusted by TSOs as being
part of curative remedial actions avoiding more costly remedial actions including re-dispatch, and
lost active load, therefore the effect on market designs can shed more light on the overall costs of
the entire system;

• It would be interesting to analyze how the optimal placement of an HVDC transmission line can
affect the overall system risk, or risk in adjacent areas. This can be simulated by generating critical
snapshots, by looking at different combinations of load and RES: i) high load - min RES, ii) high
RES - min load, iii) high RES on one side, and iv) combination of the above. For each of these
snapshots appropriate dispatch for conventional generation (CG) by solving a default AC OPF cost
problem can be computed. Finally, the risk for different HVDC systems, i.e., different placement /
different capacity of the HVDC transmission line can be then evaluated.

51
Appendix A

System parameters

A.1 Bus parameters

Table A.1: Bus parameters.

Pload Qload baseKV Vmin Vmax

Bus Type Zone
(MW) (MVAr) (kV) (p.u.) (p.u.)
1 2 108 22 138 1 1.05 0.95
2 2 97 20 138 1 1.05 0.95
3 1 180 37 138 1 1.05 0.95
4 1 74 15 138 1 1.05 0.95
5 1 71 14 138 1 1.05 0.95
6 1 136 28 138 1 1.05 0.95
7 2 125 25 138 1 1.05 0.95
8 1 171 35 138 1 1.05 0.95
9 1 175 36 138 1 1.05 0.95
10 1 195 40 138 1 1.05 0.95
11 1 - - 230 2 1.05 0.95
12 1 - - 230 2 1.05 0.95
13 3 265 54 230 2 1.05 0.95
14 2 194 39 230 2 1.05 0.95
15 2 317 64 230 2 1.05 0.95
16 2 100 20 230 2 1.05 0.95
17 1 - - 230 2 1.05 0.95
18 2 333 68 230 2 1.05 0.95
19 1 181 37 230 2 1.05 0.95
20 1 128 26 230 2 1.05 0.95
21 2 - - 230 2 1.05 0.95
22 2 - - 230 2 1.05 0.95
23 2 - - 230 2 1.05 0.95
24 1 - - 230 2 1.05 0.95

52
A.2 Generator parameters

Table A.2: Generator parameters.

Pgen Qgen Voltage Pmax Pmin

Bus Unit
(MW) (MVAr) (p.u.) (MW) (MW)
1 1 10 0 1.035 20 16
1 2 10 0 1.035 20 16
1 3 76 0 1.035 76 15.2
1 4 76 0 1.035 76 15.2
2 1 10 0 1.035 20 16
2 2 10 0 1.035 20 16
2 3 76 0 1.035 76 15.2
2 4 76 0 1.035 76 15.2
7 1 80 0 1.025 100 25
7 2 80 0 1.025 100 25
7 3 80 0 1.025 100 25
13 1 95.1 0 1.02 197 69
13 2 95.1 0 1.02 197 69
13 3 95.1 0 1.02 197 69
14 1 0 35.3 1.014 0 0
15 1 12 0 1.014 12 2.4
15 2 12 0 1.014 12 2.4
15 3 12 0 1.014 12 2.4
15 4 12 0 1.014 12 2.4
15 5 12 0 1.014 12 2.4
15 6 155 0 1.014 155 54.3
16 1 155 0 1.017 155 54.3
18 1 400 0 1.05 400 100
21 1 400 0 1.05 400 100
22 1 50 0 1.05 50 10
22 2 50 0 1.05 50 10
22 3 50 0 1.05 50 10
22 4 50 0 1.05 50 10
22 5 50 0 1.05 50 10
22 6 50 0 1.05 50 10
23 1 155 0 1.05 155 54.3
23 2 155 0 1.05 155 54.3
23 3 350 0 1.05 350 140

53
A.3 Transmission line parameters

Table A.3: Transmission line parameters.

Transmission
Resistance Reactance Susceptance Rating
Circuits From bus To bus distance
(p.u.) (p.u.) (p.u.) (MVA)
(miles)
1 1 2 3 0.0026 0.0139 0.4611 175
2 1 3 55 0.0546 0.0211 0.0572 175
3 1 5 22 0.0218 0.0845 0.0229 175
4 2 4 33 0.0328 0.1267 0.0343 175
5 2 6 50 0.0497 0.1920 0.0520 175
6 3 9 31 0.0308 0.119 0.0322 175
7 3 24 0 0.0023 0.0839 - 400
8 4 9 27 0.0268 0.1037 0.0281 175
9 5 10 23 0.0228 0.0883 0.0239 175
10 6 10 16 0.0139 0.0605 2.4590 175
11 7 8 16 0.0159 0.0614 0.0166 175
12 8 9 43 0.0427 0.1651 0.0447 175
13 8 10 43 0.0427 0.1651 0.0447 175
14 9 11 0 0.0023 0.0839 - 400
15 9 12 0 0.0023 0.0839 - 400
16 10 11 0 0.0023 0.0839 - 400
17 10 12 0 0.0023 0.0839 - 400
18 11 13 33 0.0061 0.0476 0.0999 500
19 11 14 29 0.0054 0.0418 0.0879 500
20 12 13 33 0.0061 0.0476 0.0999 500
21 12 23 67 0.0124 0.0966 0.2030 500
22 13 23 60 0.0111 0.0865 0.1818 500
23 14 16 27 0.0050 0.0389 0.0818 500
24 15 16 12 0.0022 0.0173 0.0364 500
25 15 21 34 0.0063 0.0490 0.1030 500
26 15 21 34 0.0063 0.0490 0.1030 500
27 15 24 36 0.0067 0.0519 0.1091 500
28 16 17 18 0.0033 0.0259 0.0545 500
29 16 19 16 0.0030 0.0231 0.0485 500
30 17 18 10 0.0018 0.0144 0.0303 500
31 17 22 73 0.0135 0.1053 0.2212 500
32 18 21 18 0.0033 0.0259 0.0545 500
33 18 21 18 0.0033 0.0259 0.0545 500
34 19 20 27.5 0.0051 0.0396 0.0833 500
35 19 20 27.5 0.0051 0.0396 0.0833 500
36 20 23 15 0.0028 0.0216 0.0455 500
37 20 23 15 0.0028 0.0216 0.0455 500
38 21 22 47 0.0087 0.0678 0.1424 500

54
A.4 Generation cost parameters

Table A.4: Generation cost parameters.

Startup cost Shutdown cost Cost function Cost function Cost function
Bus
($ ) ($ ) (c2) (c1) (c0)
1 1500 0 0 130 400.6849
1 1500 0 0 130 400.6849
1 1500 0 0.01414 16.0811 212.3076
1 1500 0 0.01414 16.0811 212.3076
2 1500 0 0 130 400.6849
2 1500 0 0 130 400.6849
2 1500 0 0.01414 16.0811 212.3076
2 1500 0 0.01414 16.0811 212.3076
7 1500 0 0.052672 43.6615 781.521
7 1500 0 0.052672 43.6615 781.521
7 1500 0 0.052672 43.6615 781.521
13 1500 0 0.00717 48.5804 832.7575
13 1500 0 0.00717 48.5804 832.7575
13 1500 0 0.00717 48.5804 832.7575
15 1500 0 0.32841 56.564 86.3852
15 1500 0 0.32841 56.564 86.3852
15 1500 0 0.32841 56.564 86.3852
15 1500 0 0.32841 56.564 86.3852
15 1500 0 0.32841 56.564 86.3852
15 1500 0 0.008342 12.3883 382.2391
16 1500 0 0.008342 12.3883 382.2391
18 1500 0 0.000213 4.4231 395.3749
21 1500 0 0.000213 4.4231 395.3749
22 1500 0 0 0.001 0.001
22 1500 0 0 0.001 0.001
22 1500 0 0 0.001 0.001
22 1500 0 0 0.001 0.001
22 1500 0 0 0.001 0.001
22 1500 0 0 0.001 0.001
23 1500 0 0.008342 12.3883 382.2391
23 1500 0 0.008342 12.3883 382.2391
23 1500 0 0.004895 11.8495 665.1094

55
A.5 HVDC line specifications

Table A.5: HVDC line specifications.

HVDC line capacity [MW] 500

HVDC transfer in base case [MW] 500

Maximum reactive power at converter sending end [MVAr] 250

Minimum reactive power at converter sending end [MVAr] -250

Maximum reactive power at converter receiver end [MVAr] 250

Minimum reactive power at converter receiver end [MVAr] -250

Voltage at converter receiver end [p.u.] 1

Voltage at converter sending end [p.u] 1

Constant losses [MW] 11

Linear losses coefficient [MW] 0

HVDC power received at other end of line [MW] 489

56
Appendix B

Visualization tool

B.1 Detailed analysis when HVDC set-point is set to 325 MW

IEEE RTS 24-bus
266.8
480.0 480.0 360.0
1.050 1.050 1.050

<
<

< < 39.6

1.040 39.6 185.4
278.2

< 165.4 < 1.050

775.8
436.6 <
79.9 144.8 74.4 <
109.9
120.5 > < < 30.1
30.1
74.4
<

<

1.027 0.01.031 1.042

<

<
>

488.8
138.2 306.5
286.6 286.6 102.7
<<
>

155.4
<
1.009 325.0
<

1.023
<

322.1
253.8
131.6 179.9 < < 1.049
337.2
<
<

212.3
161.3
>

0.958 0.997 1.031

207.0
<

<
216.8 210.4 <
<

<

318.2 201.6 217.1 273.5 280.1

0.967 1.026
< 187.7
> 47.7 > 0.997
<
<

163.3
<

< 106.7 1.005

< 87.2
<

89.0 129.4 62.0

>

0.992
>

76.6 85.3 <

30.3 > <
<
1.016
14.9
<

29.3 0.958
>

205.2
>

<
>

24.6
<

> 54.1 >

1.039 1.041 0.993
153.6 153.6 75.0
129.7 116.3 149.9

Figure B.1: Network parameters when HVDC set-point is set to 325 MW.

57
B.2 Detailed analysis HVDC set-point is set to 500 MW

IEEE RTS 24-bus

266.8
480.0 480.0 360.0
1.050 1.050 1.050

<

<
< < 43.3
1.042 43.3 184.3
285.3

< 166.9 < 1.050

717.2
445.8 <
79.9 144.8 16.0 <
162.7
229.2 > > > 52.8
52.8
16.0
<

<
1.032 0.01.033 1.043

<
<
>

463.9
176.3 201.7
282.3 282.3 102.7
<<
>

155.4
<
1.026 500.0

<
1.025

<
309.0
253.8
169.5 150.1 < < 1.010
329.5
<
< 212.3
62.9
>

0.955 0.983 0.986

207.0
<

<
216.8 210.4 <
<

<
309.8 190.9 243.1 256.3 310.5
0.951 0.995
< 184.6
> 32.7 > 0.971
<
<

163.3
<

< 109.9 0.972

< 90.7
<

89.0 136.8 67.7

>

0.963
>

70.3 85.3 <

23.9 > <
<
0.984
7.0
<

26.1 0.958
>

205.2
>

<
>

55.1
<

> 49.4 >

1.006 1.009 1.016
121.0 150.2 75.0
129.7 116.3 149.9

Figure B.2: Network parameters when HVDC set-point is set to 500 MW.

58
Bibliography

[1] E. Commission, COM(2010) 639 final: Energy 2020: A strategy for competitive, sustainable and
secure energy, (2010).

[2] G. Papaefthymiou and K. Dragoon, Towards 100% renewable energy systems: Uncapping power
system flexibility, Energy Policy 92, 69 (2016).

[3] J. McCalley, Security assessment: decision support tools for power system operators, (2000).

[4] C. V. Fabien Roques, Options for the future of power system regional coordination, (2016).

[5] A. L’Abbate, G. Migliavacca, U. Hager, C. Rehtanz, S. Ruberg, H. Ferreira, G. Fulli, and A. Purvins,
The role of facts and HVDC in the future paneuropean transmission system development, in AC and
DC Power Transmission, 2010. ACDC. 9th IET International Conference on (IET, 2010) pp. 1–8.

[6] P. Buijs, D. Bekaert, S. Cole, D. Van Hertem, and R. Belmans, Transmission investment problems in
Europe: Going beyond standard solutions, Energy Policy 39, 1794 (2011).

[7] E. Commission, Renewable energies in the 21st century: building a more sustainable future, (2007).

[8] E. Commission, Energy Roadmap 2050 – Impact assessment and scenario analysis, (2011).

[9] Entso-e, Press release: Flow-Based methodology for CWE market coupling successfully launched,
(2015).

[10] J. van Leeuwen, A scenario-based voltage stability analysis for external constraints in flow-based
capacity calculation, (Master thesis).

[11] D. Van Hertem and M. Ghandhari, Multi-terminal VSC HVDC for the European supergrid: Obstacles,
Renewable and sustainable energy reviews 14, 3156 (2010).

[12] M. O. Dessouky, The environmental impact of large scale solar energy projects on the MENA deserts:
Best practices for the DESERTEC initiative, in EUROCON, 2013 IEEE (IEEE, 2013) pp. 784–788.

[13] J. M. Birkebæk, A. Jäderström, R. Paprocki, and O. Ziemann, Using HVDC Links to cope with
Congestions in Neighbouring AC Networks, PSCC, Stockholm (2011).

[14] P. L. Francos, S. S. Verdugo, H. F. Álvarez, S. Guyomarch, and J. Loncle, INELFE—Europe’s first

integrated onshore HVDC interconnection, in Power and Energy Society General Meeting, 2012 IEEE
(IEEE, 2012) pp. 1–8.

[15] U. Project, FP7, Deliverable 1.3: Final Report, Tech. Rep. (2016).

59
[16] i. Project, Deliverable 9.2: Project Final Report, (2016).

[17] P. Kundur, J. Paserba, V. Ajjarapu, G. Andersson, A. Bose, C. Canizares, N. Hatziargyriou, D. Hill,

A. Stankovic, C. Taylor, et al., Definition and classification of power system stability IEEE/CIGRE joint
task force on stability terms and definitions, IEEE transactions on Power Systems 19, 1387 (2004).

[18] Z. Liu, L. van der Sluis, W. Winter, H. Paeschke, R. Becker, C. Weber, J. Eickmann, C. Schroeders,
F. Oldewurtel, L. Roald, et al., Challenges, experiences and possible solutions in transmission system
operation with large wind integration, in 11th International Workshop on Large-Scale Integration of
Wind Power into Power Systems (11th International Workshop on Large-Scale Integration of Wind
Power into Power Systems, 2012).

[19] Entso-e, P5 – Policy 5: Emergency Operations, (2015).

[20] J. McCalley, S. Asgarpoor, L. Bertling, R. Billinion, H. Chao, J. Chen, J. Endrenyi, R. Fletcher,

A. Ford, C. Grigg, et al., Probabilistic security assessment for power system operations, in Power
Engineering Society General Meeting, 2004. IEEE (IEEE, 2004) pp. 212–220.

[21] J. D. McCalley, V. Vittal, and N. Abi-Samra, An overview of risk based security assessment, in Power
Engineering Society Summer Meeting, 1999. IEEE, Vol. 1 (IEEE, 1999) pp. 173–178.

[22] Umbrella, Deliverable D 4.1 - Risk-based assessment concepts for system security – State-of-the-art
review and concept extensions, (2012).

[23] K. Habur and D. O’Leary, FACTS-flexible alternating current transmission systems: for cost effective
and reliable transmission of electrical energy, Siemens-World Bank document–Final Draft Report,
Erlangen (2004).

[24] K. Meah and S. Ula, Comparative evaluation of HVDC and HVAC transmission systems, in Power
Engineering Society General Meeting, 2007. IEEE (IEEE, 2007) pp. 1–5.

[25] C. J. Wallnerström, Y. Huang, and L. Söder, Impact from dynamic line rating on wind power
integration, IEEE Transactions on Smart Grid 6, 343 (2015).

[26] A. D. Little, Untapped multi-billion market for grid companies, aggregators, utilities and industrials?
(2016).

[27] J. J. Grainger and W. D. Stevenson, Power system analysis (McGraw-Hill, 1994).

[28] R. D. Zimmerman, C. E. Murillo-Sánchez, and R. J. Thomas, MATPOWER: Steady-state operations,

planning, and analysis tools for power systems research and education, IEEE Transactions on power
systems 26, 12 (2011).

[29] F. Capitanescu, J. M. Ramos, P. Panciatici, D. Kirschen, A. M. Marcolini, L. Platbrood, and L. We-

henkel, State-of-the-art, challenges, and future trends in security constrained optimal power flow,
Electric Power Systems Research 81, 1731 (2011).

[30] U. E. F. project, Deliverable D3.1: Report on deterministic algorithms for EOPF, (2013).

[31] L. Castaing, M.-S. Debry, G. Bareux, and O. Beck, Optimal operation of HVDC links embedded in
an AC network, in PowerTech (POWERTECH), 2013 IEEE Grenoble (IEEE, 2013) pp. 1–6.

[32] W. Wang and M. Barnes, Power flow algorithms for multi-terminal VSC-HVDC with droop control,
IEEE Transactions on Power Systems 29, 1721 (2014).

60
[33] J. Cao, W. Du, H. F. Wang, and S. Bu, Minimization of transmission loss in meshed AC/DC grids
with VSC-MTDC networks, IEEE Transactions on Power Systems 28, 3047 (2013).

[34] E. Iggland, R. Wiget, S. Chatzivasileiadis, and G. Anderson, Multi-area DC-OPF for HVAC and
HVDC grids, IEEE Transactions on Power Systems 30, 2450 (2015).

[35] M. de Jong, G. Papaefthymiou, D. Lahaye, C. Vuik, and L. van der Sluis, Impact of correlated
infeeds on risk-based power system security assessment, in Power Systems Computation Conference
(Wroclaw, Poland, 2014).

[36] M. de Jong, G. Papaefthymiou, and P. Palensky, A framework for incorporation of infeed uncertainty
in power system risk-based security assessment, IEEE Transactions on Power Systems (2017).

[37] G. Papaefthymiou, J. Verboomen, and L. van der Sluis, Estimation of power system variability due
to wind power (IEEE, 2007).

[38] G. Papaefthymiou, P. Schavemaker, L. Van der Sluis, W. Kling, D. Kurowicka, and R. Cooke,
Integration of stochastic generation in power systems, International Journal of Electrical Power &
Energy Systems 28, 655 (2006).

[39] M. Sklar, Fonctions de répartition à n dimensions et leurs marges (Université Paris 8, 1959).

[40] M. Ni, J. D. McCalley, V. Vittal, and T. Tayyib, Online risk-based security assessment, IEEE Transac-
tions on Power Systems 18, 258 (2003).

[41] M. A. Rios, D. S. Kirschen, D. Jayaweera, D. P. Nedic, and R. N. Allan, Value of security: modeling
time-dependent phenomena and weather conditions, IEEE Transactions on Power Systems 17, 543
(2002).

[42] G. Papaefthymiou and D. Kurowicka, Using copulas for modeling stochastic dependence in power
system uncertainty analysis, IEEE Transactions on Power Systems 24, 40 (2009).

[43] G. Papaefthymiou and P. Pinson, Modeling of spatial dependence in wind power forecast uncertainty,
in Probabilistic Methods Applied to Power Systems, 2008. PMAPS’08. Proceedings of the 10th Inter-
national Conference on (IEEE, 2008) pp. 1–9.

[44] F. Gonzalez-Longatt, C. Carmona-Delgado, J. Riquelme, M. Burgos, and J. L. Rueda, Risk-based

DC security assessment for future DC-independent system operator, in Energy Economics and Environ-
ment (ICEEE), 2015 International Conference on (IEEE, 2015) pp. 1–8.

[45] X. Li, X. Zhang, L. Wu, P. Lu, and S. Zhang, Transmission line overload risk assessment for power
systems with wind and load-power generation correlation, IEEE Transactions on Smart Grid 6, 1233
(2015).

[46] P. Pourbeik, B. Chakrabarti, T. George, J. Haddow, H. Illian, R. Nighot, et al., Review of the current
status of tools and techniques for risk-based and probabilistic planning in power systems, CIGRE
(2010).

[47] R. H. Renner and D. Van Hertem, Ancillary services in electric power systems with HVDC grids, IET
Generation, Transmission & Distribution 9, 1179 (2015).

[48] J. Beerten, D. Van Hertem, and R. Belmans, VSC MTDC systems with a distributed DC voltage
control-A power flow approach, in PowerTech, 2011 IEEE Trondheim (IEEE, 2011) pp. 1–6.

61
[49] J. Beerten and R. Belmans, Development of an open source power flow software for high voltage direct
current grids and hybrid AC/DC systems: MATACDC, IET Generation, Transmission & Distribution
9, 966 (2015).

[50] L. Roald, S. Misra, T. Krause, and G. Andersson, Corrective Control to Handle Forecast Uncertainty:
A Chance Constrained Optimal Power Flow, IEEE Transactions on Power Systems (2016).

[51] S. Wang, J. Zhu, L. Trinh, and J. Pan, Economic assessment of HVDC project in deregulated en-
ergy markets, (in Proc. 3rd Int. Conf. Electric Utility Deregulation and Restructuring and Power
Technologies, Nanjing, China, Apr. 2008).

[52] S. De Boeck and D. Van Hertem, Coordination of multiple HVDC links in power systems during alert
and emergency situations, in PowerTech (POWERTECH), 2013 IEEE Grenoble (June 2013) pp. 1–6.

[53] Entso-e, HVDC Link Implementation Guide, (2016).

[54] Entso-e, System Operation Agreement, (2016).

[55] Entso-e, ENTSO-E Draft Network Code on High Voltage Direct Current Connections and DC-connected
Power Park Modules, , 1 (2014).

[56] A.-K. Marten, F. Sass, T. Krause, and D. Westermann, Fast local converter set point adaption after
AC grid disturbances based on a priori optimization, Cigré International Symposium Across Borders
– HVDC Systems and Markets Integration (2015).

[57] Entso-e, Principles for Determining the Transfer Capacities in the Nordic Power Market, (2016).

[58] A. Pizano-Martinez, C. R. Fuerte-Esquivel, H. Ambriz-Perez, and E. Acha, Modeling of VSC-based

HVDC systems for a Newton-Raphson OPF algorithm, IEEE Transactions on Power Systems 22, 1794
(2007).

[59] P. Haugland, It’s time to connect: Technical description of HVDC Light® technology, (2006).

[60] M. C. Imhof, Voltage Source Converter Based HVDC–Modelling and Coordinated Control to Enhance
Power System Stability, Ph.D. thesis (2015).

[61] S. S. Chatzivasileiadis, Power system planning and operation methods integrating the controllability
of HVDC, Ph.D. thesis, ETH ZURICH (2013).

[62] S. Cole, Steady-state and dynamic modelling of VSC HVDC systems for power system simulation,
Katholieke Universiteit Leuven, Leuven (2010).

[63] R. D. Zimmerman, C. E. Murillo-Sánchez, and R. J. Thomas, MATPOWER: Steady-state operations,

planning, and analysis tools for power systems research and education, IEEE Transactions on power
systems 26, 12 (2011).

[64] W. Feng, A. Le Tuan, L. B. Tjernberg, A. Mannikoff, and A. Bergman, A new approach for benefit
evaluation of multiterminal VSC–HVDC using a proposed mixed AC/DC optimal power flow, IEEE
Transactions on Power Delivery 29, 432 (2014).

62