FULLTEXT01

Modeling and Control of VSC-HVDC Links Connected to Weak AC Systems
Lidong Zhang
ROYAL I NSTITUTE OF T ECHNOLOGY S CHOOL OF E LECTRICAL E NGINEERING E LECTRICAL M ACHINES AND P OWER E LECTRONICS Stockholm 2010
Submitted to the School of Electrical Engineering in partial fulllment of the requirements for the degree of Doctor of Philosophy.
Stockholm 2010
TRITAEE 2010:022 ISSN 1653-5146 ISRN KTH-EE10/22SE ISBN 978-91-7415-640-9
AT X. This document was prepared using L E
To Yibin, Karin, Vivianne and my parents and my sister Lixia
iv
Abstract
For high-voltage direct-current (HVDC) transmission, the strength of the ac system is important for normal operation. An ac system can be considered as weak either because its impedance is high or its inertia is low. A typical high-impedance system is when an HVDC link is terminated at a weak point of a large ac system where the short-circuit capacity of the ac system is low. Low-inertia systems are considered to have limited number of rotating machines, or no machines at all. Examples of such applications can be found when an HVDC link is powering an island system, or if it is connected to a windfarm. One of the advantages of applying a voltage-source converter (VSC) based HVDC system is its potential to be connected to very weak ac systems where the conventional linecommutated converter (LCC) based HVDC system has difculties. In this thesis, the modeling and control issues for VSC-HVDC links connected to weak ac systems are investigated. In order to fully utilize the potential of the VSCHVDC system for weak-ac-system connections, a novel control method, i.e., powersynchronization control, is proposed. By using power-synchronization control, the VSC resembles the dynamic behavior of a synchronous machine. Several additional functions, such as high-pass current control, current limitation, etc. are proposed to deal with various practical issues during operation. For modeling of ac/dc systems, the Jacobian transfer matrix is proposed as a unied modeling approach. With the ac Jacobian transfer matrix concept, a synchronous ac system is viewed upon as one multivariable feedback system. In the thesis, it is shown that the transmission zeros and poles of the Jacobian transfer matrix are closely related to several power-system stability phenomena. The similar modeling concept is extended to model a dc system with multiple VSCs. It is mathematically proven that the dc system is an inherently unstable process, which requires feedback controllers to be stabilized. For VSC-HVDC links using power-synchronization control, the short-circuit ratio (SCR) of the ac system is no longer a limiting factor, but rather the load angles. The righthalf plane (RHP) transmission zero of the ac Jacobian transfer matrix moves closer to the origin with larger load angles, which imposes a fundamental limitation on the achievable bandwidth of the VSC. As an example, it is shown that a VSC-HVDC link using powersynchronization control enables a power transmission of 0.86 p.u. from a system with an SCR of 1.2 to a system with an SCR of 1.0. For low-inertia system connections, simulation studies show that power-synchronization control is exible for various operation modes v
related to island operation and handles the mode shifts seamlessly. Keywords: Control, modeling, multivariable feedback control, HVDC, power systems, stability, subsynchronous torsional interaction, voltage-source converter, weak ac systems.
vi
Acknowledgements
First of all, my deepest gratitude goes to my supervisors, Prof. Hans-Peter Nee and Prof. Lennart Harnefors. It is an honor and a pleasure for me to have Prof. Hans-Peter Nee as my supervisor. His patience and support helped me to go through the hardest moments of the research work. It is also a privilege for me to be a student of Prof. Lennart Harnefors. I am grateful for his generosity to share with me his deep understanding on scientic work. Without his guidance, this project cannot reach the same level as it is today. This work has been carried out within Elektra Project 30630 and has been funded by Energimyndigheten, ELFORSK, ABB Power Systems, ABB Corporate Research, Banverket. The nancial funding is greatly acknowledged. My acknowledgements also go to the members of the steering group: Gunnar Asplund (ABB Power Systems), Pablo Rey (ABB Power Systems), Hongbo Jiang (Banverket), Torbj orn Thiringer (Chalmers University of Technology). During the last two and half years, I had many inspiring discussions with the steering group members. Their fruitful comments and inputs have greatly improved the quality of the research. Especially, I would like to thank Gunnar Asplund, who was the chairman of the group before his retirement from ABB Power Systems. Gunnar Asplund initiated the project and gave valuable suggestions at the beginning of the project. I would like also to thank my supervisor, Prof. Math Bollen, during my Licentiate study at Chalmers. Prof. Math Bollen brought me into the scientic world. I received endless support from him during my study at Chalmers and after graduation. To my colleagues at ABB, I am grateful for all the supports I have received during this period. In particular, I would like to thank Ying-Jiang H afner, Magnus Ohrstr om, Cuiqing Du, and Rolf Ottersten for interesting discussions as well as many helps with thesis writing. Ying-Jiang H afner carefully reviewed the manuscript of the thesis and gave important suggestions. I would like to give a special thank to Pablo Rey, my group manager at ABB, for allowing me to be absent from the group for the Ph.D study. At KTH, I would like to thank all the colleagues in the Electrical Machines and Power Electronics department. In particular, I would like to thank Prof. Chandur Sadarangani for reviewing the manuscript of the thesis. I am also grateful to Hailian Xie for her help with thesis writing, to Peter L onn for his computer support, to Eva Pettersson and Brigitt H ogberg for their help with the administrative work. vii
Many thanks to my parents and my parents-in-law for their love and support. My mother-in-law, Prof. Renmu He, is a renowned professor in power systems in China. I received many helps from her in my professional life as well as my family life for the past years. Her valuable suggestions during her stay in Sweden shed light on my research and inuenced the content of this thesis. I would like also to thank my sister and nephew for their love and encouragement for all the time. Last but not least, I would like to thank my beloved wife and daughters. Yibin, thank you so much for your endless love, support and understanding. Thank you, Karin and Vivianne, for the joys you have brought to my life. Lidong Zhang Stockholm, Sweden April 2010
viii
Contents
Abstract Acknowledgements Contents 1 Introduction 1.1 Background . . . . . . . . . . . . . . . . 1.2 Project objectives and outline of the thesis 1.3 Scientic contributions of the thesis . . . 1.4 List of publications . . . . . . . . . . . . v vii ix 1 1 3 4 6 9 9 11 14 21 21 25 30 31 32 35 41 43 47 47 47 50 52 ix
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
2 High-Voltage Direct-Current Transmission 2.1 DC versus AC transmission . . . . . . . . . . . . . . . . . . . . . . . . 2.2 HVDC transmission using line-commutated current-source converters . . 2.3 HVDC transmission using forced-commutated voltage-source converters . 3 Control Methods for VSC-HVDC Systems 3.1 Power-angle control . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Vector current control . . . . . . . . . . . . . . . . . . . . . . . 3.3 Power-synchronization control . . . . . . . . . . . . . . . . . . 3.3.1 Power-synchronization mechanism in ac systems . . . . 3.3.2 Power-synchronization control of grid-connected VSCs . 3.3.3 Bumpless-transfer and anti-windup schemes . . . . . . . 3.3.4 Negative-sequence current control . . . . . . . . . . . . 3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Dynamic Modeling of AC/DC Systems 4.1 Jacobian transfer matrix for ac-system modeling . . . . . . 4.1.1 Power-system stability and dynamic modeling . . 4.1.2 Feedback-control view of power systems . . . . . 4.2 Grid-connected VSCs using power-synchronization control
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
Contents
4.2.1 Impedance-source neglecting the ac capacitor at the lter bus . 4.2.2 Impedance-source including the ac capacitor at the lter bus . 4.2.3 AC-source feeding from a series-compensated ac line . . . . . Grid-connected VSCs using vector current control . . . . . . . . . . . Jacobian transfer matrix for dc-system modeling . . . . . . . . . . . . Summary of the properties of the Jacobian transfer matrix . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 63 70 74 81 87 88 89 89 91 100 102 107 111 117 117 119 122 130 131 131 133 133 134 135 136 147 150 152 155 155 159 162 164 165 173 177 178
4.3 4.4 4.5 4.6
5 Control of VSC-HVDC Links Connected to High-Impedance AC Systems 5.1 General aspects of high-impedance ac systems . . . . . . . . . . . . . . 5.2 Comparison of power-synchronization control and vector current control 5.3 Multivariable feedback designs . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Internal model control . . . . . . . . . . . . . . . . . . . . . . 5.3.2 H control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Performance and robustness comparison . . . . . . . . . . . . . 5.4 Direct-voltage control . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Controller design . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 DC-capacitance requirement . . . . . . . . . . . . . . . . . . . 5.5 Interconnection of two very weak ac systems . . . . . . . . . . . . . . 5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Control of VSC-HVDC Links Connected to Low-Inertia AC Systems 6.1 General aspects of low-inertia ac systems . . . . . . . . . . . . . . 6.2 Controller design . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Frequency droop control . . . . . . . . . . . . . . . . . . . 6.2.2 Alternating-voltage droop control . . . . . . . . . . . . . . 6.3 Dynamic modeling and linear analysis of a typical island system . . 6.3.1 Jacobian transfer matrix . . . . . . . . . . . . . . . . . . . 6.3.2 Integrated linear model . . . . . . . . . . . . . . . . . . . . 6.3.3 Linear analysis . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Simulation studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Jacobian transfer matrix for other input devices . . . . . . . . . . . 6.5.1 Synchronous generator . . . . . . . . . . . . . . . . . . . . 6.5.2 Induction motor . . . . . . . . . . . . . . . . . . . . . . . . 6.6 Subsynchronous characteristics . . . . . . . . . . . . . . . . . . . . 6.6.1 Frequency-scanning method . . . . . . . . . . . . . . . . . 6.6.2 Large ac-system connection . . . . . . . . . . . . . . . . . 6.6.3 Island operation . . . . . . . . . . . . . . . . . . . . . . . . 6.6.4 Summary of the subsynchronous characteristics . . . . . . . 6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
. . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
Contents
7 Conclusions and Future Work 181 7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 References A Fundamentals of the Phasor and the Space-Vector Theory A.1 Fundamentals of the phasor theory . . . . . . . . . . . A.2 Fundamentals of the space-vector theory . . . . . . . . A.3 Implementation of and dq transformations . . . . . A.3.1 abc- transformation . . . . . . . . . . . . . A.3.2 -dq transformation . . . . . . . . . . . . . . 185 197 197 198 200 200 201 203 203 203 205 206 207 207 208 208
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
B Jacobian Transfer Matrix B.1 Derivation of the transfer functions in Table 4.1 . . . . . B.1.1 Transfer function JP (s) . . . . . . . . . . . . . B.1.2 Transfer function JQ (s) . . . . . . . . . . . . . B.1.3 Transfer function JUf (s) . . . . . . . . . . . . B.1.4 Transfer function JPV (s) . . . . . . . . . . . . . B.1.5 Transfer function JQV (s) . . . . . . . . . . . . . B.1.6 Transfer function JUf V (s) . . . . . . . . . . . . B.2 Proof of the instability of the dc Jacobian transfer matrix
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
C Technical Data of the Test System 211 C.1 The VSC-HVDC link . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 C.2 The synchronous generator . . . . . . . . . . . . . . . . . . . . . . . . . 212 C.3 The induction motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 D List of Symbols and Abbreviations 215
xi
Contents
xii
Chapter 1 Introduction
This chapter describes the background of the thesis. The aim and the outline, as well as the major scientic contributions of the thesis are presented. Finally, a list of publications is given.
1.1 Background
In 1954, the rst commercial high-voltage direct-current (HVDC) link between mainland Sweden to Gotland island was commissioned. Since then, the accumulated installed power of HVDC transmission worldwide has increased steadily, and recently a dramatic increase in volume has been initiated. So far, most of the HVDC systems installed worldwide are line-commutated converter (LCC) systems using thyristor valves. However, with gradually reduced losses and costs, the recently developed voltage-source converter (VSC) technology has shown to be more advantageous in many aspects [14]. The conventional line-commutated HVDC technology has an inherent weakness, i.e., the commutation of the converter valves is dependent on the stiffness of the alternating voltage. The converter cannot work properly if the connected ac system is weak. Substantial research has been performed in this eld [59]. The most outstanding contribution on this subject is [5], which recommends to use short-circuit ratio (SCR) as a description of the strength of the ac system relative to the power rating of the HVDC link. Both [8] and [9] conclude that, for ac systems with an SCR lower than 1.5, synchronous condensers have to be installed to increase the short-circuit capacity of the ac system. However, synchronous condensers can substantially increase the investment and maintenance costs of an HVDC project. In contrast to the conventional LCC-HVDC system, the VSC-HVDC system is based on self-commutated pulse-width modulation (PWM) technology, i.e., a VSC can produce its own voltage waveform independent of the ac system. Thus, a VSC-HVDC system has the potential to be connected to very weak ac systems. However, with the traditional vector current control the potential of the VSC is not fully utilized [1012], 1
Chapter 1. Introduction
e.g., Ref. [11] shows that the maximum power that a VSC-HVDC link using vector current control can transmit to the ac system with SCR = 1.0 is 0.4 p.u. Ref. [12] shows that the inner-current controller of vector current control may interact with low-frequency resonances that are typically present in weak ac systems. In addition, the phase-locked loop (PLL) dynamics of vector current control might also have a negative impact on the performance of VSC-HVDC links for weak ac-system connections [10, 11, 13]. The poor performance of vector current control for weak-ac-system connections has become an obstacle for VSC-HVDC transmission to be applied in more challenging ac-system conditions. The application of high power-electronic devices, such as HVDC systems and FACTS devices also imposes new challenges for power-system stability analysis and dynamic modeling. For classical power-system stability analysis, the phasor theory is the major mathematical tool. With the phasor approach, the electromagnetic transients of the ac network are neglected. This is a practical solution for conventional power systems where the electromagnetic transients have negligible effects on the stability issue of concern. However, for high power-electronic devices, such a simplication is not acceptable. The dynamic frequency range of high power-electronic devices is much higher than that of the conventional power-system components. In this frequency, the phasor theory cannot properly reect the dynamic interaction between the ac system and the powerelectronic devices on the one hand, and between different power-electronic devices on the other hand. For example, it has been shown by [14] that the conclusions drawn by the phasor-based quasi-static analysis might not always agree with the results obtained by time simulations with electromagnetic-transient programs. The space-vector theory is based on instantaneous values, and therefore it is able to represent the electromagnetic transients of the ac network [15]. Traditionally, the spacevector theory is mainly applied for analyzing electrical machines and control of powerelectronic devices [1517]. Several methodologies for dynamic modeling of three-phase systems based on the space-vector theory have been been proposed. In [18], the complex transfer functions are applied for analyzing three-phase ac machines. In [19], a threephase linear current controller is analyzed in the frequency domain based on the spacevector approach. In [20], the space-vector theory is applied for modeling of three-phase dynamic systems using the transfer matrix concept. For subsynchronous torsional interaction (SSTI) analysis, the ac network is normally required to be modeled by the spacevector approach to take into account the electromagnetic transients [21, 22]. In recent years, the space-vector theory has also been applied to study the dynamic interactions between high power-electronic devices. In [12], the dynamic interaction between an LCCHVDC link and a VSC-STATCOM in the frequency domain is analyzed based on the space-vector theory. While the space-vector theory has been applied successfully for analyzing highfrequency stability phenomena in power systems, the theoretical work to connect the 2
1.2. Project objectives and outline of the thesis

high-frequency stability to the classical power-system stability dened by the phasor approach is missing in the literature. Such a connection is, however, necessary for HVDC systems and FACTS applications. On the one hand, the dynamic frequency range of such devices is high. On the other hand, the ratings of those devices are often high enough to have a signicant impact on most of the classical power-system stability phenomena, such as angle stability and voltage stability. In the foreseeable future, the number of such devices in power systems is expected to increase considerably. Thus, there is a need for a unied modeling approach to address both the high-frequency and low-frequency stability phenomena.
1.2 Project objectives and outline of the thesis

The objectives of the project are: 1. Develop a new control method for VSC-HVDC links connected to weak ac systems. 2. Develop a unied approach for dynamic modeling of ac/dc systems. 3. Investigate various modeling and control issues for VSC-HVDC links connected to high-impedance ac systems. 4. Investigate various modeling and control issues for VSC-HVDC links connected to low-inertia ac systems. The project is conducted by both theoretical analysis and time simulations. The outline of the thesis is: Chapter 2 A short introduction of various technologies for HVDC transmission is given. Chapter 3 Two existing control methods for VSC-HVDC systems, i.e., power-angle control and vector current control are described. A novel control method, i.e., powersynchronization control, is proposed to solve the problem for VSC-HVDC links connected to weak ac systems. Chapter 4 A unied dynamic modeling approach, i.e., the Jacobian transfer matrix, is proposed for modeling of ac/dc systems. Grid-connected VSCs using powersynchronization control and vector current control are modeled by the proposed concept. Chapter 5 The control issues for VSC-HVDC links connected to high-impedance ac systems are investigated. The dynamic performance of a VSC-HVDC link using 3
power-synchronization control and vector current control are compared. Two multivariable feedback-control designs, i.e., internal model control (IMC) and H control are investigated. A direct-voltage controller is proposed. A control structure for interconnection of two very weak ac systems is proposed. Chapter 6 Power-synchronization control is applied to VSC-HVDC links connected to low-inertia ac systems. A frequency droop controller and a voltage droop controller are proposed. A linear model of a typical island system is developed for tuning the control parameters of the VSC-HVDC link. The subsynchronous characteristics of a VSC-HVDC converter are analyzed for both the large ac-system connection and island operation. Chapter 7 Summarizes the thesis and provides suggestions for future work.
1.3 Scientic contributions of the thesis

The main contributions of the thesis are: A novel control method for grid-connected VSCs, i.e., power-synchronization control, is proposed. The VSC using power-synchronization control basically resembles the dynamic behavior of a synchronous machine. A group of additional control functions, such as high-pass current control, current limitation function, antiwindup schemes, etc. are proposed to deal with various practical issues during operation. A novel modeling concept, i.e., the Jacobian transfer matrix, is proposed as a unied dynamic modeling technique for ac/dc systems. With the proposed concept, a synchronous power system is viewed upon as a multivariable feedback control system. The proposed concept is intended to be a unied framework for analyzing both the low-frequency and high-frequency stability phenomena in power systems. The theoretical connections between the stability dened by the Jacobian transfer matrix concept and the classical power-system stability dened by the phasor approach are analyzed. It is discovered that the transmission zeros of the Jacobian transfer matrix have a close relationship with angle and voltage stability in power systems. A similar modeling concept, i.e., the dc Jacobian matrix, is proposed for modeling of dc systems. By using a -link dc model, it is mathematically proven that the dc system (constructed by VSCs) is an inherently unstable process, where the dc resistance gives a destabilizing effect. 4
1.3. Scientic contributions of the thesis

Grid-connected VSCs using power-synchronization control and vector current control are modeled by the Jacobian transfer matrix concept. The transfer functions are validated with frequency-scanning results from PSCAD/EMTDC. The Jacobian transfer matrix concept is also applied for modeling of two conventional power components, i.e., the synchronous generator and the induction motor. It is discovered that the transmission zeros of the Jacobian transfer matrix are useful for interpreting some classical concepts, such as the synchronizing torque for the synchronous generator and the pull-out slip for the induction motor, from a feedback control point of view. The transfer functions of the Jacobian transfer matrices are also validated with frequency-scanning results from PSCAD/EMTDC. The performance of power-synchronization control and vector current control are compared for VSC-HVDC links connected to weak ac systems, where it is concluded that power-synchronization control is more suitable for weak-ac-system connections. Two multivariable feedback-control design methods, i.e., IMC and H control are investigated for VSC-HVDC links connected to high-impedance ac systems. The performance and robustness of various control designs are compared and discussed. A two-degree-of-freedom direct-voltage controller for VSC-HVDC system is proposed where a prelter is applied to remove the overshoot of the direct voltage. A notch lter is proposed to reduce the dc-resonance peak. The requirement of dc capacitance for VSC-HVDC links connected to weak ac systems is derived. A control structure for VSC-HVDC links interconnecting two very weak ac systems is proposed. The linear model is validated with time simulations from PSCAD/EMTDC for each major design step. A frequency droop controller and an alternating-voltage droop controller are proposed for VSC-HVDC links connected to low-inertia systems. A complete linear model is developed for a typical island system which includes a synchronous generator, an induction motor, a VSC-HVDC link and some RLC loads. The root-locus technique is applied to tune the control parameters of the VSC-HVDC link. Simulation studies are performed to demonstrate the exibility of power-synchronization control for various operation modes related to island operation. The subsynchronous characteristics of a VSC-HVDC converter using power-synchronization control are analyzed using the frequency-scanning method. 5
1.4 List of publications

- The doctoral thesis has resulted in the following publications: I L. Zhang and H.-P. Nee, Multivariable feedback design of VSC-HVDC connected to weak ac systems, in PowerTech 2009, Bucharest, Romania, 2009. II L. Zhang, L. Harnefors and H.-P. Nee, Power-synchronization control of gridconnected voltage-source converters, IEEE Trans. Power Systems, vol. 25, no. 2, pp. 809-820, May 2010. III L. Zhang, L. Harnefors and H.-P. Nee, Modeling and control of VSC-HVDC links connected to island systems accepted for publication at IEEE Power and Energy Society General Meeting, 2010, Minneapolis, USA. IV L. Zhang, L. Harnefors and H.-P. Nee, Interconnection of two very weak ac systems by VSC-HVDC links using power-synchronization control, accepted for publication in IEEE Trans. Power Systems. V L. Zhang, H.-P. Nee and L. Harnefors, Analysis of stability limitations of a VSC-HVDC link using power-synchronization control, submitted to IEEE Trans. Power Systems. - The author has co-authored the following publication during the course of the Ph.D study: VI L. Harnefors, L. Zhang and M. Bongiorno, Frequency-domain passivity-based current controller design, IET Power Electron., vol. 1, no. 4, pp. 455-465, 2008. - During the course of the licentiate study, the author has authored and co-authored the following publications: VII L. Zhang and M. H. J. Bollen, A method for characterizing unbalanced voltage dips with symmetrical components, IEEE Power Engineering Letter, pp. 50-52, July 1998. VIII L. Zhang and M. H. J. Bollen, A method for characterization of three-phase unbalanced dips from recorded voltage waveshapes, in International Telecommunication Energy Conference, Copenhagen, Danmark, 1999. IX L. Zhang and M. H. J. Bollen, Characteristics of voltage dips in power systems, IEEE Trans. Power Delivery, vol. 15, no. 2, pp. 827-832, April 2000. X L. Zhang, Three-phase unbalance of voltage dips, Licentiate thesis, Technical Report no. 322L, ISBN 91-7197-855-0, Chalmers University of Technology, G oteborg, Sweden, 1999. 6
1.4. List of publications

XI M. H. J. Bollen and L. Zhang, Analysis of voltage tolerance of ac adjustable-speed drives for three-phase balanced and unbalanced sags, IEEE Trans. Ind. Applicat., vol. 36, no. 3, pp. 904-910, May/June 2000. XII M. H. J. Bollen, J. Svensson, and L. Zhang, Testing of grid-connected power converters for the effects of short circuits in the grid, in European Power Electronics Conference, Lausanne, Switzerland, 1999. - In addition, during the working time in ABB, the author has authored and co-authored the following publications that are relevant to the subjects of the doctoral and licentiate studies: XIII L. Zhang and L. D ofn as, A novel method to mitigate commutation failures in HVDC systems, in International Conference on Power System Technology, Kunming, China, 2002. XIV L. Zhang, L. Harnefors and P. Rey, Power system reliability and transfer capability improvement by VSC-HVDC (HVDC Light), in Cigre Regional meeting, Tallin, Estonia, 2007. XV M. H. J. Bollen and L. Zhang, Different methods for classication of three-phase unbalanced voltage dips due to faults, Electric Power Systems Research, vol. 66, no. 1, pp. 59-69, July 2003.
Chapter 2 High-Voltage Direct-Current Transmission

This chapter presents general aspects of HVDC transmission. Two major HVDC technologies, i.e., HVDC transmission using line-commutated current-source converters and HVDC transmission using forced-commutated voltage-source converters are described.
2.1 DC versus AC transmission

The history of electric power systems began with direct-current (dc) transmission. In 1882, Thomas Edison built the rst power system with dc transmission with a low voltage level. However, dc transmission was quickly replaced by three-phase alternating current (ac) transmission because of several advantages of the latter. The most prominent advantage of ac transmission is that power can be transformed to different voltage levels. By using transformers, long-distance power transmission becomes possible. In addition, circuit breakers for alternating current can take advantage of the natural current zeros that occur twice per cycle, and ac motors are cheaper and more robust than dc motors. In spite of the principal use of ac transmission in power systems, the interests on dc transmission still remain [23]. In 1954, the rst commercial HVDC link between mainland Sweden to Gotland island was commissioned. Since then, the accumulated installed power of HVDC transmission systems worldwide has increased steadily, and recently a dramatic increase in volume has been initiated. Given the extra costs and losses related to the converter stations, HVDC transmission is justied by some particular conditions where the dc technology is the most feasible or may be the only solution: Power transmission via cables. Due to their physical structures, cables have much higher capacitance than overhead lines. The capacitive current in cables created by the alternating voltage makes ac power transmission over long distance impossible. Even for a moderate length (50km), the losses created by the capacitive current can 9
Chapter 2. High-Voltage Direct-Current Transmission

be so high that reactive compensating equipment has to be installed in the middle of the cable [24]. However, installation of reactive compensating equipment is expensive and not always practical, e.g., with submarine cable transmission under sea. On the other hand, if the power is transmitted by direct currents, there will be no losses related to capacitive currents. Therefore, for long-distance submarine cable transmission, HVDC transmission is the only feasible technical solution. Bulk-power transmission over long-distance. Interestingly, given the fact that ac won the battle of currents due to its possibility to transmit power over long distance [25], HVDC transmission wins the battle back after a century. To transmit the same amount of power, dc transmission needs fewer power lines than ac transmission. Accordingly, the costs and losses of the converter stations get balanced by savings on the overhead lines where the break-even distance is around 400 km to 700 km depending on the land conditions and project specications [26]. Besides, dc transmission does not have the stability limitation related to ac transmission over long distance. Unsynchronized ac-system connection. AC transmission is only possible if the two interconnected ac systems have the same nominal frequency and operate synchronously, but dc transmission does not have such requirements. Many back-toback HVDC links have been built for such purposes. Besides the above essential arguments, there are additional benets by having embedded HVDC links in ac systems: Power-system stability improvement. One of the major features of the HVDC technology is its capability to manipulate large amount of power in a very short time, which can often be utilized to improve the stability of the ac system. One example is the improvement of transient stability by running up or running back the dc power for emergency power supports [4, 27]. Another example is that HVDC system can be used to damp low-frequency oscillations in ac systems by having an auxiliary damping controller [28]. Firewall function. Large interconnected ac systems have many well-known advantages, e.g., the possibility to use larger and more economical power plants, reduction of reserve capacity in the systems, utilization of the most efcient energy resources, as well as achieving an increase in system reliability [25,29]. However, larger interconnected ac systems also increase the system complexity from the operation point of view. One of the consequence of such complexity is the large blackouts in America and Europe [30]. In this aspect, HVDC links have the rewall function in preventing cascaded ac-system outages spreading from one system to another [31]. 10
2.2. HVDC transmission using line-commutated current-source converters

idc L
V1 Ua Ub Uc
V3
V5
+ + +
L L L
Ia Ib Ic
udc
V4
V6
V2
Fig. 2.1 Graetz bridge for LCC-HVDC system.
2.2 HVDC transmission using line-commutated currentsource converters

The converter technology used for HVDC transmission in the early days was based on mercury valves. The major problem with mercury-arc technology was ark-back fault which destroyed the rectifying property of the converter valve and consequently triggered other problems [23]. In the late 1960s, the thyristor valve technology was developed that overcame the problems of mercury-arc technology. Converters based on either mercury valves or thyristor valves are called line-commutated converters (LCCs), or current-source converters (CSCs). The basic module of an LCC is the three-phase full-wave bridge circuit shown in Fig. 2.1. This topology is known as the Graetz bridge. Although there are several alternative congurations possible, the Graetz bridge has been universally used for LCC-HVDC converters as it provides better utilization of the converter transformer and a lower voltage across the valve when not conducting [32]. The Graetz bridge can be used for transmitting power in two directions, i.e., the rectier mode and the inverter mode. This is achieved by applying different ring angles on the valves. If the ring angle is lower than 90 , the direct current is owing from the positive terminal of the dc circuit, thus the power is following from the ac side to the dc side; If the ring angle is higher than 90 , the direct voltage changes polarity, thus the direct current is owing from the negative terminal of the dc circuit. The power is then owing from the dc side to the ac side. An HVDC link is essentially constructed by two Graetz bridges, which are interconnected on the dc sides. The interconnection could be an overhead line, a cable, or a back-to-back connection. The application of LCC-HVDC technology has been very successful and the installations of LCC-HVDC links are expected to grow at least in the near future. However, the LCC technology suffers from several inherent weaknesses. 11

1 0
Uab Uac Ubc Uba Uca Ucb Uab Uac Ubc Uba Uca Ucb
udc (p.u.)
-1 0 2 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Iv (p.u.)
0
V4
V5 V6
V1 V2
V3 V4
V1 V6 V2
V3
-2 0 1 0
0.005
0.01
0.015
0.02
Uab
0.025
Uac
0.03
0.035
0.04
Uab Uac
Uv (p.u.)
-1 0 0.005 0.01 0.015 0.02 0.025 time (sec) 0.03 0.035 0.04
Fig. 2.2 Commutation failures of an LCC-HVDC inverter. Upper plot: direct voltage. Middle plot: valve currents. Lower plot: valve voltage.
One problem is that the LCC always consumes reactive power, either in rectier mode or in inverter mode. Depending on the ring angles, the reactive power consumption of an LCC-HVDC converter station is approximately 50 60% of the active power. The reactive-power consumption requires compensation by connecting large ac lters/capacitors at the converter stations. For a common LCC-HVDC link, the lters/capacitors not only increase the costs, but also occupy large amounts of space of the converter stations. Besides, large lters/capacitors also contribute to the temporary overvoltage (TOV) and low-order harmonic resonance problems of the HVDC link when connects to a weak ac system [5]. Another well-known problem of the LCC-HVDC system is the occurrence of commutation failures at the inverter station typically caused by disturbances in the ac system. Either depressed voltage magnitude or phase-angle shift of the alternating voltage may reduce the extinction volt-time area of the inverter valve [33, 34]. If the extinction angle of the inverter valve is smaller than 5 6 , the previously conducted valve will regain current, which will end up with a commutation failure. Fig. 2.2 shows plots of a typical commutation failure. A disturbance on Uab appears during the commutation between V1 and V3. Because V1 does not get the reverse voltage that is needed to switch off the current, V1 continues to be conducted and the valve current of V3 goes down to zero again. When the next commutation occurs between V2 and V4, V1 and V4 conduct at the same time. From Fig. 2.2 it can be observed that the commutation failure, in fact, creates a short circuit on the dc side, which essentially temporarily stops the power transmission. Commutation failures are common phenomena of LCC-HVDC systems. A single 12
2.2. HVDC transmission using line-commutated current-source converters

commutation failure generally does no harm to either the converter valves or the ac system. However, a number of repeated commutation failures may force the HVDC link to trip [35]. While the above two problems can be mitigated by relatively easy measures, the third problem is more fundamental, which can become a limiting factor for LCC-HVDC applications. For LCCs, the successful commutation of the alternating current from one valve to the next relies on the stiffness of the alternating voltage, i.e., the network strength of the ac system. If the ac system has low short-circuit capacity relative to the power rating of the HVDC link, i.e., low SCR, more problematic interactions between the ac and the dc systems are expected. Besides, the SCR of the ac system also imposes an upper limitation on the HVDC power transmission, which is often described by the wellknown maximum power curve (MPC) [5]. As mentioned before, an LCC-HVDC link normally requires reactive-power compensation by connecting larger ac lters/capacitors at the converter stations. These ac lters/capacitors create additional problems in weak ac systems as described below. One such problem is the aforementioned TOV issue. In case of a sudden change in the active power, or the blocking of converter, the large lters/capacitors at the converter station together with the high inductance of the ac system cause a temporary overvoltage before the protection system disconnects the lters/capacitors. The magnitude of the overvoltage is directly related to the strength of the ac system. Ref. [5] gives the following estimation of the fundamental components of TOV (TOVfc ) regarding the SCR of the ac system: SCR > 3: TOVfc lower than 1.25 p.u. 2 < SCR < 3: TOVfc higher than 1.25 p.u. but lower than 1.4 p.u. SCR < 2: TOVfc higher than 1.4 p.u. TOVs can inuence the design and costs of the dc stations. TOVs can also lead to saturation of the converter transformer or transformers close to the dc station. Another problem with weak-ac-system connections is the low-order harmonic resonance. The high inductance of the ac system and the large lters/capacitors of the HVDC link create a resonance with low frequency fres , which can be estimated approximately by fres f1 2 SCR (2.1)
where f1 is the nominal frequency of the ac system. That is to say, the resonance frequency tends to be lower for weak ac systems. Generally speaking, the lower the resonance frequency, the greater the risk for harmful interaction with the converter control system. An improved topology of the LCC-HVDC system to overcome part of the above mentioned problems is the capacitor-commutated converter (CCC)-HVDC technology, 13

idc
Ua Ub Uc
+ + +
L L L
Ia Ib Ic
udc
Fig. 2.3 Two-level voltage-source converter.
where ac capacitors are inserted in series between the valves and converter transformers [36]. The series-connected capacitors not only supply the reactive power consumed by the valves, it also improve the dynamic performance of the HVDC system. However, the major drawback of the CCC concept is that the series capacitors increase the insulation costs of the valves. Thus, the CCC-HVDC technology has been so far only applied to back-to-back HVDC links, where the voltage level of the valves is much lower.
2.3 HVDC transmission using forced-commutated voltagesource converters

Voltage-source converters are a new converter technology for HVDC transmission [2]. The rst commercial VSC-HVDC (HVDC Light) link with a rating of 50 MW was commissioned in 1999 in Gotland island of Sweden, close to the worlds rst LCC-HVDC link. Voltage-source converters (VSCs) utilize self-commutating switches, e.g., gate turn off thyristors (GTOs) or insulated-gate bipolar transistors (IGBTs), which can be turned on or off freely. This is in contrast to the LCC where the thyristor valve can only be turned off by reversed line voltages. Therefore, a VSC can produce its own sinusoidal voltage waveform using pulse-width modulation (PWM) technology independent of the ac system. Many different topologies have been proposed for VSCs. However, for HVDC applications, they have been so far limited to three major types: two-level converter, threelevel converter, and modular multilevel converter (M2C) [3739]. Fig.2.3 shows a two-level grid-connected VSC. The two-level bridge is the simplest topology that can be used in order to build up a three-phase forced-commutated VSC bridge. The bridge consists of six valves and each valve consists of a switching device 14
2.3. HVDC transmission using forced-commutated voltage-source converters

idc
Ua Ub Uc
+ + +
L L L
Ia Ib Ic
udc
Fig. 2.4 Three-level neutral-point-clamped voltage-source converter.
and an anti-parallel diode. For an HVDC link, two VSCs are interconnected on the dc side. For high-voltage applications, series connection of switching devices is necessary. The operation principle of the two-level bridge is simple. Each phase of the VSC can be connected either to the positive dc terminal, or the negative dc terminal. By adjusting the width of pulses, the reference voltage can be reproduced, as shown in the upper plot of Fig. 2.6. After ltering by phase reactors and shunt lters, this series of voltage pulses resembles the voltage waveform of the reference voltage. The three-level VSC shown in Fig. 2.4 is also called neutral-point-clamped (NPC) converter. The key components that distinguish this topology from the two-level converter are the two clamping diodes in each phase. These two diodes clamp the switch voltage to half of the dc voltage. Thus, each phase of the VSC can switch to three different voltage levels, i.e., the positive dc terminal, the negative dc terminal and the mid-point. Consequently, voltage pulses produced by a three-level VSC match closer to the reference voltage. Therefore, the three-level NPC converter has less harmonic content as shown in the middle plot of Fig. 2.6. Additionally, the three-level NPC converter has lower switching losses. Compared to two-level VSCs, three-level NPC VSCs require more diodes for neutral-point clamping. However, the total number of switching components does not necessarily have to be higher. The reason for this is that, for HVDC applications, a valve consists of many series-connected switches. In the two-level case a valve has to withstand twice as high voltage than in the three-level case. Accordingly, the total number of 15

A V1 B V2 (a)
A B
Positive arm
A A B B ... B
Ua
+
A B A B ... A B
udc
Negative arm (b)
Fig. 2.5 Modular multilevel voltage-source converter. (a) One M2C module. (b) One phase topology.
switches is approximately equal. The NPC concept can be extended to higher number of voltage levels, which can result in further improved harmonic reduction and lower switching losses [40]. However, for high-voltage converter applications, the neutral-clamped diodes complicate the insulation and cooling design of the converter valve. Therefore, NPC concepts with a number of voltage levels higher than three has never been considered for HVDC applications [37]. The recently proposed modular multilevel converter (M2C) concept [39, 4143] attracts signicant interests for high-voltage converter applications. Fig. 2.5 shows the M2C topology for one phase. Compared to the above two topologies, one major feature of the M2C is that no common capacitor is connected at the dc side. Instead, the dc capacitors are distributed into each module, while the converter is built up by cascadeconnected modules. Fig. 2.5(a) shows an M2C module. Each M2C module consists of two valves which can be switched in three different ways: V2 is turned on and V1 is turned off, the capacitor is inserted into the circuit from A to B. The module contributes with voltage to the phase voltage. The capacitor is charged if the current is from A to B, and discharged otherwise. V1 is turned on and V2 is turned off, the capacitor is by passed. The module does not contribute with voltage to the phase voltage. Both V1 and V2 are turned off, the module is blocked. 16

100 0 -100 0 100 0 -100 0 100 0 -100 0 0.01 0.02 0.03 time (sec) 0.04 0.05 0.06 0.01 0.02 0.03 0.04 0.05 0.06 0.01 0.02 0.03 0.04 0.05 0.06
Fig. 2.6 Pulse-width modulation for different converter topologies. Upper plot: two-level converter. Middle plot: three-level converter. Lower plot: M2C with ve modules .
The M2C concept is especially attractive for high-voltage applications, since the converter can be easily scaled up by inserting additional modules in each arm. If considerable amounts of modules are cascaded (approximately 100 modules would be common for HVDC applications), each module theoretically only needs to switch on and off once per period, which greatly reduces the switching losses of the valves. However, preliminary investigation indicates that slightly higher switching frequencies are necessary. The lower plot of Fig. 2.6 shows the voltage waveform produced by a ve-module (ve for each arm) M2C. With only ve modules, the waveform already resembles much better the sinusoidal voltage reference than the other two topologies. With M2C, the harmonic content of the voltage produced by the VSC is so low that additional ltering equipment is almost unnecessary. An additional benet of the M2C is that the control system has an extra freedom in dealing with faults at the dc side. The dc capacitors are not necessarily discharged during faults. Thus, the fault recovery can be faster [39]. Compared to the other two topologies, the major drawback of the M2C topology is that the required switching components are doubled since only one of the valves of each module contributes to the phase voltage when the module is inserted in. In addition, the design and control of the M2C are generally more complex at least than the two-level converter. However, since the switching frequency of the M2C can be kept very low switches with higher blocking voltages may be used, which in turn limits the increase in number of switches. On the other hand, the reduction of switching losses and savings on ltering equipment of the M2C may eventually justify its application for HVDC transmission. 17
UL1 (kV)
UL1 (kV)
UL1 (kV)

Reactive power Over-voltage limitation
1 .0
Converter current limitation

0 .5
0.5
1.0
Active power
Under-voltage limitation
Fig. 2.7 PQ diagram for a typical VSC-HVDC converter.
No matter what converter topology is used, the VSC can always be treated as an ideal voltage source where the control system has the freedom to specify the magnitude, phase, and frequency of the produced sinusoidal voltage waveform. However, for control design and stability analysis, it is important to take into account the limitation of the converter in terms of active and reactive power transfer capability. One such limit is the converter-current limitation, which is imposed by the current carrying capability of the VSC valves. Since both the active power and the reactive power contribute to the current owing through the valves, this limitation is manifested as a circle in a PQ diagram. Accordingly, if the converter is intended to support the ac system with reactive-power supply/consumption, the maximum active power has to be limited to make sure that the valve current is within the limit. Another limitation which determines the reactive-power capability of the VSC is the over/under voltage magnitude of the VSC (modulation index limitation). The overvoltage limitation is imposed by the direct-voltage level of the VSC. The under-voltage limit, however, is limited by the main-circuit design and the active-power transfer capability, which requires a minimum voltage magnitude to transmit the active power. In this respect, the tap-changer of the converter transformer can play an important role to extend the reactive-power limitation of the VSC. This could be an argument to have converter transformers in VSC-HVDC systems. Fig. 2.7 shows the PQ diagram with the above mentioned limitations for a typical VSC-HVDC converter [44]. VSC-HVDC technology overcomes most of the weaknesses of the LCC-HVDC technology. In addition, it supports the ac system with reactive-power supply/consumption. 18

Similar to an LCC-HVDC system, a VSC-HVDC system can quickly run up or run back the active power for ac system emergency-power support, but it can also instantly reverse the active power [4]. Since the direct voltage of a VSC-HVDC system varies in a much smaller range than that of a LCC-HVDC system, extruded cables can be used for VSC-HVDC systems. The extruded cable reduces the cable cost and the construction cost. The latter makes long-distance land-cable transmission possible [2]. Besides the above features, the most essential one is that a VSC-HVDC system has an unlimited connection capability with ac systems, i.e., with properly designed control systems, VSC-HVDC system has the potential to be connected to any kind of ac system with any number of links. This outstanding property will eventually bring the dctransmission technology to ever broader application elds!
19
20
Chapter 3 Control Methods for VSC-HVDC Systems

This chapter describes various control methods used for VSC-HVDC systems. In Section 3.1 and Section 3.2 two existing control methods, i.e., power-angle control and vector current control are described. A novel control method, i.e., the so-called powersynchronization control, is introduced in Section 3.3. The major results of this chapter are summarized in Section 3.4. Some results of this chapter are included in [45].
3.1 Power-angle control

Power-angle control is also called voltage-angle control. It is perhaps the most straightforward controller for grid-connected VSCs [4648]. The principle of power-angle control is based on the following well-known equations U1 U2 sin X 2 U1 U1 U2 cos Q= X P =
(3.1)
where P and Q are the active and reactive powers between two electrical nodes in ac systems with voltage magnitudes U1 and U2 . The quantities and X are the phase-angle difference and line reactance between the two nodes. From (3.1) it follows that the active power is mainly related to the phase angle , while the reactive power is more related to the voltage-magnitude difference. These mathematical relationships are the foundation of power-angle control, i.e., the active power is controlled by the phase angle of the VSC voltage, while the reactive power or lter-bus voltage is controlled by the magnitude of the VSC voltage. Fig. 3.1 shows the main-circuit and control block diagram of a VSC-HVDC converter using power-angle control. Lc is the inductance of the phase reactor, and Ln is the 21
Chapter 3. Control Methods for VSC-HVDC Systems

VSC
Ln P, Q
Lc
Cf
+
uf
ic + v
Uf U ref + Q Qref +
P Pref + -
PLL
t
ref va ref vb vcref
RPC/AVC
Voltage reference control
APC
Fig. 3.1 Main-circuit and control block diagram of a VSC-HVDC converter using power-angle control.
inductance of the ac system. Cf is the ac capacitor connected at the lter bus. The bold letter symbols E, uf , and v represent the voltage vectors of the ac source, the lter bus, and the VSC respectively. P and Q are the active power and reactive power from the VSC to the ac system. The quantity ic is the current vector of the phase reactor. To produce three-phase alternating voltages, the VSC needs three variables: magnitude, phase angle and frequency. With power-angle control, these three variables are given by three different controllers, i.e., the reactive-power controller (RPC) or the alternatingvoltage controller (AVC), the active-power controller (APC), and the phase-locked loop (PLL). The above controllers are briey described in below: Reactive-power controller. The reactive power to/from the VSC is controlled by the magnitude of the VSC voltage. A proportional-integral (PI) controller can be applied, e.g., KiQ Q V = Kp + [Qref Q] . (3.2) s where the output V gives the change in magnitude of the VSC reference voltage. Alternating-voltage controller. Alternatively, the VSC-HVDC converter controls the lter-bus voltage instead of the reactive power. The output of the AVC is the same as that of the RPC. A PI controller can be applied, e.g., V =
U Kp +
KiU s
[Uref Uf ].
(3.3)
Active-power controller. The active power to/from the VSC is controlled by the phase angle of the VSC voltage. A proportional-integral (PI) controller can be ap22
3.1. Power-angle control

plied, i.e., v =
P Kp +
KiP s
[Pref P ].
(3.4)
where the output v gives the change in phase angle of the VSC reference voltage. Phase-locked loop. The function of the PLL is to synchronize the VSC to the ac system. Below a description of a PLL design suitable for power-angle control is given. If 1 is the angular frequency of the ac system, and is the angular frequency of the VSC, a PLL controller has the objective to follow the phase angle of the lter-bus voltage by minimizing e = (1 )t. (3.5) If the error e in (3.5) is sufciently small, (3.5) can be approximated by e sin(1 t t)
= sin 1 t cos t cos 1 t sin t
(3.6)
where sin 1 t and cos 1 t can be obtained by transformation of the lter-bus voltage as shown below. The phase quantities of the lter-bus voltage can be dened as uf a = Uf 0 cos(1 t) uf b = Uf 0 cos(1 t 120 ) uf c = Uf 0 cos(1 t 240 ). (3.7)
The corresponding real and imaginary parts of the vector uf in the stationary reference frame (see Appendix A) can be written as uf = Uf 0 cos 1 t, uf = Uf 0 sin 1 t. Substituting (3.8) into (3.6), yields e uf uf cos t sin t. Uf 0 Uf 0 (3.9) (3.8)
A PI controller can be used to minimize the error e, i.e., PLL =

PLL Kp +
KiPLL s
e.
(3.10)
Fig. 3.2 shows the control block diagram of the PLL. The angle change PLL is added to a reference frequency signal ref t. 23

ref t + t
+
u f 1 Uf0
COS
u fa u fb u fc
abc
+ -
PLL + Kp
SIN
K iPLL s
PLL
u f
1 Uf0
Fig. 3.2 PLL for power-angle control.
Voltage-reference control. By having V , v and t, the three-phase reference voltages of the VSC can be formulated as:
ref va = (V0 + V ) cos(t + v ) ref vb = (V0 + V ) cos(t + v 120 )
ref vc = (V0 + V ) cos(t + v 240 )
(3.11)
where V0 is a nominal voltage reference, e.g., V0 = 1.0 p.u. As shown in this section, the design and implementation of power-angle control is simple and straightforward. However, power-angle control practically has never been applied to any real VSC-HVDC system, since it suffers from two fundamental problems: 1. The control system has no general means to damp the various resonances in the ac system. Therefore, the bandwidth of the controller is very much limited by the resonances in the ac system, especially the one at the grid frequency [48]. Fig. 3.3 shows a plot of active-power and reactive-power step responses with powerangle control. The resonance at the grid frequency can be easily observed. Although the resonance can be damped out by applying notch lters in the active-power and reactive-power controllers, or canceled by some model-based control designs [49], the effects of such measures are doubtful since the ac system is a highly uncertain process where not all of the resonance frequencies are known. 2. The control system does not have the capability to limit the valve current of the converter. This is a serious problem, as the converters of a VSC-HVDC link usually do not have over-current capability. It is very important for the control system to limit the valve current to prevent the converters from being blocked (tripped) at disturbances. 24
3.2. Vector current control

0.15 Pref, P (p.u.) 0.1 0.05 0 -0.05 0 0.15 Qref, Q (p.u.) 0.1 0.05 0 -0.05 0 0.5 1 time (sec) 1.5 2 Q Q
ref
Pref P
0.5
1 time (sec)
1.5
Fig. 3.3 Step response of active power (upper plot) and reactive power (lower plot) with powerangle control. Observe the resonance at the grid frequency.
3.2 Vector current control

Vector current control of VSCs has initially been applied to variable-speed drives, where the VSC is connected to an ac motor [50,51]. By utilizing the dq decoupling technique, the control system is able to control the torque and ux independently. Once eld orientation is obtained the ac motor can be controlled using principles quite similar to those of dcmotor control. Vector current control is generally considered as a substantial step in acmotor control [52]. The application of vector current control on grid-connected VSCs is often considered as a dual problem of drive control [17,53]. The basic principle is to control the active power and the reactive power independently through an inner-current control loop [54,55]. As shown in Fig. 3.4, the essence of vector current control is that the control system creates a converter dq frame, where a PLL is applied to make sure that the d -axis of the converter dq frame is always aligned with the lter-bus voltage in order to synchronize the VSC to the ac system. A simple implementation of a current controller is achieved by the proportional-type control law
c c c vref = c Lc (iref ic c ) + j1 Lc ic + HLP (s)uf
(3.12)
where c is the desired closed-loop bandwidth of the inner-current controller, iref is the c converter current reference, and vref is the voltage reference of the VSC. The superscript c denotes the converter dq frame. The term j1 Lc ic c is used to remove the so-called crosscoupling. The function HLP (s) is a low-pass lter to improve the disturbance rejection 25

Grid-q VSC-q
v
uf
E
VSC-d
Grid-d
Fig. 3.4 Converter dq frame of vector current control.
capability of the current controller. HLP (s) has the following expression f HLP (s) = s + f
(3.13)
where f is typically chosen with a bandwidth (40 100 rad/s) [13]. By applying the control law in (3.12), vector current control is claimed to be able to control the active power and the reactive power independently by the d and q components of the current reference iref . However, this is a conditional correct conclusion, which is only valid if the lter-bus voltage is sufciently stiff such that its dynamics are negligible. A brief analysis is given below. In a stationary frame, the dynamic equation of the phase reactor of the VSC can be described by Kirchhoffs voltage law as dis = vs us (3.14) f dt where the superscript s denotes the stationary reference frame. If it is assumed that Lc = 1 (3.15)
at all times, i.e., the angular frequency of the converter dq frame equals the angular frequency 1 of the grid, the following relations are established
c j1 t c j1 t us , is , vs = vc ej1 t . f = uf e c = ic e
(3.16)
Substituting (3.16) into (3.14) yields the dynamic equation in the converter dq frame dic c c = vc uc (3.17) f j1 Lc ic . dt c If the switching-time delay is neglected and it is assumed that |vref | does not exceed the c c maximum voltage modulus, then v = vref . Substituting (3.12) into (3.17) yields Lc Lc 26 dic s c = c Lc (iref ic uc . c) dt s + f f (3.18)

By writing (3.18) in component form and applying Laplace transform (s = d/dt), yields ic cd = s c ref id uc s + c Lc (s + c )(s + f ) f d s c ref ic iq uc . cq = s + c Lc (s + c )(s + f ) f q
(3.19)
The above design approach for the inner-current controller is often referred to as internalmodel control (IMC) design [51], since the bandwidth of the inner-current control is explicitly specied in the control parameters. Another common design approach is the deadbeat-current control design [56, 57], which can only be realized by digital implementations. Generally speaking, if the bandwidth c of IMC is chosen sufciently high, IMC and deadbeat-current control give similar results. For either of the control design, the control bandwidth is basically limited by the switching frequency of the PWM and the sampling period of the computer. Moreover, both methods rely on a good knowledge of the value of Lc . The following analysis will establish the relations between the active/reactive power and the current references of the inner-current control. Assuming per unit quantities, the instantaneous active power and reactive power from the VSC to the lter bus are given by
c c P = Re uc , Q = Im uc . f (ic ) f (ic )
(3.20)
Linearizing (3.20) yields the following expressions P = ic cd0 ic cq 0 ic d0 ic q0

T T
uc fd uc fq uc fq uc fd
uc f d0 uc f q0 uc f d0 uc f q0
ic cd ic cq ic cq ic cd
Q =
(3.21)
where the subscript 0 denotes the operating-point value. In the converter dq frame, in the steady state, the q component of the lter-bus voltage equals zero and the d component equals the voltage magnitude, i.e.,
c uc f d0 = Uf 0 , uf q 0 = 0.
(3.22)
If the dynamics of the lter-bus voltage are neglected, it follows that

c uc f d = uf q = 0.
(3.23)
By substituting (3.22) and (3.23) into (3.21), the expressions of P and Q can be simplied as c P = Uf 0 ic (3.24) cd , Q = Uf 0 icq . 27

Linearizing (3.19) and further substituting it into (3.24) [with the condition in (3.23)] yields the following relationship in transfer matrix form P Q =
c Uf 0 s+ c
0
c Uf 0 s+ c
iref d iref q
(3.25)
J(s)
The transfer matrix J(s) is called Jacobian transfer matrix in this thesis, which is a general concept for dynamic modeling of ac/dc systems that is to be introduced in Chapter 4. Eq. (3.25) shows that the Jacobian transfer matrix J(s) is diagonal, i.e., P is ref only related to iref d while Q is only related to iq , and no cross-coupling between the two loops exists. However, (3.25) is derived based on the assumptions of (3.15) and (3.23). Both of the assumptions are related to the stiffness of the lter-bus voltage. If the ac system is strong enough, i.e., Ln Lc , the dynamics of the lter-bus voltage can be neglected. However, if the ac system is weak, the assumptions in (3.15) and (3.23) no longer hold. Therefore, the weaker the ac system, the higher the off-diagonal elements in J(s), i.e., the more interactions between the active-power and the reactive-power control. Consequently, to analyze the stability of vector current control for VSC-HVDC links connected to weak ac systems, the dynamics of the lter-bus voltage have to be considered. That is, the Jacobian transfer matrix J(s) in (3.25) should take into account the grid inductance Ln and the dynamics of the PLL. Such a model will be developed in Chapter 4. An in-depth analysis of the difculty with vector current control for weak-acsystem connections will be given in Chapter 5. Fig. 3.5 shows the main-circuit and control block diagram of a VSC-HVDC converter using vector current control. The active-power controller and the reactive-power or the alternating-voltage controller of vector current control can be designed in a similar ref way as power-angle control but with iref d and iq as outputs. However, the PLL can be designed in a more concise way by utilizing the concept of the converter dq frame , i.e., a PI controller is applied to minimize the q component of the lter-bus voltage in the converter dq frame K PLL PLL PLL = Kp + i Im{uc (3.26) f }. s In this way, the VSC is synchronized to the ac system. Fig. 3.6 shows the control block diagram of the PLL for vector current control. With vector current control, the voltage ref ref reference of the VSC is formulated by vd , vq and t. This is essentially the same as power-angle control where the reference of the VSC voltage is formulated by the magnitude, the phase angle and the frequency. In the former case the rectangular form is used while the latter uses polar form. The mathematical expressions of the dq- and abc blocks in Fig. 3.5 are given in Appendix A.3. For vector current control, given sufciently high bandwidth, the dq components of the converter current always follow the corresponding current references. Consequently, 28

VSC E Ln P, Q Cf Lc
+
uf
ic + v
Uf U ref + -
Q Qref + P Pref + -
RPC/AVC APC PLL
ref iq
uf
ref id
ref ref v vd Inner v ref ref vq current dq controller
ic
abc
ref va ref vb vcref
Fig. 3.5 Main-circuit and control block diagram of a VSC-HVDC converter using vector current control.
ref t
u
c f
Im{u }
c f
PLL p
K PLL PLL + + i s
Fig. 3.6 PLL for vector current control.
by limiting the modulus of the current references, the valve current of the converter is limited. The simulation results in Fig. 3.7 show the fault ride-through capability of a VSCHVDC link using vector current control. A three-phase ac-system fault with 0.2 s duration is applied at 0.1 s close to the lter bus. The modulus of the current reference [|iref c | =
2 ref 2 (iref d ) + (iq ) ] reaches the current limit Imax immediately after the fault occurrence. The control system automatically limits the converter current. After the fault is detected, the control system reduces the fault current to half of the maximum load current (or any other desired values to minimize the short-circuit current contribution to the ac system) except a very short current spike at the fault inception. In VSC applications, regardless of the control principle, the converter always tries to protect itself from excessive over currents. This fast protection is often implemented as a low-level hardware system, and its objective is to protect the converter in cases where the higher levels of control fail. Since the current spike in Fig. 3.7 is so short ( < 1.6 pu in magnitude and < 5 ms in duration ), it neither does any harm to the converter valve, nor contributes much to the short-circuit current to the ac system. Besides the fault-current limitation capability, the current control also has a damping effect on resonances in the ac system. Therefore, vector current control overcomes the two fundamental problems of power-angle control. In practice, vector current control
29

1 P (p.u.) 0.5 0 0 id , iq (p.u.) 1 0.5 0 0 2 | ic| (p.u.) 1 0 0 0.2
0.2 i ref
d
0.4
0.6
0.8
ref
i ref
q
ref
0.4
0.6
0.8
0.2
0.4 time (sec)
0.6
0.8
Fig. 3.7 Fault ride-through capability of vector current control. Upper plot: active power from the VSC. Middle plot: dq components of the current reference. Lower plot: modulus of the converter current.
has been successfully applied to a number of commercial VSC-HVDC links. However, a major drawback of vector current control is its poor performance for VSC-HVDC links connected to weak ac systems, which becomes an obstacle for VSC-HVDC transmission to be applied in more challenging ac-system conditions. To overcome the problem of vector current control with weak-ac-system connections, a novel control method, i.e., power-synchronization control, is proposed in the next section. In some sense, power-synchronization control might be viewed upon as a combination of power-angle control and vector current control.
3.3 Power-synchronization control

With the two control methods described in the previous sections, a PLL is used to synchronize the VSC with the ac system. This has since long been believed to be a pre-condition for any grid-connected VSC. However, there is, in fact, an internal synchronization mechanism in ac systems that the VSC can utilize to synchronize with the ac system. In this section, a control method based on such type of synchronization is proposed. The major goal of the proposed control method is to overcome the problem of vector current control with weak-ac-system connections. 30
3.3. Power-synchronization control

J1 dm1 d m 2 = Tm1 Te1 J2 = Te 2 Tm 2 dt dt SM 2 SM1 X
Fig. 3.8 Synchronization mechanism between SMs in an ac system.
3.3.1 Power-synchronization mechanism in ac systems

In this sub-section, the power-synchronization mechanism between synchronous machines (SMs) in ac systems is described. The mechanism is illustrated by a simple system consisting of two interconnected SMs as shown in Fig. 3.8. SM1 operates as a generator and SM2 operates as a motor. The reactance X is the sum of the reactances of the SMs and that of the line interconnecting the two SMs. All resistances and other damping effects are disregarded. Initially, it is assumed that the two SMs operate at steady state, as described by the phasor diagram in Fig. 3.9(a). The phasors E1 and E2 represent the line-to-line equivalents of the inner emfs of the two SMs respectively. These emfs are assumed to be constant at all times (even during transients). The electric power transmitted from SM1 to SM2 is given by E1 E2 sin P = (3.27) X where is the electrical angle separating the two emfs E1 and E2 . The mechanical torque Tm1 of SM1 is now increased by a certain amount for a short duration and then brought back to its initial value. As a consequence of the temporary increase of Tm1 , the mechanical angle of the rotor of SM1 advances, as predicted by the generator-mode swing equation dm1 = Tm1 Te1 (3.28) J1 dt where J1 is the total inertia of the shaft-system of SM1 , m1 is the rotor speed, and Te1 is the electromagnetic torque of of SM1 . Since the emf of a synchronous machine is tightly connected to the rotor position, the advance of the mechanical angle of the rotor of SM1 inevitably causes an advance of the phase of the emf of SM1 , as indicated by the phasor E1 in Fig. 3.9(b). The initial position of E1 is shown as a dashed line in Fig. 3.9(b). Due to the phase advancement indicated by E1 , the phase difference between the emfs of the two SMs is increased. According to (3.28), this translates into an increase of the electric power transmitted from SM1 to SM2 . This increase in power is equivalent to an increase in the electromagnetic torque Te2 of SM2 . Assuming that SM2 has a constant load torque Tm2 , the rotor of SM2 starts to accelerate as dictated by J2 dm2 = Te2 Tm2 dt (3.29) 31

E1 E2 E1 ' E2
(a) (b)
Fig. 3.9 Phasor diagrams describing power synchronization.
where J2 is the total inertia of the shaft-system of SM2 , and m2 is the mechanical angular velocity of SM2 . As the rotor of SM2 starts to accelerate, the same thing occurs with the phase of E2 , as indicated by the arc-shaped arrow in Fig. 3.9(b). The acceleration of the phasor E2 results in a reduction of the phase difference between the emfs of the two SMs. After a transient, which in reality involves a certain amount of damping, the phase difference between the emfs of the two SMs is brought back to its initial value (as the transmitted electric power), and the system is again at steady state. The synchronization mechanism described above is known to all power system specialists, i.e., the synchronization process is achieved by means of a transient power transfer. The same kind of synchronization also appears in large systems of interconnected synchronous machines. Due to the fact that the synchronous machines can maintain operation in various acnetwork conditions while the vector-current-controlled VSCs are prone to fail, it makes sense to suggest a control method based on a synchronization process where the electric power is the communicating medium. In the next sub-section, a controller based on power synchronization is proposed.
3.3.2 Power-synchronization control of grid-connected VSCs

From the discussion in the preceding sub-section it is known that the SMs in an ac system maintain synchronism by means of power synchronization, i.e., a transient power transfer. This power transfer involves a current which is determined by the interconnecting network. Generally, this current is unknown. If power synchronization should be used to control a VSC, therefore, it cannot be combined with a vector current controller, which requires a known current reference. As will be shown below, the active power output from the VSC is instead controlled directly by the power-synchronization loop and the reactive power (or alternating voltage) is controlled by the magnitude of the VSC voltage. Consequently, an inner current loop is not necessary from a power and voltage control point of view. However, as it was mentioned in Section 3.2, besides power and voltage control, the current controller is also important in 1. Providing damping effects to poorly-damped resonances in ac systems. 2. Limiting the valve current of the converter during severe ac-system faults. 32

To make the essence of power-synchronization control easier to be captured, the initial proposal of the control law only considers the damping issue, while the fault-currentlimitation issue is proposed as a modication of the initial design. Accordingly, an initial control design based on the power-synchronization law is proposed as Power-synchronization loop (PSL). The control law is given by v = kp (Pref P ). (3.30) s where v supplies the synchronization input to the VSC, i.e., t = ref t + v . The power-synchronization loop is essentially an emulation of the swing equation, however, not an exact copy. Since the mechanical angular velocity m is the derivative of the angular position, (3.28) represents a double integration when going from torque (or electric power) to angular position. This double integration, inherently, yields a poor phase margin even with considerable damping. Therefore, the proposed power-synchronization law in (3.30) employs only a single integration.
Alternating-voltage control (AVC). The control law is given by V =
ku (Uref Uf ). (3.31) s where V gives the change in magnitude of the VSC reference voltage. The AVC can also be viewed as an emulation of the exciter control of a synchronous machine. A normal exciter control of a synchronous machine is of proportional type. However, it is found to be more benecial to have integral process for the VSC to suppress high-frequency disturbances. If there are other voltage-controlling devices connected close to the lter bus, a load compensation should be applied to avoid voltage hunting. This issue will be discussed in Chapter 6.
Reactive-power control (RPC). When operating against a weak ac system, the VSC-HVDC converter should preferably be operated in AVC mode to give the ac system the best possible voltage support. In case reactive-power control is required, the output of this controller should be added to the alternating-voltage reference, and the added amount should be limited. The PI-type controller proposed in (3.2) can be used for the RPC but with voltage reference change Uref as output. Voltage-vector control law. The control law of the voltage vector of the VSC is proposed as c vref = (V0 + V ) HHP (s) ic (3.32) c where V0 is the nominal value, e.g., V0 = 1.0 p.u., and V is given by the AVC. HHP (s) is a high-pass lter for damping purpose, which is expressed by HHP (s) = kv s s + v
(3.33) 33

where v should be chosen low enough to cover all the possible resonances in the ac system. Typically v should be chosen between 30 rad/s and 50 rad/s to also cover the subsynchronous resonance in ac systems. The gain kv determines the level of the damping effect with typical values between 0.2 p.u. and 0.6 p.u. The effects of HHP (s) for resonance damping will be further analyzed in Chapter 4. In the following, a current limitation scheme is proposed as a modication of the initial design. The principle is to have the control system in current-limitation mode automatically once the converter current is above the limit. In case that the converter current is above the maximum limit Imax , the desired control law of the VSC is the inner-current control law of vector current control in (3.12). However, instead of giving a constant current reference to (3.12), the value of iref in (3.12) is given by iref = 1 c c c (V0 + V ) HHP (s) ic c HLP (s)uf j1 Lc ic + ic . c Lc (3.34)
The current reference in (3.34) is designed in such a way that the control law in (3.12) becomes (3.32) in normal operation. This can be easily veried by substituting (3.34) into (3.12). However, the current reference iref in (3.34) gives an indication of the actual converter current. During ac-system faults, current limitation is automatically achieved by limiting the modulus of iref to the maximum current limit Imax . A brief analysis of this is given below. The dynamics of the converter current in the converter dq frame can be described by dic c Lc c = vc uc (3.35) f j1 Lc ic . dt
c Assuming vc = vref , substituting (3.12) into (3.35) yields,
dic s Lc = c Lc (iref ic uc . c) dt s + f f By setting the time derivative and the Laplace operator s to zero, it is found that iref = ic c.
(3.36)
(3.37)
That is, the current reference is identical to the actual converter current in the steady state. In other words, by limiting the modulus of the current reference, the converter current is limited. Fig. 3.10 shows the overview of power-synchronization control. The Current reference control block corresponds to the control law described by (3.34), while the Current controller block corresponds to the control law described by (3.12). 34

U ref + Qref + QE Ln P, Q Cf RPC U ref + Uf -
k u V s AVC
VSC Lc
+
uf
ic
+
v
i ref Current- d ref reference iq control P Pref + -
ref vd ref v ref va
Current v ref controller q
dq
ref v
abc
ref vb vcref
kp s PSL
ref t
Fig. 3.10 Main-circuit and control block diagram of a VSC-HVDC converter using powersynchronization control.
3.3.3 Bumpless-transfer and anti-windup schemes

In some situations, power synchronization cannot be applied, and a back-up PLL has to be used instead. Those situations are: 1. Before the VSC is de-blocked, the back-up PLL provides the synchronization input to the VSC. After the converter is de-blocked, the PLL is replaced by the PSL. This procedure is similar to the auto-synchronization process used for synchronous machines before they are connected to the grid. 2. During severe ac-system faults, the control system has to limit the converter current. Thus, the PSL cannot be applied. According to the control law given by (3.32), the converter dq frame created by the PSL has the d -axis aligned with the voltage vector v of the VSC. However, the dq frame created by the PLL has the d -axis aligned with the voltage vector uf of the lter bus. As shown in Fig. 3.11, the two dq frames differ by an angle c . Consequently, the dq frame created by the PLL should be corrected to minimize the bump of transfer when the synchronization input of the VSC is switched from the PSL to the PLL. The angle c can be derived from the power-angle equation in (3.27) with the p.u. expresseion c = arcsin( P Xc ) arcsin(P Xc ) Uf V (3.38) 35

where Xc is the reactance of the phase reactor. Uf and V are the magnitudes of the lterbus voltage and the VSC voltage, which are close to 1.0 p.u. in normal operation. Fig. 3.12 shows the block diagram of the backup PLL, which is essentially the same as the PLL for vector current control in Fig. 3.6 but with angle correction. However, if the synchronization input of the VSC is switched to the PLL, the PSL still controls the angle output v . At the time when the synchronization input is switched back from the PLL to the PSL, v might not necessarily be the same as PLL . This most likely happens during severe ac-system faults, where the PSL tends to increase v rapidly. In Fig. 3.12, a bumpless-transfer scheme is proposed to handle the above situation. Once v , the scheme integrates the error and feeds back a term there is an error between v and to cancel the integral action of the PSL. From the scheme, the following relationship is established p kp v = (Pref P ) + v . (3.39) s + p s + p v . In other words, once the synchronization If p is sufciently large, (3.39) yields v
input of the VSC is switched from the PSL to the PLL, the bumpless-transfer scheme will make sure that the output of the PSL tracks the output of the PLL to avoid the transfer bump when the synchronization input is switched back to the PSL. 1 In the following, the anti-windup issue of the alternating-voltage controller is discussed. With power-synchronization control, the alternating-voltage controller controls ref the d component of the voltage reference vd [cf. (3.32)], which is essentially the magniref tude of the VSC voltage. However, vd might be limited by either of the two fundamental limitations of the VSC: 1. Converter-current limitation. This mainly happens during ac-system faults as it has been described above.
ref 2. Modulation-index limitation. Occasionally, the magnitude of vd is above the maximum value of voltage that the VSC can produce. Thus, the control has to limit the ref magnitude of vd to prevent over-modulating the valve. Modulation-index limitation tends to be reached more often with weak-ac-system connections, where a relatively higher VSC voltage magnitude is necessary to keep the lter-bus voltage to be nominal. Usually, the tap-changer of the converter transformer is used to prevent modulation-index limitation. However, it can still occur with transient voltage swings, or if the tap-changer of the converter transformer has reached the limit of the tapping range.
The principle of the anti-windup scheme of the alternating-voltage control is basically the same as the bumpless-transfer scheme of the PSL, as shown in Fig. 3.13. Whatever the
Interestingly, this is essentially the same problem as transient stability phenomena in ac systems where synchronous machines accelerate the speed of their rotors during ac-system faults. Thus, the well-known equal-area criterion might be viewed as an integrator-windup problem. Unfortunately, no bumpless-transfer or anti-windup scheme can ever be designed for a real machine.
1
36
Grid-q PLL-q PSL-q PSL-d

v
Angle correction
uf
PLL-d
c
E
Grid-d
Fig. 3.11 Converter dq frames created by the PSL and the PLL. The two dq frames differ by the angle c .
Angle correction
c
u
c f
Im{u cf }
PLL Kp +
PLL i
PLL +
+
' PLL
s kp s PSL + +
Backup-PLL Pref +
ref t
v
+ v
p
s
Fig. 3.12 Bumpless-transfer scheme for switching the synchronization input of the VSC.
37

Currrent limitation U ref + ku V + s + AVC
c V0 H HP ( s )icd
Modulation index limitation

ref d v
+ +
ref vd
Uf
E
s
Fig. 3.13 Anti-windup scheme for the alternating-voltage controller.
ref reason of the voltage limitation, the anti-windup scheme integrates the error of vd and ref v d , and feeds back a term to cancel the integrator of the alternating-voltage controller. By ref ref choosing E large enough, the output vd tracks the limited d component reference v d . Once the voltage limitation lifts, the alternating-voltage control gets a smooth re-start. In the following, some simulation results from PSCAD/EMTDC are shown to demonstrate the performance of a VSC-HVDC link using power-synchronization control. The VSC-HVDC link is connected to a very weak ac system with SCR = 1.0. The converter station at the other side of the VSC-HVDC link controls the direct voltage, and the converter is assumed to be connected to a strong system. It is also assumed that the bandwidth of the direct-voltage controller at the other station is high enough such that the variation of the dc-link voltage has negligible effects on the dynamics of the converter connected to the weak ac system. The issue of interconnection of two very weak ac systems will be discussed in Chapter 5, where the direct-voltage control plays a central role. Fig. 3.14 shows the converter deblocking and power ramping-up process. In the lower plot, it is shown that the VSC uses the PLL as its synchronization input before the converter is deblocked. The output of the PSL, v , is forced to be equal to the output of the PLL, PLL . After the converter is deblocked at 0.1 s, the PSL takes over the synchronization input of the VSC, and ramps up the active power to 0.86 p.u. The PLL tracks the PSL as the power is ramping up, and is exactly identical to the PSL when the system reaches the steady state. In Fig. 3.15, a three-phase ac-system fault with duration 0.2 s is applied close to the lter bus. By detecting the ac-system fault (by current limitation or magnitude drop of the lter-bus voltage), the control system switches the synchronization input of the VSC to the backup-PLL. In the middle plot, it is shown that the PSL initially quickly increases the angle output v , but after the back-up PLL takes over the synchronization input of the VSC, the bumpless transfer scheme re-directs the output of the PSL to follow the output of the PLL. Once the ac-system fault is cleared, the PSL takes over the synchronization input and brings the power back to the pre-fault level. The PLL again tracks the PSL and
38

1 0 -1 0 1 P (p.u.) 0.5 0 0 v, PLL (rad) 1.5 1 0.5 0 0 0.2 0.4 time (sec) PLL 0.6 0.8 1 PSL 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
Fig. 3.14 Converter deblocking and power ramping up of the VSC-HVDC link using powersynchronization control. Upper plot: three-phase lter-bus voltages. Middle plot: active power of the VSC-HVDC link. Lower plot: outputs of the PSL and the PLL.
equals the PSL in the steady state. The lower plot of Fig. 3.15 demonstrates the anti-windup scheme of the alternatingvoltage controller. During the fault period, the d component of the VSC voltage reference ref vd is limited by the current controller. Similar to the bumpless-transfer scheme of the PSL, the anti-windup scheme of the alternating-voltage controller also re-directs the voltage output of the AVC to follow the limited voltage reference. After the fault is cleared, the AVC takes over the voltage control. Fig. 3.16 shows the actions of the current controller during the fault. After the threephase ac-system fault is applied at 0.1 s, the modulus of the current reference [|iref | =
2 ref 2 (iref d ) + (iq ) ] reaches the current limit Imax . The control system seamlessly switches ref to the control law (3.12), where the limited iref d and iq become the inputs to the current controller. Only a short current spike is observed on the valve current |ic | at the fault occurrence stage, which usually does no harm to the converter valve. After detecting the faults, the current controller reduces the valve current to half of the maximum load current Imax or any other desired values to minimize the short-circuit current contribution to the ac system. After the fault is cleared at 0.3 s, the current limitation lifts, and the voltagevector control law is back to (3.32). As shown by the time simulation in this section, power-synchronization control, in fact, uses vector current control and the PLL during severe ac-system faults, since the power-synchronization law is not applicable. A question is naturally raised: if vector
U (p.u.)
39
1 U (p.u.) 0 -1 0 v, PLL (rad) 3 2 1 0 0 1.5 vd (p.u.) 1 0.5 0 0 0.2 Controller vref

d
0.2
0.4
0.6
0.8
PSL
PLL
0.2
0.4
0.6
0.8
ref
Limited vd 0.4
ref
0.6 time (sec)
0.8
Fig. 3.15 Bumpless-transfer and anti-windup schemes of power-synchronization control during a three-phase ac-system fault. Upper plot: three-phase lter-bus voltages. Middle plot: outputs of the PSL and the PLL. Lower plot: d component of the VSC voltage reference.
1 P (p.u.) 0.5 0 0
ref ref , iq (p.u.) id
0.2
ref id
0.4
0.6
0.8
1 0 -1 0 2
ref iq
0.2
0.4
0.6
0.8
| ic| (p.u.)
1 0 0
0.2
0.4 time (sec)
0.6
0.8
Fig. 3.16 Current limitation of power-synchronization control during a three-phase ac-system fault. Upper plot: active power of the VSC-HVDC link. Middle plot: dq components of the current reference. Lower plot: modulus of the converter current.
40

current control does not work well with weak-ac-system connections, could instability occur during ac-system faults? As will be shown in Chapter 5, the stability of the current controller deteriorates with weak-ac-system connections but to an acceptable level. What is not acceptable is using current references to control the active power and the alternating voltage. The justication of using power-synchronization control is that it is meaningless to just control the current during normal operation. However, it is enough during severe ac-system faults. As shown in Fig. 3.16, vector current control works fairly stable during ac-system faults given the fact that the ac system only has SCR = 1.0. VSCs using power-synchronization control basically emulate the operation of a synchronous machine. Therefore, it contributes with short-circuit capacity to the ac system at the PCC. In order to obtain a simple estimate of the impact of the VSC on the short-circuit capacity of ac system, the effects of the alternating-voltage control and the ac lter can be disregarded. Doing so, the short-circuit capacity of the ac system including the VSC at the PCC can be expressed as
2 Uf 0 (Ln + Lc ) . Sac = 1 Ln Lc
(3.40)
However, VSCs do not necessarily increase short-circuit currents to the ac system during ac-system faults thanks to the current limitation function.
3.3.4 Negative-sequence current control

In reality, there is often phase unbalance in the three-phase voltages of the ac system. The phase unbalance may appear in the steady state due to unevenly connected loads in the ac system, but more often caused by unbalanced voltage dips [34]. The zero-sequence voltage in the ac system cannot pass through the converter transformer where Y type connection is commonly used, but the negative-sequence voltage can create unbalanced converter currents, which may overload the converter and produce double-grid-frequency ripples on the direct voltage [58]. In this section, a negative-sequence current controller is proposed to mitigate the phase-unbalance problem. Fig. 3.17 shows the control block diagram of the proposed negative-sequence current controller, which is principally the same as the positive-sequence current controller in Fig. 3.5 and Fig. 3.10 but with negative-sequence variables. The outputs of the negativeref ref ref sequence current controller va , vb and vc are directly added to the outputs of the ref ref ref positive-sequence current controller va , vb and vc . Sometimes, the two current controllers together are referred to as dual vector current controller [59, 60]. The minus sign in the subscripts of the variables in Fig. 3.17 represents the negative-sequence quantities. However, the dq frame of the negative-sequence current controller rotates in the opposite direction to the converter dq frame. Therefore, t is used as the synchronization input. The proposed control law of the negative-sequence current controller is expressed as 41

u f ic
ref id ref iq ref vd ref v ref va ref vb vcref
Innerref vq current controller
dq t
ref v
abc
Fig. 3.17 Negative-sequence current controller.

ref c c c v = c Lc iref ic + j1 Lc ic + uf
(3.41)
where c is the desired bandwidth of the negative-sequence current controller. Eq. (3.41) is essentially the same expression as (3.12) but with negative-sequence variables. The negative-sequence current reference iref is usually set to zero if there is no other special purpose. Thus, the principle of the negative-sequence current controller is to create phase unbalance on the VSC voltages so as to minimize the negative-sequence current owing through the converter. There are different ways to obtain negative-sequence variables. The following is a common approach [61]. In a stationary reference frame, if the zero-sequence component is disregarded, a complex vector ys (t) can be expressed as ys (t) = Y+ ejt + Y ejt
s (t) y+ s (t) y
(3.42)
s where Y+ is the magnitude of the positive-sequence vector y+ (t), while Y is the mags nitude of the negative-sequence vector y (t). The quantity is the angular speed of the s rotating vector y+ (t). It follows from (3.42) that a voltage vector delayed with a quarter period (Tp /4) and multiplied by j is
j ys (t Tp /4) = j [Y+ ej(tTp /4) + Y ej(tTp /4) ] or j ys (t Tp /4) = Y+ ejt Y ejt .
(3.43)
(3.44)
s Accordingly, the negative sequence vector y (t) can be obtained from (3.42) and (3.44) as 1 s y (t) = Y ejt = [ys (t) j ys (t Tp /4)]. (3.45) 2
Fig. 3.18 demonstrates the effect of the negative-sequence current controller by time simulations in PSCAD/EMTDC. At 0.1 s, a negative-sequence voltage source with 10% magnitude of the positive-sequence voltage is added on the ac source E, which makes the 42
3.4. Summary
1 Uf (p.u.) 0 -1 0 v ref, v ref (p.u.) 0.1 0 -0.1 -0.2 0 0.8 | i | (p.u.) 0.6 0.4 0 0.05 0.1 0.15 0.2 0.25 time (sec) 0.3 0.35 0.4 0.05 0.1 0.15 vdref
0.05
0.1
0.15
0.2 vqref
0.25
0.3
0.35
0.4
d-
q-
0.2
0.25
0.3
0.35
0.4
Fig. 3.18 Effects of negative-current control in the steady state. Upper plot: three-phase lter-bus voltages. Middle plot: dq components of the negative-sequence VSC voltage reference. Lower plot: modulus of the converter current.
converter current unsymmetrical. The lower plot in Fig. 3.18 shows the modulus of the converter-current vector in the converter frame that oscillates with double-grid frequency caused by the negative-sequence component. If the operating point of the VSC is close to the maximum valve-current limit, this phase unbalance might overload the converter. At 0.3 s, the negative-sequence current controller is activated, which successfully removes the negative-sequence current component. Fig. 3.19 and Fig. 3.20 show the effect of the negative-sequence current controller during a single-line-to-ground fault in the ac system. With the negative-current controller activated, as shown in Fig. 3.20, the unbalance of valve currents is much reduced. It should be noted that, since unbalanced VSC voltage need to be produced in order to reduce the negative-sequence current, the magnitudes of some phases of the VSC voltage might go above the modulation-index limitation. This is the reason why there is still remaining negative-sequence component on the converter current during the unbalanced fault even with the negative-sequence current controller applied in Fig. 3.20.
3.4 Summary
In this chapter, two existing control methods, i.e., power-angle control and vector current control, for VSC-HVDC systems are reviewed, and a novel control method, i.e., powersynchronization control, is proposed. Power-angle control is simple and straightforward to implement. However, power-angle control has no general means to damp resonances in 43
1 Uf (p.u.) 0 -1 0
0.2
0.4
0.6
0.8
1 0 -1 0 2 iq
ref
id
ref
0.2
0.4
0.6
0.8
| ic| (p.u.)
1 0 0
0.2
0.4 time (sec)
0.6
0.8
Fig. 3.19 Single-line-to-ground fault without the negative-sequence current controller applied. Upper plot: three-phase lter-bus voltages. Middle plot: dq components of the current reference. Lower plot: modulus of the converter current.
1 Uf (p.u.) 0 -1 0
0.2
0.4
0.6
0.8
1 0 -1 0 2
ref iq ref id
0.2
0.4
0.6
0.8
| ic| (p.u.)
1 0 0
0.2
0.4 time (sec)
0.6
0.8
Fig. 3.20 Single-line-to-ground fault with the negative-sequence current controller applied.
44
3.4. Summary
the ac system and the converter may get over current during ac-system faults. Vector current control is the most popular control method used for grid-connected VSCs. However, for HVDC applications, vector current control has poor performance for weak-ac-system connections. VSCs using power-synchronization control basically resemble the dynamic behavior of a synchronous machine. Several additional functions, such as high-pass current control, current limitation, etc. are proposed to deal with various practical issues during operation.
45
46
Chapter 4 Dynamic Modeling of AC/DC Systems

This chapter discusses the dynamic modeling issue for ac/dc systems. In Section 4.1, a new concept, i.e., the so-called Jacobian transfer matrix, is introduced for modeling of ac systems. By the proposed modeling concept, an synchronous ac system is viewed upon as one multivariable feedback control system where the Jacobian transfer matrix is dened as the controlled process. In Section 4.2 and Section 4.3, the Jacobian transfer matrix concept is applied to model grid-connected VSCs using power-synchronization control and vector current control. In Section 4.4, the Jacobian transfer matrix concept is extended to modeling of dc systems. The properties of the Jacobian transfer matrix are summarized in Section 4.5. The major results of this chapter are summarized in Section 4.6. Some results of this chapter are included in [62, 63].
4.1 Jacobian transfer matrix for ac-system modeling

In this section, a new concept for dynamic modeling of ac systems is described. In Section 4.1.1, the existing concepts for power-system stability and dynamic modeling are reviewed. In Section 4.1.2, the Jacobian transfer matrix is introduced as a unied dynamicmodeling concept for ac systems.
4.1.1 Power-system stability and dynamic modeling

Power-system stability has been a major subject for power engineering for many years. The denitions of power-system stability and dynamic-modeling techniques have continuously been extended with the evolvement of power systems [32, 6466]. In the early days, the stability problem was one of maintaining synchronous operation. Thus, transient stability was the most classical stability problem of concern. Transient stability is dened as the ability of a power system to maintain synchronism when subjected to severe transient disturbance [67]. Transient stability is basically an issue of energy balance of the rotors of synchronous machines, where the equal-area criterion is 47
Chapter 4. Dynamic Modeling of AC/DC Systems

often used to describe the mechanism. Small-signal stability is another stability phenomenon related to synchronous machines. In todays power systems, small-signal stability is largely a problem of insufcient damping of oscillations [32]. The fast excitation control has been the traditional culprit for reduced or even negative damping torque [68]. Insufcient damping may also be a consequence of large power systems interconnected by weak ties [69]. The power-system stabilizer (PSS) are generally considered as an easy and effective solution for small-signal stability issues, at least for the local-mode oscillations [70]. HVDC systems and exible ac transmission systems (FACTS) have also been shown to be effective in damping interarea oscillations [4, 28, 71]. For small-signal stability, modal analysis based on linearized ac-system models is the most popular tool for analyzing the phenomenon and evaluating various damping solutions [32]. The above two forms of stability are mainly related to the dynamic behavior of synchronous machines. Consequently, they are often referred to as rotor-angle stability. Voltage stability was identied as a problem in the 1970s. Voltage stability is dened as the ability of a power system to maintain acceptable voltage at all buses in the system under normal operating conditions and after being subjected to a disturbance [67]. Voltage stability is often caused by lack of reactive-power support in heavy loaded areas [66]. In contrast to rotor-angle stability, voltage stability is more related to the dynamic behavior of loads and their interactions with the ac system. Dynamic load modeling is generally believed to be of central importance for voltage-stability study [72, 73]. For voltage-stability analysis, an interesting methodology was proposed, i.e., the Jacobian matrix. The Jacobian matrix was originally applied for solving power equations using the Newton-Raphson algorithm in power-ow programs. It has been found that the singularities of the Jacobian matrix have close relationships with voltage-stability problems. For instance, it can easily be shown that the critical points on the well-known P V curves exactly correspond to the operation conditions when the Jacobian matrix becomes singular. Based on the Jacobian matrix, a modal analysis technique that is similar to the one used for small-signal analysis was developed [74]. However, the applications of the Jacobian matrix and other related methods in voltage-stability analysis have been questioned for their mathematical foundations [75]. Voltage instability, like any other kind of instability, is a dynamic phenomenon. A rigorous analysis, therefore, would require a dynamic system formulation and the application of appropriate dynamic criteria, but the Jacobian matrix is merely a static matrix that describes the power-ow deviations related to voltage and angle variations in the system at a certain operating point. One argument to support the Jacobian matrix and the related methods is that the Jacobian matrix might be viewed as a dynamic description of the ac system if the frequency range of the stability of concern is quasi-static [76]. As it will be shown later by the Jacobian transfer matrix modeling technique introduced by this thesis, such an explanation does make sense. However, the question is, if the Jacobian matrix can be viewed as 48
4.1. Jacobian transfer matrix for ac-system modeling

a dynamic description of the ac system that is valid in the quasi-static frequency range, then what is the dynamic description of the ac system that is valid in the whole frequency range, and how are they related mathematically ? The subsynchronous resonance (SSR) problem appeared also in the 1970s. The rst SSR problem was experienced in 1970 resulting in the failure of a turbine-generator shaft at the Mohave plant in southern California, USA [32]. In the beginning, it was believed that SSR occurs only in series-compensated transmission systems, where the long shaft of the rotor of thermal units may interact with the series capacitor at a frequency lower than the fundamental frequency. However, it was later discovered that similar interactions can also occur with an HVDC converter [77]. Therefore, subsynchronous torsional interaction (SSTI) is suggested to be a more general denition [78]. For rotor-angle and voltage stability, the phasor theory is the major mathematical tool for dynamic modeling and analysis. With the phasor approach, electromagnetic transients of the ac network are neglected. Such a simplication is acceptable for rotor-angle and voltage stability but not for SSTI, since the dynamic frequency range of SSTI is much higher than rotor-angle and voltage stability. For analyzing SSTI, the electromagnetic transients of the ac network have to be properly represented. Thus, the space-vector theory is commonly applied to formulate the dynamic model of the ac network. There are predominantly two methods for SSTI analysis: eigenvalue analysis and frequency scanning [22, 79, 80]. Eigenvalue analysis is a rigorous tool for power-system modeling, since the whole power system is modeled as one single state-space representation. The computation of eigenvalues and eigenvectors is an excellent method of providing crucial information about the nature of the power system [21]. The major drawback of the eigenvalue analysis is its complexity once the order of the system is high. The frequencyscanning method is based on the complex-torque concept [81]. The risk of SSTI can thus be evaluated by plotting the damping-torque curve of the generator in the frequency domain. Although frequency scanning is a frequency-domain concept, the damping-torque curve is also possibly obtained from time-simulation programs by perturbing the rotor speed with a small sweeping-frequency sinusoidal signal. As the aforementioned reason, the time-simulation program used for frequency scanning should be of electromagnetictransient type, such as EMTP or PSCAD/EMTDC. In the foreseeable future, the main challenges for power-system stability analysis and dynamic modeling are imposed by high power-electronic devices, such as HVDC systems and FACTS if the number of such devices is considerably increased in power systems. On the one hand, the ratings of those devices are often high enough to have a signicant impact on all the above-mentioned stability phenomena. On the other hand, their dynamic frequency range are much higher than the traditional power-system components. They might interact with power systems and with each other in a more complex way. From the above brief review of power-system stability and the related dynamic49

Generator
K1
K5
FACTS Jacobian transfer matrix J ( s)
K2
HVDC system
K3
Constant power load K 4
Induction motor
Fig. 4.1 An ac network connected to various input devices.
modeling techniques, it can be concluded: 1. Due to historical and practical reasons, power-system stability has traditionally been dened mainly based on physical phenomena, but the mathematical relationships between different forms of stability are not rigourously claried. The dynamic modeling techniques are also versatile for various forms of stability. 2. For any dynamic system, the stability of the system is usually closely related to feedback control. However, power-system stability has rarely been interpreted from the feedback-control point of view. 3. High power-electronic devices might interact with the ac system and each other in a more complex way than conventional power components. New modeling and analysis techniques are required to meet the challenges in the future. In the next sub-section, the Jacobian transfer matrix is introduced as a unied ac-system dynamic modeling technique to address these issues.
4.1.2 Feedback-control view of power systems

The fundamental idea of the Jacobian transfer matrix modeling is that a synchronous power system is modeled as one multivariable feedback-control system, where the feedback controllers and the controlled process of the power system are explicitly dened. As shown in Fig. 4.1, for a normal power system, there are various power components 50
4.1. Jacobian transfer matrix for ac-system modeling

r +
K ( s) J ( s) y
Fig. 4.2 Feedback-control view of a power system.
and the rest of the power system, i.e., the ac network and the electrical parts of those power components form another transfer matrix J(s). A synchronous power system thus is modeled by the feedback-control system shown in Fig. 4.2. The transfer matrix J(s) is called Jacobian transfer matrix in this thesis, i.e., the Jacobian transfer matrix is dened as the controlled process of a power system.1 With the proposed modeling concept, the stability of a power system is uniquely dened as the stability of the closed-loop system formed by the controller transfer matrix K(s) and the Jacobian transfer matrix J(s). Some terminologies related to the proposed modeling concept are dened in below: Input device: Any power component connected to a power system which has a feedbackcontrol property. The aforementioned synchronous generators, induction motors, HVDC systems, and FACTS are all treated as input devices. An input device consists of two parts: the electrical part and the controller. AC network: The passive power components in a power system, such as transmission lines, transformers, line inductors, shunt capacitors, resistive-inductive-capacitive (RLC) loads, etc. Controller transfer matrix: A transfer matrix formed by all the controllers from each individual input device. The controller transfer matrix is usually a diagonal or blockdiagonal transfer matrix unless there is cross controls between the input devices. The controller as dened in this thesis is a generalized terminology. For example, the rotors of synchronous generators and induction motors are treated as controllers even though physically they are not intentionally implemented controllers as such.
1
connected to it, such as generators, induction motors, HVDC systems, FACTS devices, and loads, etc. The controllers, as will be dened later, are pulled out from those power components and form a controller transfer matrix as K1 ( s ) K ( s ) 2 K (s ) = . .. Kn ( s )
The name Jacobian will be justied in the following sections.
51
Pref +
kp s PSL
v Jacobian transfer matrix J ( s)
U ref +
ku s AVC
V V0
U f
Fig. 4.3 Closed-loop system for grid-connected VSCs using power-synchronization control
Jacobian transfer matrix: A transfer matrix represents the ac network and the electrical parts of the input devices. One exception is vector control of grid-connected VSCs, which is also formed into the Jacobian transfer matrix for mathematical convenience even though they are not electrical. In the next few sections, the Jacobian transfer matrix concept will be applied for modeling of grid-connected VSCs connected to some simplied ac-network congurations, where the VSC is the only input device of the ac system. In Chapter 6, some other input devices, such as synchronous generators and induction motors, are also modeled to create a complete linear model for an island system with several input devices.
4.2 Grid-connected VSCs using power-synchronization control

In this section, the Jacobian transfer matrix is developed for grid-connected VSCs using power-synchronization control. Fig. 4.3 shows the closed-loop system for grid-connected VSCs using power-synchronization control, i.e., the Jacobian transfer matrix is the controlled process which represents the ac circuit of the VSC and the ac network.
4.2.1 Impedance-source neglecting the ac capacitor at the lter bus

Fig. 4.4 shows the main circuit of a VSC connected to an ac network represented by a simple impedance source. Lc and Rc are the inductance and resistance of the phase reactor of the VSC, and Ln and Rn are the inductance and resistance of the ac system. Cf is the ac capacitor connected at the lter bus. The bold letter symbols E, uf , and v represent the voltage vectors of the ac source, the lter bus, and the VSC respectively. E0 , Uf 0 , and V0 are their corresponding voltage magnitudes. The ac source, which is a constant-frequency stiff voltage source, is used as the voltage reference, and the phase angles of uf and v are 52
4.2. Grid-connected VSCs using power-synchronization control

VSC
E000
Ln
Rn
P, Q in
U f 0 u 0 Rc
Lc
V0v 0 ic
+
E
+
Cf
+
v
uf
Fig. 4.4 Main circuit of a VSC connected to an impedance ac source.

P, Q E = E 0
R X = 1L I
V = V
Fig. 4.5 Equivalent circuit by neglecting the ac capacitor for angle-stability analysis.
u0 and v0 respectively. P and Q are the active and reactive powers from the VSC to the ac system. The quantity ic is the current vector of the phase reactor, and in is the current vector to the ac source. At the dc side, the VSC is assumed to connect to a stiff directvoltage source. The ac capacitor at the lter bus of the VSC is neglected in the analysis in this subsection. Such a simplication is useful for explaining some basic concepts of the proposed modeling technique in an analytical way. Especially, it is easier to demonstrate how angle and voltage stability dened in the classical power-system theory can be cast into the Jacobian transfer matrix modeling frame work. Stability analysis by the phasor theory In the following, the stability of the small power system is analyzed by the classical phasor theory. Fig. 4.5 shows the equivalent circuit of Fig. 4.4 by neglecting the ac capacitor, and V represent the phasor of the ac source and the VSC. R and X are the where E equivalent resistance and reactance between the ac source and the VSC. Neglecting the resistance, the angle stability of the equivalent system is given by the well-known powerangle equation EV P = sin . (4.1) X From (4.1) it follows that the power that can be transmitted to/from the VSC is limited by the fact that 1 sin 1. Fig. 4.6 shows the power-angle curve with different line reactances. For voltage-stability analysis, suppose that the VSC controls the reactive power to the ac system. Neglecting the resistance, the reactive power Q can be expressed by the 53

4 3.5 3 2.5 P (p.u.) 2 X = 0.6 p.u. 1.5 1 0.5 0 0 X = 1 p.u. X = 0.3 p.u.
50
100 (deg.)
150
200
Fig. 4.6 Power-angle curves with different line reactance.
following expression: V 2 EV cos . X By eliminating the angle in (4.1) and (4.2), the following equation is derived: Q= V2
2
(4.2)
2QX + E 2 V 2 + X 2 P 2 + Q2 = 0.
(4.3)
To get real solutions of V 2 in (4.3), the required condition is 2QX + E 2

2
4X 2 P 2 + Q2 0. E . 2 cos
(4.4)
If (4.1) and (4.2) are substituted back into (4.4), as V (4.5)
If the condition (4.5) is fullled, two solutions of V in (4.3) are obtained V = E2 + QX 2 E4 X 2 P 2 + XE 2 Q. 4 (4.6)
The relation in (4.6) is plotted in Fig. 4.7 where tan = Q . |P | (4.7)
These plots are commonly called P V curves, or nose curves. Besides the P V 54
tan= -0.5 1 tan= -0.2 0.8 tan= 0.4 V/E 0.6 tan= 1.5 tan= 0
0.4
0.2
0 0
0.2
0.4 |P|X/E2
0.6
0.8
Fig. 4.7 P V curves with different tan of the load.
curve, another way to determine the voltage-stability limit is by the singularity of the power-ow Jacobian matrix [74]. Jacobian matrix J is expressed as P Q = JP JPV JQ JQV
J
v
V V
(4.8)
where P , Q, v , and V are the incremental changes in the active power, the reactive power, the phase angle, and the magnitude of the bus respectively. JP , JPV , JQ , JQV are the partial derivatives of the power-ow equations with respect to the voltage angle and magnitude of the bus. For the simple ac system in Fig. 4.5, the Jacobian matrix J at the VSC bus can be derived analytically from the active-power and reactive-power expressions (4.1) and (4.2). Accordingly, JP = JPV JQ JQV P EV cos = X P EV sin = V = V X Q EV sin = = X Q 2V 2 EV cos = V = . V X
(4.9) 55

The singularity of the Jacobian matrix J is satised if its determinant equals zero, i.e., det [J] = JP JQV JPV JQ = 0 giving V = E . 2 cos (4.11) (4.10)
Comparing (4.11) to (4.5), it is easily found that the operating conditions where the Jacobian matrix becomes singular exactly correspond to the critical voltage points on the P V curves in Fig. 4.7. Stability analysis by the Jacobian transfer matrix approach In the next, the ac system and the VSC in Fig. 4.4 is modeled by the proposed Jacobian transfer matrix concept. As mentioned before, VSCs have two feasible operation modes, i.e., alternating-voltage control and reactive-power control. For the two operation modes, the process models can be written in the following input-output form: P Uf P Q JP (s) JPV (s) v
V V0
= =
J U f (s ) J U f V (s )
JPU (s)
(4.12)
JP (s) JPV (s) JQ (s) JQV (s)

JPQ (s)
v
V V0
where the denitions of P , Q, v , and V , in fact, are the same as previously dened for the power-ow Jacobian matrix (4.8), and Uf is the incremental change of the voltage magnitude of the lter bus. The transfer matrices JPU (s) and JPQ (s) are the Jacobian transfer matrices for the two operation modes. JPQ (s) has the exact input and output variables as the Jacobian matrix in (4.8). However, a fundamental difference is that Jacobian transfer matrix is a dynamic description of the ac system in all frequency range, while Jacobian matrix is only valid in the quasi-static frequency range. Appendix B.1 gives the detailed procedure for deriving the transfer functions in (4.12) by applying the space-vector theory. The six transfer functions of JPU (s) and JPQ (s) generally have the following form Jxx (s) = 56 a0 s2 + a1 s + a2 (sL + R)2 + (1 L)2 (4.13)

where the a-coefcients are summarized in Table 4.1. The k parameters are dened as k1 = V0 Uf 0 cos (v0 u0 ) k2 = V0 Uf 0 sin (v0 u0 ) k3 = E0 V0 cos v0 k4 = E0 V0 sin v0 .
Table 4.1 Coefcients of the transfer functions in the Jacobian transfer matrices.
(4.14)
a0 JP (s) a1 a2 a0 J Q (s ) a1 a2 a0 J U f (s ) a1 a2 a0 JPV (s) a1 a2 a0 JQV (s) a1 a2 a0 J U f V (s ) a1 a2
L 1 R 1
(k 3 k 1 ) (k3 k1 ) Lk2 (k 4 k 2 ) (k4 k2 ) Lk1
1 Lk3 Rk4
L 1 R 1
1 L (k4 2k2 ) + R (k3 2k1 ) LLn k2 /Uf 0 RLn k2 /Uf 0

L 1 R 1 2 1 LLn k2 /Uf 0 RLn k1 /Uf 0
(k 4 k 2 ) (k4 k2 ) + Lk1 (k 1 k 3 ) (k1 k3 ) Lk2
1 Lk4 + Rk3
L 1 R 1
1 L (2k1 k3 ) + R (k4 2k2 ) LLn k1 /Uf 0 RLn k1 /Uf 0

2 1 LLn k1 /Uf 0 RLn k2 /Uf 0
It should be noted that the output variables of the transfer functions derived for the Jacobian transfer matrices are the active power, the reactive power and the voltage magnitude at the lter bus. To compare with the results from the phasor analysis, it is necessary to substitute Uf 0 = V0 and u0 = v0 into the coefcients in Table 4.1, i.e., the condition for the phasor analysis is considered as a special case that the lter bus is identical to the VSC bus. 57

By substituting Uf 0 = V0 and u0 = v0 into the transfer functions of JPQ (s) , and further assuming R = 0 and s = 0, the four transfer functions JP (s) , JPV (s) , JQ (s), and JQV (s), are identical to the four elements of the Jacobian matrix J dened in (4.8). In other words, the power-ow Jacobian matrix is the static form of the Jacobian transfer matrix JPQ (s) that is valid in the quasi-static frequency range. Eq. (4.13) shows that all the six transfer functions have a pair of resonant (complex) poles R p1,2 = j1 . (4.15) L and two zeros. The resonant poles will be further discussed in the next subsection. In this subsection, the transmission zeros of JPU (s) and JPQ (s) are analyzed, since their locations have close relationships with angle and voltage stability dened by the phasor approach. The computation of the transmission zeros for a general multi-input multi-output (MIMO) process is a complex issue [82]. However, for a square matrix, the transmission zeros can be obtained simply by its determinant, i.e., the transmission zeros of JPU (s) are the values of s = z that satisfy det [JPU (s)] = JP (s) JUf V (s) JPV (s) JUf (s) = 0. If the resistance R is neglected, the solutions to (4.16) are given by z1,2 = 1 E0 cos u0 . Uf 0 E0 cos u0 (4.17) (4.16)
Graphically, the locations of the transmission zeros of JPU (s) can be divided by the following borders: The border where JPU (s) gets zeros at the origin. This is equivalent to E0 cos u0 = 0 giving u0 = 90 . The border where JPU (s) gets zeros at innity. This is equivalent to Uf 0 E0 cos u0 = 0 giving u0 = arccos Uf 0 E0 . (4.20) (4.19) (4.18)
The border where JPU (s) gets zeros at 1 . This is equivalent to E0 cos u0 =1 Uf 0 E0 cos u0 58 (4.21)

100 80 Origin border 60 40 (deg.) 20 0 -20 -40 -60 Origin border -80 -100 0 0.5 1 U /E
f0
Infinity border
1 border
u0
1.5
Fig. 4.8 Locations of the transmission zeros of JPU (s) .
giving v0 = arccos Uf 0 2E0 . (4.22)
The 1 border gives an idea about how much the zeros limit the achievable bandwidth of the control system, even though it is not a real border. From Fig. 4.8, it is easily observed that the transmission zeros of JPU (s) depend very much on the operating points. Both Uf 0 /E0 and u0 affect their locations. However, within moderate voltage levels, e.g., 0.8 Uf 0 /E0 1.2, they are affected mainly by u0 . The tendency is that, with higher u0 , the zeros get closer to the origin border. In control theory, a process which has right-half plane (RHP) zeros is called non-minimum-phase system. From a feedback-control perspective, the RHP zero of the process causes an additional time delay, which imposes a fundamental limitation on the achievable bandwidth of the control loop [82], i.e., the closed-loop system cannot achieve a higher bandwidth than the location of the RHP zero. When the RHP zero reaches the origin, i.e., u0 = 90 , it causes 180 phase shift even in the steady state, which means that tight control at low frequencies is not possible [82]. If u0 is replaced by v0 in (4.18), it is easily found that the operating conditions where the transmission zeros of the Jacobian transfer matrix JPU (s) reach the origin exactly correspond to the angle-stability limit dened by the phasor approach. From the above analysis, it can be clearly seen that the ac system is, in fact, a non-minimum-phase system, where the angle-stability limitation is a consequence of the non-minimum-phase effect. How can this phenomenon be interpreted physically? A brief 59

Imaginary
V'
jL I '
U'f
Instant V j L I
Uf
I' I
Slow
E Real
Fig. 4.9 Non-minimum-phase phenomena hidden by the phasor diagram.
analysis of the physical mechanism is given below. Fig. 4.9 shows the phasor diagram of the ac system, where the resistance is neglected. Now, suppose that the voltage vector of the VSC rotates with the phase angle v and its magnitude is kept constant during the angle rotation. The current magnitude | naturally increases to |I | in response to such angular rotation, but the voltage magni|I | < |U f |. This can be easily observed from tude along the ac line drops instead, e.g., |U f the phasor diagram but it can also be proven by the transfer function JUf (s) in (B.28). However, the phasor diagram hides one important fact, i.e., the voltage change is abrupt, but the current does not increase instantly due to the inductance of the ac line. Similar to JPU (s), the transmission zeros of JPQ (s) are the values of s = z that satisfy det [JPQ (s)] = JP (s) JQV (s) JPV (s) JQ (s) = 0. If the resistance R is neglected, the solutions to (4.23) are given by z1,2 = 1
2 2E0 Uf 0 cos u0 E0 . 2 2 E0 + Uf 0 2E0 Uf 0 cos u0
(4.23)
(4.24)
The locations of the transmission zeros of JPQ (s) can also be divided by the following borders: The border where JPQ (s) gets transmission zeros at the origin. This is equivalent to
2 2E0 Uf 0 cos u0 E0 =0
(4.25)
giving cos u0 = E0 . 2Uf 0 (4.26)
60

The border where JPQ (s) gets transmission zeros at 1 . This is equivalent to
2 2E0 Uf 0 cos u0 E0 =1 2 2 E0 + Uf 0 2E0 Uf 0 cos u0
(4.27)
giving u0 = arccos E0 Uf 0 + 2Uf 0 4E0 . (4.28)
The 1 border gives an idea about how much the transmission zeros limit the achievable bandwidth of the VSC controller. From Fig. 4.10, it is easily observed that the locations of the transmission zeros of JPQ (s) also depend on the operating points, in the same way as for JPU (s). But besides the dependency on the phase angle, the locations of the transmission zeros of JPQ (s) also strongly depend on the voltage magnitude of the lter bus. The tendency is that the zeros are closer to the origin border with higher u0 and lower Uf 0 /E0 . For the special condition where the lter bus is identical to the VSC bus, the transmission zeros of JPQ (s) become s = 1
2 2E0 V0 cos v0 E0 . 2 E0 + V02 2E0 V0 cos v0
(4.29)
Accordingly, the origin border of JPQ (s) becomes cos v0 = E0 . 2 V0 (4.30)
Comparing (4.30) with (4.5) and (4.11), it is easily found that the operating conditions where the transmission zeros of JPQ (s) reach the origin exactly correspond to the voltage-stability limit dened by the phasor approach. By comparing Fig. 4.8 with Fig. 4.10, it can be found that within moderate voltage levels, the transmission zeros of JPQ (s) are closer to the origin than the transmission zeros of JPU (s), which implies that the achievable bandwidth of the VSC is higher in alternating-voltage control mode than in reactive-power control mode. To interpret the physical mechanism of voltage stability, the phasor diagram in Fig. 4.9 can still be used. Different from in alternating-voltage control mode, in reactivepower control mode, the voltage magnitude of the VSC is no longer held constant during the phase-angle rotation. On the contrary, in order to keep the reactive power constant, the VSC decreases the voltage magnitude following the phase-angle rotation. This behavior obviously makes the non-minimum-phase effect worse, and the effect also becomes more dependent on the VSC voltage magnitude. Remark 1: The above analysis clearly shows the close relationships of the transmission zeros of the Jacobian transfer matrices with angle and voltage stability in power 61
80 60 40 20 u0 (deg.) 0 -20 -40 -60 -80 0 0.5 1 1.5 Uf0/E0 2 2.5 3 3.5 1 border Origin border
Fig. 4.10 Locations of the transmission zeros of JPQ (s).
systems. It also explains why the singularity of the power-ow Jacobian matrix can be used as a criterion to determine voltage instability in power systems. But, strictly speaking, the only correct criterion for power-system instability is the poles of the closed-loop system. A Jacobian transfer matrix having transmission zeros at the origin only indicates that the process is impossible to be tightly controlled at low frequencies. Instability, however, can only be caused by the feedback control. Remark 2: For the particular case in this section, the physical reason to the nonminimum phase phenomena are explained. However, non-minimum phenomena cannot be generalized to the whole power system as a reason to angle and voltage stability, since zeros are input dependent. For instance, in Chapter 6, it will be shown that the Jacobian transfer matrix for induction motors only has a left-half plane (LHP) real zero moving towards the origin as the slip of the induction motor increases. Remark 3: For the particular case in this section, it has been shown that the powerow Jacobian matrix is the static form of the Jacobian transfer matrix. However, this cannot be generalized either. One reason is that the Jacobian transfer matrix includes the ac network as well as the electrical parts of the input devices, while the power-ow Jacobian matrix normally only includes the ac network. Another reason is that, as it will be shown in later chapters, the outputs of the controllers of different input devices are versatile, while the power-ow Jacobian matrix only has voltage magnitude and phase angle as inputs. Remark 4: Based on the analysis in this section, it can be concluded that reactivepower control basically is not a suitable operation mode for VSC-HVDC links connected 62

to weak ac systems. Therefore, in the later sections, only the transfer matrix JPU (s) will be derived, and with a short name J (s).
4.2.2 Impedance-source including the ac capacitor at the lter bus

In this section, the Jacobian transfer matrix will be derived for the main circuit shown in Fig. 4.4 by including the ac capacitor at the lter bus. If the high-pass current control HHP (s) is disregarded, the voltage-vector control law of the VSC in (3.32) becomes
c vref = V0 + V.
(4.31)
In a synchronous grid dq reference frame with the d axis chosen aligned with the ac source c E, if the switching-time delay is neglected and it is assumed that |vref | does not exceed the maximum voltage modulus, then
c v = vref ejv = (V0 + V )ejv .
(4.32)
With the voltage vector v expressed by (4.32), the dynamic equations of the main circuit in Fig. 4.4 can be written as dic = (V0 + V )ejv uf Rc ic j1 Lc ic dt duf Cf = ic in j1 Cf uf dt din Ln = uf E Rn in j1 Ln in dt Lc and in dq-component form dicd dt dicq Lc dt duf d Cf dt duf q Cf dt dind Ln dt dinq Ln dt Lc = (V0 + V ) cos v uf d Rc icd + 1 Lc icq = (V0 + V ) sin v uf q Rc icq 1 Lc icd = icd ind + 1 Cf uf q = icq inq 1 Cf uf d = uf d E0 Rn ind + 1 Ln inq = uf q Rn inq 1 Ln ind . (4.36) (4.33) (4.34) (4.35)
The output variables are the active power P and the voltage magnitude Uf at the lter bus. In per unit form these two quantities are dened as P = Re {uf i n } , Uf =
2 u2 f d + uf q .
(4.37) 63

The state-space model can be obtained by linearizing (4.36) and (4.37), which yields the following form d x = Ax + B u dt y = C x + Du where 1 1 Rc 0 L 0 0 Lc c 1 1 0 0 1 C 0 Cf f A= 1 1 1 0 0 C 0 Cf f 1 R n 0 0 0 Ln 1 Ln 1 n 0 0 0 1 R Ln Ln B=
v 0 V0 sin Lc V0 cos v0 Lc V0 cos v0 Lc V0 sin v0 Lc
(4.38)
c R Lc
1 L c
0 0 0 0 0 0 0 0
C=
0 0 ind0 0 0 0 0 0 0
u f d0 Uf 0
inq0 uf d0 uf q0
uf q 0 Uf 0
0 v
V V0
0 , y= P Uf
T
D= x=
, u=
icd icq uf d uf q ind inq
(4.39)
The state-space representation (4.38) can also be written in input-output transfer matrix form y = C (sI A)1 B + D u which yields P Uf JP (s) JPV (s)
V V0
(4.40)
J U f (s ) J U f V (s )
J(s)
(4.41)
Due to the order of the system, the analytical expressions of the transfer functions of the Jacobian transfer matrix J (s) in (4.41) are difcult to obtain. However, the analytical expressions of the transmission zeros and poles are possible to be derived from the statespace representation. 64

The transmission zeros of J (s) can be derived by the QZ method suggested in [83, 84]. A polynomial system matrix P (s) is dened as P (s ) = A sI B C D . (4.42)
The transmission zeros are then the values s = z for which P (s) loses rank, resulting in zero output for some non-zero input values, i.e., the zeros are found by the values s = z that satisfy det[P (s)] = 0. (4.43) By neglecting the resistances Rc and Rn , the solutions to (4.43) are found to be exactly identical to those of (4.17). In other words, the ac capacitor at the lter bus has no inuence on the transmission zeros of the Jacobian transfer matrix! As will be explained below, this somewhat surprising result is reasonable. In Section 4.2.1, it was shown that the transmission zeros of the Jacobian transfer matrix have a close relationship with angle stability of the ac system. As long as the lter-bus voltage is controlled, the angle stability of the ac system is indeed only determined by the phase angle between the ac source and the lter bus, and independent of the ac capacitor. Of course, in practice, the ac capacitor will affect how much reactive power the VSC needs to supply in order to keep the lter-bus voltage constant, but this will not affect the stability unless the converter is forced into voltage limitation. The poles of J (s) are the eigenvalues of the A matrix. If the resistances Rc and Rn are neglected also in this case, the poles can be solved analytically with the expressions p1,2 = j1 p3,4 = j p5,6 = j 1 1 + 1 Ln Cf Lc Cf 1 1 + + 1 Ln Cf Lc Cf . (4.44)
The two pairs of poles p3,4 and p5,6 are related to the ac capacitor connected at the lter bus. However, the pole pair p1,2 at the grid frequency is not related to any ac capacitor. In Fig. 3.3, the resonance at the grid frequency has already been observed during the power step response of the VSC-HVDC link using power-angle control. In contrast to the transmission zeros, the poles of the Jacobian transfer matrix are independent of the operating points, but are usually very poorly damped due to the low resistance in transmission systems. For grid-connected VSCs, the control system has to provide damping to these poles to achieve a reasonable bandwidth. In the following, it is demonstrated how these resonant poles are damped by the high-pass current control HHP (s) proposed for power-synchronization control in Chapter 3. 65

By including HHP (s), the voltage vector v of the VSC is expressed as
c v = vref ejv jv = [(V0 + V ) HHP(s)ic c ]e
= (V0 + V )ejv HHP (s)ic . Thus, the dynamic equation in (4.33) is modied as Lc
(4.45)
dic = (V0 + V )ejv HHP (s) ic uf Rc ic j1 Lc ic . dt
(4.46)
To eliminate the Laplace transform variable s in HHP (s) which in (4.46) should be interpreted as s = d/dt, a new state variable c needs to be introduced. With the new state variable c , the dynamic equation of the phase reactor in (4.46) is expressed as dic = Lc c + (V0 + V )ejv (Rc + kv + v Lc ) ic uf j1 Lc ic dt dc Lc = v (V0 + V )ejv Rc ic uf j1 Lc ic . (4.47) dt Lc Replacing the dynamic equations of the phase-reactor in (4.33) by (4.47) and following the same procedure, the Jacobian transfer matrix J (s) including HHP (s) can be obtained. For mathematical convenience, HHP (s) has been treated as a part of the Jacobian transfer matrix, i.e., the controlled process, even though it physically belongs to the control system. Fig. 4.11 shows the pole-zero map of J (s) with a variation of the gain kv of the high pass-current control. The system parameters of the Jacobian transfer matrix are given in Table 4.2. As shown in Fig. 4.11, the effect of HHP (s) is to shift the resonant poles of J (s) towards the left-half plane (LHP). This damping effect is general for any resonances in the ac system since HHP (s) basically emulates the dynamic behavior of a physical resistor. From Fig. 4.11 it can also be observed that the two transmission zeros are not affected by HHP (s). This property, in fact, can also be analytically veried by applying the QZ method for zero calculation. Figs. 4.12-4.15 show the Bode plots of the four transfer functions JP (s), JPV (s), JUf (s), and JUf V (s) overlapped with plots produced by frequency-scanning results from PSCAD/EMTDC (dashed lines). The transfer functions with and without HHP (s) are placed side by side in the gures. It clearly shows the damping effects of HHP (s) to all the resonant poles for all the transfer functions. The resonant poles p3,4 and p5,6 caused by the ac capacitor at the lter bus appear in all the four transfer functions. However, it seems that the resonant pole pair p1,2 at the grid frequency only affects JP (s) and JPV (s), but not JUf (s) and JUf V (s). The reason is that, by neglecting the resistances Rn and Rc , JUf (s) and JUf V (s) get exactly pole-zero cancelation at the grid frequency. 66
2000 1500 1000 (rad/sec) 500 0 -500 -1000 -1500 -2000 z1
p5 p3 p1
z2
p2
p4 p6 -2000 -1500 -1000 -500 0 500 (rad/sec) 1000 1500 2000
Fig. 4.11 Damping effects of HHP (s) on the resonant poles of J (s). Variations of the gain kv from 0.0 p.u. to 0.6 p.u.
Table 4.2 Parameters of the Jacobian transfer matrix for a grid-connected VSC using powersynchronization control connected to an impedance-source. Per unit based on 350 MVA and 195 kV.
Parameters Rc 1 Lc Main-circuit parameters 1 Cf Rn 1 Ln E0 Initial conditions V0 u0 (v0 ) kv High-pass current control v
Value 0.01 p.u. 0.2 p.u. 0.17 p.u. 0.01 p.u. 1.0 p.u. 1.0 p.u. 1.0 p.u. 30 (35.8 ) 0.45 p.u. 40 rad/s
67
10 |JP(j)|
10 |JP(j)|
1 2 3
10
10
10 ARG J (j) (deg.)
-2
10
10
10
10 ARG J (j) (deg.)
-2
10
10
10
0 -200 -400 -600 -800
10
10 (rad/sec)
10
0 -200 -400 -600 -800
10
10 (rad/sec)
10
(a) Without HHP (s).
(b) With HHP (s).
Fig. 4.12 Bode plots of JP (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
10 |JPV(j)|
10 |JPV(j)|
1 2 3
10
10
10 (j) (deg.)
-2
10
10
10
10 (j) (deg.)
-2
10
10
10
100 0 -100 -200 -300 10

1
100 0 -100 -200 -300 10

1
PV
ARG J
10 (rad/sec)
10
ARG J
PV
10 (rad/sec)
10
(b) With HHP (s).
Fig. 4.13 Bode plots of JPV (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
68
10 |JU (j)|
10 |JU (j)|
1 2 3
10
10
10 ARG J (j) (deg.) U f
-2
10
10
10
10 ARG J (j) (deg.) U f
-2
10
10
10
-200 -400 -600 10

1
-200 -400 -600 10

1
10 (rad/sec)
10
10 (rad/sec)
10
(b) With HHP (s).
Fig. 4.14 Bode plots of JUf (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
10 |JU V(j)|
10 |JU V(j)|
1 2 3
10 ARG J V(j) (deg.) U f
-2
10
10
10 0
10
10
10 ARG J V(j) (deg.) U f
-2
10 0
10
10
-100 -200 -300 10

1
-100 -200 -300 10

1
10 (rad/sec)
10
10 (rad/sec)
10
(b) With HHP (s).
Fig. 4.15 Bode plots of JUf V (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
69

P, Q E000 Cn
un Ln
VSC
Lc
Rn
in
U f 0u 0 Rc
V0 v 0
+
E
+
uf ic
+
v
Fig. 4.16 AC-source feeding from a series-compensated ac line.
This property can also be conrmed by the analytical expressions of JUf (s) and JUf V (s) in Table 4.1 for the network conguration without considering the ac capacitor at the lter bus. Substituting R = 0 into the transfer functions it is found that JUf (s) = JUf V (s) = Ln V0 sin (v0 u0 ) L (4.48)
Ln V0 cos (v0 u0 ) L
i.e., the poles are canceled by the zeros.
4.2.3 AC-source feeding from a series-compensated ac line

Fig. 4.16 shows a network conguration with an ac source feeding from a series-compensated transmission line, where Cn is the capacitance of the series ac capacitor, while un is the voltage vector across the series ac capacitor. The ac capacitor at the lter bus of the VSC is neglected in this conguration in order to obtain analytical expressions of the transmission zeros and poles of the Jacobian transfer matrix. In a synchronous dq reference frame where the d axis is chosen aligned with the ac source, the dynamic equations of the main circuit in Fig. 4.16 can be expressed as di = v un E Ri j1 Li dt dun Cn = i j1 Cn un dt L (4.49) where L = Lc + Ln , R = Rc + Rn , and i = ic = in . In component form, (4.49) can be 70

written as did dt diq L dt dund Cn dt dunq Cn dt L = vd und E0 Rid + 1 Liq = vq unq Riq 1 Lid = id + 1 Cn unq = iq 1 Cn und . (4.50) Since the ac capacitor at the lter bus is neglected, the lter-bus voltage vector uf is neither a state variable nor an input. Such a variable has to be represented by other states and/or inputs. For this network conguration, uf can be solved from the following two equations di = uf un Rn i j1 Ln i (4.51) dt di Lc = v uf Rc i j1 Lc i. (4.52) dt By dividing (4.51) with Ln and (4.52) with Lc , and further subtracting (4.52) from (4.51), it is possible to express uf as Ln uf = Ln Lc v + Rx i + un . L L (4.53)
where Rx = (Rn Lc Rc Ln )/L. Following the same procedure as the previous section, the state-space model in the form of (4.38) can be obtained by linearizing (4.50) and (4.37) with uf d and uf q derived from the linearized component form of (4.53). The corresponding matrices and vectors in (4.38) are expressed as R 1 L 1 L 0 T V0 cos v0 v 0 1 1 R 0 L V0 sin 0 0 L Lc Lc A= , B = V0 cos v0 V0 sin v0 1 0 0 0 0 1 Lc Lc Cn C= D= x= 0
1 Cn
id0 Rx + uf d0 iq0 Rx + uf q0
uf d0 Rx Uf 0 Ln V0 (iq0 L Ln V0 (u f q 0 LUf 0 uf q0 Rx Uf 0
id0 Lc L
iq0 Lc L
uf d0 Lc Uf 0 L
uf q0 Lc Uf 0 L Ln V0 (id0 L
cos v0 id0 sin v0 ) cos v0 uf d0 sin v0 )

T
cos v0 + iq0 sin v0 ) cos v0 + uf q0 sin v0 ) (4.54) 71
Ln V0 (uf d0 LUf 0
id iq und unq

The denitions of u and y are the same as in (4.39). Similarly, the Jacobian transfer matrix J(s) can be obtained from the state-space representation by applying (4.40). The poles of J(s) are the eigenvalues of the A matrix. These are given by p1,2 = p3,4 = R 1 2L 2 1 R 2L 2 R2 4 2 41 +4 2 L LCn R2 4 2 41 4 2 L LCn
2 2 41 R 2 1 LCn L2 2 2 41 R 2 1 . LCn L2
(4.55)
If the resistance R is neglected, (4.55) can be simplied as p1,2 = j 1 p3,4 1 LCn 1 = j 1 + LCn
(4.56)
For series compensation, the reactance of the series capacitor is always smaller than the reactance of the ac line. Consequently, 1/ LCn is always smaller than 1 . Thus, (4.56) shows that J(s) has two pairs of poles, one pair (p1,2 ) in the subsynchronous frequency range, and the other pair (p3,4 ) in the supersynchronous frequency range. If Cn = , these two pairs of poles correspond to the pole pair at the grid frequency for the impedancesource conguration in the previous subsection. The smaller the ac capacitance, i.e., the higher degree that the ac line is compensated, the further the pole pairs are separated. The pole pair p1,2 in the subsynchronous frequency range is usually troublesome. Besides the SSR problem with the rotor shaft of the thermal plant as mentioned before, they can also create problems for power-electronic devices connected in the vicinity since the subsynchronous frequency range is also where the control systems of most powerelectronic devices are active. If the resistances Rc and Rn are neglected, the transmission zeros of J(s) can be derived analytically by applying the QZ method, which gives z1,2 = z3,4 = 2 Uf 0 (1 Ln Cn 1) + b 1 2 1 + 2Ln Cn (Uf 0 E0 cos u0 ) Ln Cn 2 Uf 0 (1 Ln Cn 1) b 1 2 1 + 2Ln Cn (Uf 0 E0 cos u0 ) Ln Cn
(4.57)
where the term b is expressed as

2 2 2 2 b = 8E0 cos u0 1 Ln Cn (2E0 cos u0 3Uf 0 ) Uf 0 + Uf 0 (1 Ln Cn + 3) .
(4.58)
Fig. 4.17 shows the loci of the two pairs of zeros as the load angle u0 is increased. For the Jacobian transfer matrix, the main-circuit parameters are chosen as given by Table 4.2 72
600 400 200 0 -200 -400 z1 u0=32
z3 p1 p z2 u0=54 u0=41 z4

(rad/sec)
u0=41
u0=54
u0=32
p -600 -600 -400 -200
0 (rad/sec)
200
400
600
Fig. 4.17 Loci of the transmission zeros with increased load angles.
but with a 60% degree of series compensation, i.e., the capacitance of the series capacitor is chosen as 1 Cn = 1.667 p.u. The zero pair z1,2 is on the real axis, while z3,4 is on the imaginary axis. With increased load angles, the resonant zero pair z3,4 moves from p1,2 towards p3,4 . The real zero pair z1,2 moves towards the origin. If u0 = 90 is substituted into the analytical expression of z1,2 in (4.57), it precisely gives z1,2 = 0. It can also be analytically veried that, if Cn = is substituted into (4.57), then the mathematical expression of z1,2 is identical to (4.17), while z3,4 = . In contrast to the resonant poles, for which HHP (s) provides signicant damping, HHP (s) has no effect on the resonant zero pair z3,4 , which might negatively affect the phase margin of the control system around the subsynchronous frequency range. Care must be taken if the bandwidth of the control system of the VSC is intended to be higher than the frequency of z3,4 . Fig. 4.18 shows the Bode plot of the open-loop transfer function of the power-synchronization loop, which shows the effect of HHP (s). The open-loop transfer function of the power-synchronization loop is expressed as HPSL (s) = JP (s) kp . s (4.59)
During the process of describing the modeling of VSCs using power-synchronization control with the concept of Jacobian transfer matrix, simplied ac-network congurations have been used. This has been done in order to develop a basic understanding of the Jacobian transfer matrix modeling concept, as well as to understand the physical meanings 73

10
2
|HPSL(j)|
10
10
-2
Without HHP(s) With HHP(s)
10
-4
10 0 arg HPSL(j) (deg.)
10
10
-100
-200
-300 10
10 (rad/sec)
10
Fig. 4.18 Bode plot of the open-loop transfer function HPSL (s), kp = 130 rad/s. Initial conditions: u0 = 24.9 (v0 = 37 ).
of the transmission zeros and poles of the Jacobian transfer matrix in an analytical way. Within the investigated ac-network congurations so far, a constant-frequency ac voltage source with a voltage vector E is assumed. The d axis of the grid dq frame has been chosen aligned with the voltage vector for analysis. Such voltage sources, of course, do not exist in the real system. However, if the VSC is connected to a large ac system, assuming such an equivalent network conguration is acceptable. In case that the VSC is connected to an island system, a common ac-network R I frame needs to be dened which is not related to any voltage sources. Such issues will be discussed in Chapter 6.
4.3 Grid-connected VSCs using vector current control

In Chapter 3, the Jacobian transfer matrix was developed for a grid-connected VSC using vector current control. However, the model was based on the assumption that the lter-bus voltage is stiff, whereas such assumption is not valid if the VSC is connected to a weak ac system. In this section, the Jacobian transfer matrix is developed taking into account the dynamics of the lter bus, as well as the dynamics of the PLL. The methodology is independent of ac-network congurations, but for simplicity the impedance-source network conguration shown in Fig. 4.4 is used for model development. Fig. 4.19 shows the small-signal block diagram of vector current control, where the Jacobian transfer matrix is dened as the transfer matrix which has the current reference ref iref d and iq as inputs, and the active power P and the the lter-bus voltage Uf as 74
4.3. Grid-connected VSCs using vector current control

Jacobian transfer matrix J ( s)
c cd
ref id
ref iq
c icq
v G cc ( s)
c dref
v Mv PLL
ref d
u c fd u
c fq
c vqref
ref vq
J n ( s)
P U f icd icq u fd u fq
u fd PLL G PLL ( s) u fq
u c fd u c fq Mu PLL
c icd
u fd u fq
icd Mi PLL icq
c icq
Fig. 4.19 Denition of the Jacobian transfer matrix for a grid-connected VSC using vector current control.
outputs.2 By dening the Jacobian transfer matrix, Fig. 4.20 shows the closed-loop system of a grid-connected VSC using vector current control, where the minus sign in the alternating-voltage control is due to the direction of the phase-reactor current reference. In the following, the mathematical expressions of the three transfer matrices Jn (s), GPLL (s), Gcc (s), and the three real matrices Mv , Mu , and Mi shown in Fig. 4.19 are described, where the superscript c on the variables represents the converter dq frame. 1. Network Jacobian transfer matrix Jn (s). This is the ac-network Jacobian transfer matrix which has the dq components of the VSC voltage vd and vq as the inputs, and the active power P , the lter-bus voltage magnitude Uf , dq components of the phase-reactor current icd and icq , and the dq components of the lter-bus voltage uf d and uf q as the outputs. To distinguish it from the nal Jacobian transfer matrix for vector current control, the subscript n is used. Jn (s)
The procedure of model development is the same if reactive power Q is used as output instead of Uf , but the alternating-voltage control mode is focused on as it is the preferred operation mode for weakac-system connections.
2
75
Pref +
KP K + i s APC
P p
ref id
P Jacobian transfer matrix
U ref +
(KU p + -
K iU ) s AVC
ref iq
J ( s)
U f
Fig. 4.20 Closed-loop system for VSCs using vector current control.
is similar to the Jacobian transfer matrix developed for power-synchronization control in Section 4.2.2, but differs by the inputs and the outputs. Those matrices in the state-space representation that are different from (4.39) are 0 0 ind0 inq0 uf d0 uf q0 0 0 u f d0 u f q 0 0 0 Uf 0 Uf 0 T 1 0 0 0 0 0 1 0 0 0 0 0 B= , C= 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 u= vd vq
T
, y=
P Uf icd icq uf d uf q
. (4.60)
The network Jacobian transfer matrix Jn (s) thus can be obtained from (4.40), i.e., y = Jn (s)u. (4.61)
2. Phase-locked loop (PLL) transfer matrix GPLL (s). As described in Chapter 3, the control law of the PLL is given by PLL =
PLL Kp +
KiPLL s
c Im{uc f } = FPLL (s)uf q .
(4.62)
FPLL (s)
The lter-bus voltage uc f in the converter dq frame is related to uf in the grid dq frame by jPLL uc (4.63) f = uf e 76

or in component form
c uc f d + juf q = (uf d + juf q )(cos PLL j sin PLL )
(4.64)
which yields uc f q = uf q cos PLL uf d sin PLL . Substituting (4.65) into (4.62), yields PLL = FPLL (s)(uf q cos PLL uf d sin PLL ) which is further linearized as PLL = FPLL (s) sin u0 uf d 1 + FPLL (s)(uf d0 cos u0 + uf q0 sin u0 )
APLL (s)
(4.65)
(4.66)
FPLL (s) cos u0 uf q 1 + FPLL (s)(uf d0 cos u0 + uf q0 sin u0 )

BPLL (s)
(4.67)
where u0 is the angle output of the PLL in the steady state, which corresponds to the phase angle of the lter-bus voltage in the grid dq frame. Eq. (4.67) can be further expressed in transfer matrix form as uf d uf q
PLL =
APLL (s) BPLL (s)

GPLL (s)
(4.68)
3. Frame-transformation matrices Mv , Mu , and Mi . The VSC voltage reference vref , the lter-bus voltage uf , and the phase-reactor current ic , are related in the converter dq frame and the grid dq frame as
c jPLL jPLL vref = vref ejPLL , uc , ic . f = uf e c = ic e
(4.69)
By writing (4.69) in component form and applying linearization, (4.69) can be ex77

pressed in input-output matrix form as
c vdref
ref vd ref vq
c c cos u0 sin u0 (vd 0 sin u0 vq 0 cos u0 )
sin u0
cos u0
uc fd uc fq
cos u0
sin u0
sin u0 cos u0 cos u0 sin u0
ic cd ic cq
sin u0 cos u0
v c qref c c (vd 0 cos u0 vq 0 sin u0 ) PLL Mv uf d (uf d0 sin u0 + uf q0 cos u0 ) uf q (uf d0 cos u0 uf q0 sin u0 ) PLL Mu icd (icd0 sin u0 + icq0 cos u0 ) icq . (icd0 cos u0 icq0 sin u0 ) PLL
Mi
(4.70)
4. Current-control transfer matrix Gcc (s). The current control law is given by (3.12). After writing (3.12) in component form and applying linearization, the current controller can be expressed in input-output transfer matrix form as
c vdref c vqref
c Lc 0
0 c Lc
c Lc 1 Lc HLP (s) 1 Lc c Lc
Gcc (s)
0 HLP (s)
iref q ic cd . ic cq uc fd uc fq (4.71)
iref d
If the switching-time delay of the converter is neglected and it is assumed that |vref | does not exceed the maximum voltage modulus, then v = vref . The above derived transfer/real matrices can be interconnected by the block diagram shown in Fig. 4.19. Accordingly, the Jacobian transfer matrix for vector current control is obtained as 78 P Uf JPId (s) JPIq (s) iref d iref q
JUf Id (s) JUf Iq (s)

J(s)
(4.72)
2000 1500 1000 (rad/sec) 500 0 -500 -1000 -1500 -2000 z1
p3
p1
z2
p2
p4 p6 -2000 -1500 -1000 -500 0 500 (rad/sec) 1000 1500 2000
Fig. 4.21 Damping effects of the current controller on the resonant poles of J (s) for vector current control. Variations of c from 1200 rad/s to 2500 rad/s.
Table 4.3 Control parameters of the Jacobian transfer matrix for a grid-connected VSC using vector current control connected to an impedance source.
Parameters Inner-current control c f PLL

PLL Kp
Value 2500 rad/s 80 rad/s 20 20
KiPLL
Similar to the HHP(s) function of power-synchronization control, the current controller of vector current control also provides damping to the resonant poles of the ac network, as shown in Fig. 4.21. The main-circuit parameters and initial conditions of the Jacobian transfer matrix for vector current control are chosen the same as those given in Table 4.2 for power-synchronization control, and the control parameters of the inner current control and the PLL are given in Table 4.3. The higher the bandwidth of the current controller, the more damping it adds. For a modern transistor PWM converter with a switching frequency of 1 2 kHz, the bandwidth of the current controller can be chosen as c = 1000 2500 rad/s [51], which is enough for damping purpose. When analyzing the transmission zeros of the Jacobian transfer matrix, an inter79

0
|JPI (j)|
10 (j) (deg.)
-1
|JPI (j)|
1 2 3
10
10
q
10
10
10
10 (j) (deg.)
-2
10
10
10
0
d
-200 -400 -600 10

1
q
PI
ARG J
-200
10
10 (rad/sec)
10
ARG J
PI
-100
10 (rad/sec)
10
(a) JPId (s).

|JU I (j)|
0
(b) JPIq (s).

|JU I (j)| 10
0
10
fd
fq
10 (j) (deg.)
-1
10
10
10
10 (j) (deg.)
-2
10
10
10
-200
fd
-200
fq
UI
ARG J
ARG J
UI
-400 -600
-400 -600
10
10 (rad/sec)
10
10
10 (rad/sec)
10
(c) JUf Id (s).
(d) JUf Iq (s).
Fig. 4.22 Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC). Parameters and initial conditions are given in Table 4.2 and Table 4.3.
esting observation can be made. With the same main-circuit parameters and initial conditions, by comparing Fig. 4.21 with Fig. 4.11, it can be easily found that the locations of the transmission zeros of the Jacobian transfer matrix for vector current control are exactly identical to those for power-synchronization control! Thus, from the fundamental limitation point of view, i.e., the bandwidth limitation by the RHP zero, there is no difference between vector current control and power-synchronization control. Fig. 4.22 shows the Bode plots of the transfer functions of the Jacobian transfer matrix with overlapped frequency-scanning plots from PSCAD/EMTDC. The two plots generally show good agreement for all the transfer functions. 80
4.4. Jacobian transfer matrix for dc-system modeling

Pac2
VSC2
Pdc2 Pdc3
VSC3
Pac3
VSC1
Pac1
Pdc1
Pac4
VSC4
Pdc4
DC-Jacobian transfer matrix G dc ( s ) Pdc5
VSC5
Pac5
Fig. 4.23 A dc network connected to multiple VSCs.

u
ref dc
r + K dc ( s ) +
K ac ( s )
-
Pac = Pdc J ( s)
y
G dc ( s )
u dc
AC-Jacobian transfer matrix
DC-Jacobian transfer matrix
Fig. 4.24 Feedback-control view of the ac/dc combined system.
4.4 Jacobian transfer matrix for dc-system modeling

The concept of Jacobian transfer matrix for dc-system modeling is similar to ac-system modeling. As shown in Fig. 4.23, a dc system is also treated as a multivariable feedback control system where the Jacobian transfer matrix Gdc (s) is dened as the controlled process. In this thesis, only VSCs are considered as input devices to the dc system. The Jacobian transfer matrix of a dc system thus represents the passive components in the dc system. The input variables to Gdc (s) are the dc powers from the VSCs, while the output variables from Gdc (s) are the direct voltages at each VSC terminal. Fig. 4.24 shows the feedback-control view of the ac/dc combined system, where Kac (s) is the controller transfer matrix of the ac system while Kdc (s) is the controller transfer matrix of the dc system, i.e., the direct-voltage controllers. In other words, the dc Jacobian transfer matrix Gdc (s) is treated as an additional controlled process of the 81

Pac1
VSC1
Pdc1
Pdc2
VSC2
Pac2
Pac1
VSC1
Pdc1
Rdc
idc
Ldc
Pdc2
VSC2
Pac2
+
udc Cdc
+
udc1 Cdc1
+
udc2 Cdc2
(a)
(b)
Fig. 4.25 DC-link representation for a two terminal VSC-HVDC link. (a) Single dc capacitor. (b) -link.
closed-loop system of the combined ac/dc system. The ac Jacobian transfer matrix J(s) and the dc Jacobian transfer matrix Gdc (s) are connected by the equivalence of the acpower vector Pac and the dc-power vector Pdc , i.e., Pac = Pdc (4.73)
where the minus sign is due to the denition of the power directions. By using the equivalence in (4.73), it should be pointed out that the losses of the converter valves can introduce an error, which will be discussed later. In the following, a dc link connected to two VSCs, i.e., a two-terminal VSC-HVDC link, is modeled by the dc Jacobian transfer matrix modeling concept. Fig. 4.25(a) shows an dc-link circuit that is represented by a single dc capacitor. The capacitor bank at the dc link is an energy storage. The time derivative of the stored energy must equal the sum of the instantaneous power infeed from the two converters (neglecting the losses). The direct-voltage dynamics can thus be written as d (u2 1 dc ) Cdc = Pdc1 + Pdc2 2 dt (4.74)
where Cdc is the dc-link capacitance and udc is the direct voltage. Pdc1 and Pdc2 are the instantaneous powers from VSC1 and VSC2 . If the direct-voltage controller were to operate directly on the error uref dc udc , the closed-loop dynamics would be dependent on the operating point udc0 . This inconvenience is avoided by selecting the direct-voltage con2 troller operating instead on the error uref u2 dc dc as suggested in [13,85]. Consequently, the dc-link dynamics can be written in the linearized form u2 dc = 2 (Pdc1 + Pdc2 ) . sCdc
Gdc (s)
(4.75)
If the dc transmission line is a long overhead line, then the resistance and the inductance of the dc line have to be taken into account. Fig. 4.25(b) shows a dc link represented by 82

a -link model. Based on Kirchhoffs voltage and current laws, the dynamic equations of the dc circuit can be written as dudc1 Pdc1 Cdc1 = idc dt udc1 didc = udc1 udc2 Rdc idc Ldc dt dudc2 Pdc2 Cdc2 = idc + (4.76) dt udc2 where Cdc1 and Cdc2 are lumped capacitances at the two VSC stations. The quantities udc1 and udc2 are the direct voltages. Pdc1 and Pdc2 are the instantaneous powers from VSC1 and VSC2 . Rdc and Ldc are the resistance and the inductance of the dc line. The quantity idc is the direct current of the dc line. If the linearized deviation variables Pdc1 2 and Pdc2 of Pdc1 and Pdc2 are chosen as the inputs, and u2 dc1 and udc2 of udc1 and udc2 are chosen as the outputs, a state-space model can be obtained by linearization of (4.76) d x = Ax + B u dt y = C x + Du (4.77) where
1 u2Pdc10 Cdc1 Cdc1
dc10
A= C= u=
0 L1 dc u2Pdc20 dc20 Cdc2 , D=
1 Ldc
dc R Ldc
0 2udc10 0 0
1 Cdc2
0
T
0 0 0 0
, B =
1 udc10 Cdc1
0 0
0
1 udc20 Cdc2
0 2udc20
, x=
T
udc1 idc udc2 .
Pdc1 Pdc2
, y=
2 u2 dc1 udc2
(4.78)
The state-space representation in (4.77) can also be written in input-output transfer matrix form by (4.40) which yields u2 Pdc1 Gdc11 (s) Gdc12 (s) dc1 = . (4.79) u2 P dc2 Gdc21 (s) Gdc22 (s) dc2
Gdc (s)
The poles of Gdc (s) can be solved analytically for the operating point where Pdc10 = Pdc20 = 0 with the expression p1 = 0 p2,3 = Rdc j 2Ldc
2 Rdc Cdc1 + Cdc2 . 2 4Ldc Ldc Cdc1 Cdc2
(4.80)
83

Table 4.4 Parameters of the dc Jacobian transfer matrix for the -link model. Per unit based on 350 MW and 150 kV.
Parameters Rdc Main-circuit parameters Ldc Cdc1 Cdc2 Udc10 Initial conditions Pdc10
Value 0.04 p.u. 0.0025 p.u. 0.0077 p.u. 0.0077 p.u. 1.0 p.u. 0.0 p.u.
As shown in Fig. 4.26, the poles of Gdc (s) are dependent on the operating point. With the increase of loading, the frequencies of the two complex poles p2,3 are reduced, and the real pole p1 at the origin moves into the right-half plane. The main-circuit parameters and initial conditions of the dc Jacobian transfer matrix for the -link are given in Table 4.4. In control theory, the RHP pole of the process imposes a fundamental lower limit on the bandwidth of the controller, i.e., the closed-loop system of the direct-voltage control has to achieve a bandwidth that is higher than the location of the RHP pole of Gdc (s) to stabilize the process. Recalling also the upper limit of bandwidth imposed by the RHP transmission zero of the ac Jacobian transfer matrix [cf. (4.17)], it is generally more complicated to operate grid-connected VSCs at high load angles. The instability of Gdc (s) is to do with the resistance of the dc link. With dc powers as the inputs, as the way how VSCs work to a dc link, the dc resistance gives a destabilizing effect. The analytical solutions of the poles of Gdc (s) with other operating point than Pdc10 = Pdc20 = 0 are difcult to obtain. However, Appendix B.2 gives a rigorous mathematical proof of the instability of Gdc (s) for other operating points than Pdc10 = 0 or Pdc20 = 0. The mathematical proof also shows the role of the dc resistance to the instability of Gdc (s). Fig. 4.27 shows the root-loci of Gdc (s) by varying the dc-resistance values. With increased dc resistance, the resonant pole pair p2,3 becomes more damped, but p0 moves towards the right-half plane. It should be noted that the destabilizing effect of the dc resistance only becomes apparent if the dc-transmission line is sufciently long, e.g., HVDC transmission over long-distance overhead lines. By applying the QZ method, the transmission zero of Gdc (s) for the -link model is derived with a surprisingly simple expression z= Rdc . Ldc (4.81)
In contrast to the zeros of the ac Jacobian transfer matrix, the location of the zero of the dc Jacobian transfer matrix is independent of the operating points. 84
400 300 200 (rad/sec) 100 z 0 -100 -200 -300 -400 -30 p
1
p2
p3 -20 -10 0 (rad/sec) 10 20 30
Fig. 4.26 Root-loci of Gdc (s) regarding variations of Pdc10 from 0.0 p.u. to 1.0 p.u.
400 300 200 (rad/sec) 100 z 0 -100 -200 -300 -400 -200 p1 p2
p -150 -100 -50
0 (rad/sec)
50
100
150
200
Fig. 4.27 Root-loci of Gdc (s) regarding variations of Rdc from 0.0 p.u. to 0.24 p.u. Initial conditions: Pdc10 = 1.0 p.u., udc10 = 1.0 p.u.
85

J u d ( s) v V V0 PVSC1 = Pdc1
2 udc1
J(s)
U f
Pdc2
G dc ( s )
2 udc2
Fig. 4.28 Denition of the transfer function Jud (s).
To verify the proposed modeling concept with frequency-scanning results from time simulations, a transfer function Jud (s) is dened as Jud (s) = JP (s)Gdc11 (s) (4.82)
which is shown graphically in Fig. 4.28. It should be noted that the active power derived previously for the ac Jacobian transfer matrix is the power from the lter bus, which is somewhat different from the active power owing from the VSC due to the energy stored in the phase reactor. Therefore, the ac power Pac should be obtained from the linearization of Pac = PVSC = Re {vi c} . (4.83)
In Fig. 4.28, the active power owing from the VSC is denoted as PVSC to distinguish it from the active power from the lter bus to the ac system. Fig. 4.29 shows the Bode plots of Jud (s) overlapped with the frequency-scanning results from PSCAD/EMTDC for the two types of dc-link representations respectively, where the impedance-source system shown in Fig. 4.4 is used as the ac-network conguration. Besides, the high-pass current control HHP (s) has been applied in the ac Jacobian transfer matrix. Some discrepancies can be observed between the plots of the linear models and the frequency-scanning results, which are mainly due to the lack of valve-loss representation of the linear models. The most noticeable one is the resonance peak of the -link model in Fig. 4.29(b). The frequency-scanning results show that the valve losses have a damping effect on the resonance peak while the linear model tends to overestimate the impact of the dc resonance. In addition, the slopes of the magnitudes of the linear models in both Fig. 4.29(a) and Fig. 4.29(b) are generally steeper than those from the frequency-scanning results, which can be considered as the resistive effects of the valve losses. To properly represent the losses of the converter valves, the linear model should take into account the topology of the converter as well as the applied PWM technique, which is of nonlinear nature. These issues certainly require further investigations in the future research. 86
4.5. Summary of the properties of the Jacobian transfer matrix

10 |JU (j)|
5
10 |JU (j)|
1 2 3
10 ARG JU (j) (deg.)
-5
10
10
10
10
10
10 ARG JU (j) (deg.)
-5
10
10
10
0 -200 -400 -600 -800
10
10 (rad/sec)
10
0 -200 -400 -600 -800
10
10 (rad/sec)
10
(a) Single dc capacitor model.
(b) -link model.
Fig. 4.29 Bode plots of Jud (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
4.5 Summary of the properties of the Jacobian transfer matrix

In this chapter, the concept of Jacobian transfer matrix is proposed for dynamic modeling of ac/dc systems. By investigating some simplied ac/dc network congurations connected to VSCs, several properties of the ac/dc Jacobian transfer matrix can be summarized below: 1. The complex poles of the ac/dc Jacobian transfer matrix generally reect the resonances in the ac/dc network. The poles of the ac Jacobian transfer matrix are independent of the operating point, while the poles of the dc Jacobian transfer matrix are dependent on the operating point except in the special case when the dc network is represented by a single dc capacitor. 2. The ac Jacobian transfer matrix is an inherently stable process, although the poles might be poorly damped and the zeros can move to the origin. Therefore, instability of an ac system can only be caused by feedback control. On the other hand, the dc Jacobian transfer matrix is an inherently unstable process or has a pole at the origin. Therefore, a dc system has to be stabilized by feedback control, i.e., the direct-voltage control. 3. The zeros (on the real axis) of the ac Jacobian transfer matrix are dependent on the operating point. By moving to the origin, the zeros impose a fundamental limitation on power transmission in the ac network. This limitation has manifested as anglestability and voltage-stability phenomena in the classical power-system stability theory. If the zero of the ac Jacobian transfer matrix appears on the right-half plane, 87

it also imposes an upper limitation on the achievable bandwidth of the input device, i.e., the bandwidth of the power controller of the input device should be lower than the location of the RHP zero. On the other hand, the dc Jacobian transfer matrix does not have such operating-point-dependent zero. The above difference of ac and dc systems explains the fact that power can be transmitted in the dc network up to the thermal limit but is often limited in the ac network, especially for long-distance power transmissions. 4. The pole on the right-half plane of the dc Jacobian transfer matrix imposes a fundamental lower limitation on the bandwidth of the direct-voltage controller, i.e., the bandwidth of the direct-voltage controller should be higher than the location of the RHP pole to stabilize the process. It should be pointed out that the above conclusions about the dc system are only valid for the dc system dened in this thesis, i.e., the VSCs are the only input devices to the dc network, and dc powers are the inputs while direct voltages are the outputs of the dc Jacobian transfer matrix. For example, for dc systems connected to LCCs, the dynamic modeling is certainly different. Such scenarios are, however, out of the scope of this thesis. In this chapter, the ac Jacobian transfer matrix has only considered the VSC as the input device. However, the concept of the ac Jacobian transfer matrix is meant for any type of input device in ac systems. In Chapter 6, the Jacobian transfer matrices for two other input devices, i.e., synchronous generators and induction motors, are developed.
4.6 Summary
In this chapter, the concept of Jacobian transfer matrix is proposed for dynamic modeling of ac/dc systems. The fundamental idea of the ac Jacobian transfer matrix modeling is that a synchronous power system is modeled as one multivariable feedback-control system, where the feedback controllers and the controlled process of the power system are explicitly dened. The concept has been applied to model grid-connected VSCs using power-synchronization control and vector current control for several simplied acnetwork congurations. The modeling concept is also extended to model dc systems constructed by multiple VSCs. By theoretical analysis, it is found that the poles and zeros of the Jacobian transfer matrix give useful information about the properties of the ac/dc system.
88
Chapter 5 Control of VSC-HVDC Links Connected to High-Impedance AC Systems

In this chapter, the control of VSC-HVDC links connected to high-impedance weak ac systems is investigated. In Section 5.1, general aspects of high-impedance ac systems are described. In Section 5.2, the performance of vector current control and powersynchronization control for weak-ac-system connections are compared based on the Jacobian transfer matrix modeling concept. In Section 5.3, multivariable feedback designs of power-synchronization control are investigated by two design approaches, i.e., internal model control and H control. In Section 5.4, a direct-voltage controller is proposed for power-synchronization control. The dc-capacitance requirement for VSC-HVDC links connected to weak ac systems is discussed. Finally, a control structure for interconnecting two very weak ac systems is proposed in Section 5.5. The major results of this chapter are summarized in Section 5.6. Some results of this chapter are included in [63, 86].
5.1 General aspects of high-impedance ac systems

As it was mentioned in Chapter 2, there is an inherent weakness with the conventional LCC-HVDC system, i.e., the commutations of the thyristor valves are dependent on the stiffness of the alternating voltage. The converter cannot work properly if the connected ac system is weak. Substantial research has been performed in this eld [5, 6, 8]. The most outstanding contribution on this subject is [5]. According to [5], an ac system can be considered as weak from two aspects: 1. ac-system impedance is high. 2. ac-system inertia is low. Either of the two system conditions may become an obstacle for conventional LCC-HVDC applications. If an HVDC link is terminated at the weak point of a large ac system, i.e., the 89
Chapter 5. Control of VSC-HVDC Links Connected to High-Impedance AC Systems

equivalent impedance of the ac system is high, the alternating voltage at the lter bus will become sensitive to power variations of the HVDC link. This difculty is usually measured by the short-circuit ratio (SCR), which is a ratio of the ac-system short-circuit capacity vs. the rated power of the HVDC link. If the ac system is represented as an impedance source as shown in Fig. 4.4, then the SCR is directly related to the ac-system inductance Ln . According to [5], SCR is dened as SCR = Sac PdcN (5.1)
where Sac is the short-circuit capacity of the ac system at the lter bus, while PdcN is the rated dc power of the HVDC link. The short-circuit capacity of the ac system Sac can be expressed as 2 2 Uf Uf 0 0 Sac = (5.2) Z 1 Ln where 1 is the angular frequency of the ac system and Z is the equivalent impedance of the ac system. To further simplify the expression of SCR in (5.1), the lter-bus voltage is assumed to be identical to the base value, i.e, Uf 0 = UacN , and the rated power of the HVDC link PdcN is used as the base power of the ac system, i.e., SacN = PdcN . If 1 Ln is expressed in per unit, it follows from (5.1) and (5.2) that SCR can be expressed as SCR = 1 . 1 Ln (5.3)
The following is a denition of the strength of an ac system based on the classication of [5]: Strong system, if the SCR of the ac system is greater than 3.0. Weak system, if the SCR of the ac system is between 2.0 and 3.0. Very weak system, if the SCR of the ac system is lower than 2.0. One of the driving forces to develop the VSC-HVDC technology is to overcome the weak-ac-system connection problem of the conventional LCC-HVDC system. By applying VSC techniques, the two notorious weak-ac-system related problems for conventional LCC-HVDC systems, i.e., the transient over voltage (TOV) and the low-order harmonic resonance, are no longer big issues. For VSC-HVDC systems, large ac capacitors are not needed for reactive-power compensation. Moreover, as shown in Chapter 4, for both vector current control and power-synchronization control, the control system can easily provide damping to resonances in ac systems at any frequency 1 . However, weak-ac-system connections still represent more challenging operation conditions for VSC-HVDC systems due to the following reasons:
However, as it is shown by [12], the problem of low-harmonics resonance might not be trivial for vector current control, but it is not a problem for power-synchronization control thanks to the high-pass current control function.
1
90
5.2. Comparison of power-synchronization control and vector current control

1. The lter-bus voltage is more sensitive to power variations of the VSC-HVDC link in weak ac systems. Thus, a weak-ac-system connection requires that the control system of the VSC-HVDC link must be less dependent on the stiffness of the lterbus voltage. 2. The SCR of the ac system imposes a theoretical limitation on the maximum power that the VSC-HVDC system is possible to transmit to or from the ac system. This limitation has traditionally been analyzed by the maximum power curve (MPC) for conventional LCC-HVDC systems [5, 87]. As it was discussed in Chapter 4, such power limitations are basic characteristics of the ac system that are related to the operating-point-dependent zeros of the ac Jacobian transfer matrix. Moreover, for a VSC-HVDC link, the RHP zero of the ac Jacobian transfer matrix imposes a fundamental limitation on the achievable bandwidth of the control system. 3. The off-diagonal elements of the Jacobian transfer matrix are larger for weak-acsystem connections, especially with high loadings. This problem is in common for vector current control and power-synchronization control, but it is more serious for vector current control. High off-diagonal elements imply more interactions between the active-power control and the alternating-voltage control. 4. Model uncertainties, either due to variations of ac-network congurations or operating points, are generally higher for weak-ac-system connections. The control system of the VSC-HVDC link is required to be robust for model variations. In the following sections, the above issues related to high-impedance ac systems will be addressed. The low-inertia ac-system connection issue will be discussed in Chapter 6.
5.2 Comparison of power-synchronization control and vector current control

In this section, the dynamic performance of power-synchronization control and vector current control are compared for VSC-HVDC links connected to high-impedance weak ac systems. For simplicity, the impedance-source system shown in Fig. 4.4 is assumed as the ac-network conguration. Except for the quantities that are varied, the parameters of the network conguration, initial conditions, and control systems, are based on those dened in Table 4.2 and Table 4.3. The comparisons are performed by analyzing the characteristics of the Jacobian transfer matrices in the frequency domain and by verications using time simulations. Fig. 5.1 shows the Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control regarding the variations of the SCRs of the ac system. The VSC-HVDC converter is assumed to operate with zero loading P = 0.0 p.u., i.e., 91
10 |JPI (j)|
d
10 |JPI (j)|
q
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JU I (j)|
fd
10 |JU I (j)|
fq
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.1 Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control with P = 0.0 p.u. (solid: SCR = 5.0, dashed: SCR = 2.0, dotted: SCR = 1.0).
10 |JPI (j)|
d
10 |JPI (j)|
q
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JU I (j)|
fd
10 |JU I (j)|
fq
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.2 Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control with SCR = 1.0 (solid: P = 0.0 p.u., dashed: P = 0.5 p.u., dotted: P = 0.7 p.u.).
92

static synchronous compensator (STATCOM) mode. For simplicity, only the magnitudes of the transfer functions are plotted. As it was mentioned before, one typical feature of high-impedance ac systems is that the lter-bus voltage is more sensitive to power variations of the VSC-HVDC link. This effect is reected in transfer functions JUf Id (s) and JUf Iq (s) where their magnitudes are getting larger with decreased SCRs of the ac system. Meanwhile, the magnitudes of the off-diagonal transfer functions, increase with lower SCRs, which imply difculties for the outer-loop controllers, i.e., more interactions between the active-power control loop and the alternating-voltage control loop are expected. However, even more severe difculties are experienced with increased loadings of the VSC-HVDC link that are shown below. Fig. 5.2 shows the Bode plots of the transfer functions of the Jacobian transfer matrix regarding the variations of the loadings of a VSC-HVDC link connected to an ac system with SCR = 1.0. It is easily observed that the magnitudes of the off-diagonal transfer functions increase dramatically with increased loadings, while the magnitude of JPId (s) decreases. Beyond 50% of the VSC-HVDC loading (the dashed curves), the magnitudes of the off-diagonal transfer functions are greater than those of the diagonal transfer functions. In control theory, such a process is called ill-conditioned process which implies particular difculties for the outer-loop controllers [82]. Of all the transfer functions, the variation of JUf Id (s) plays the most critical role. From Fig. 4.22c in Chapter 4, it can be found that JUf Id (s) has 180 phase shift at low frequencies which means that the in2 crease of iref d always results in a decreased lter-bus voltage Uf , or vise versa. This effect is greatly amplied with higher-loading conditions for a VSC-HVDC converter connected to a weak ac system. This is also the reason to that the magnitude of JPId (s) decreases with high-loading conditions as will be shown by the time-simulation results in below. Figs. 5.3-5.5 show the time-simulation results from PSCAD/EMTDC by applying a 0.1 p.u. step to iref d at 0.1 s. The pre-conditions of the three gures correspond to the three operating points of the Bode plots in Fig. 5.2 respectively. The time-simulation results agree with the conclusion drawn from the frequency-domain analysis. For example, the time-simulation results in Fig. 5.5 show that, with 70% loading of the VSC-HVDC link, the 0.1 p.u. step of iref d does not give any increase of the active-power output due to the decreased lter-bus voltage Uf , although the converter current ic cd still follows the current ref reference id . The same conclusion can be drawn by the frequency-domain analysis. In the following, similar frequency-domain analysis and time simulations are performed for power-synchronization control. The network congurations and operating points are chosen identical as those for the analysis of vector current control. Fig. 5.6 shows the Bode plots of the transfer functions of the Jacobian transfer matrix for power-synchronization control regarding variations of SCRs of the ac system.
JPIq (s) and JUf Iq (s) also have 180 phase shift at low frequencies, but they are in the same direction for iref q .
2
93
iref, ic (p.u.)
0.1 0.05 0 0 0.1 0.2 0.3 0.4
iref d ic
cd
cd
0.5
P (p.u.)
0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.5
Uf (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
Fig. 5.3 Step response of id for vector current control at P = 0.0 p.u. with SCR = 1.0.
0.65 iref, ic (p.u.) 0.6 0.55 0.5 0 0.65 P (p.u.) 0.6 0.55 0.5 0 0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 id ic
ref
cd
cd
0.5
Uf (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
94
iref, ic (p.u.)
0.8 0.75 0.7 0 0.1 0.2 0.3 0.4
id ic
ref
cd
cd
0.5
P (p.u.)
0.8 0.75 0.7 0.65 0 0.1 0.2 0.3 0.4 0.5
Uf (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
10 |JP(j)|
10 |JPV(j)|
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JU (j)|
f
10 |JU V(j)|
f
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.6 Bode plots of the transfer functions of the Jacobian transfer matrix for power-synchronization control with P = 0.0 p.u. (solid: SCR = 5.0, dashed: SCR = 2.0, dotted: SCR = 1.0).
95

It is interesting to note that the magnitudes of the off-diagonal transfer functions decrease slightly with decreased SCRs of the ac system. However, the variations of the magnitudes of JP (s) are larger than JPId (s) in Fig. 5.1. Similar to vector current control, Fig. 5.7 shows that the magnitudes of the off-diagonal transfer functions increase with higher loadings of the VSC-HVDC system for power-synchronization control, which means that the outer-loop controllers of power-synchronization control also experience difculties with increased loadings of the VSC-HVDC system. However, the magnitude of JUf (s) is much lower than JUf Id (s) with the same operating conditions. Besides, the magnitude of JP (s) does not decrease as dramatically as JPId (s) with increased loadings of the VSC-HVDC system. The above difference between JUf (s) and JUf Id (s) is the key factor that affects the dynamic performance of power-synchronization control and vector current control for VSC-HVDC links connected to weak ac systems. With power-synchronization control, the active power is controlled by the phase angle of the VSC voltage, meanwhile the magnitude of the VSC voltage is held constant. Such a control strategy minimizes dynamic variations of the lter-bus voltage. However, with vector current control, the inner-current control loop only controls the currents, while control of the lter-bus voltage is left to the outer-loop controller. Accordingly, dynamic variations of the lter-bus voltage are much higher for vector current control than for power-synchronization control. Such dynamic variations of the lter-bus voltage create substantial difculties for the outer-loop controllers. Figs. 5.8-5.10 show the simulation results from PSCAD/EMTDC when applying a 0.1 rad step to v for power-synchronization control. Comparing with the simulation results with vector current control, the variations of the lter-bus voltage are much smaller for power-synchronization control in response to the same power changes. Besides, the dynamic responses of the open-loop transfer functions for power-synchronization control are more stable than those for vector current control since the Jacobian transfer matrix for vector current control includes more feedback loops, especially the PLL introduces a delay. The difference of the dynamic response can also be explained by the Bode plots of the transfer functions, where the magnitudes are atter at low frequencies for powersynchronization control than for vector current control. Generally speaking, with power-synchronization control, the control system can maintain stable operation with load angle up to u0 = 60 , i.e., approximately 86% of the rated power, for VSC-HVDC links connected to ac systems with SCR = 1.0, whereas it is very difcult for vector current control to operate even with 50% loading. Ref. [45] gives more simulation results for such comparisons. However, if the connected ac system is strong, vector current control seems to be more advantageous than power-synchronization control. Figs. 5.11-5.12 show the Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control and power-synchronization control with SCR variations from 3 to 10. In Fig. 5.11 vector 96
10 |JP(j)|
10 |JPV(j)|
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JU (j)|
f
10 |JU V(j)|
f
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.7 Bode plots of the transfer functions of the Jacobian transfer matrix for power-synchronization control with SCR =1.0 p.u. (solid: P = 0.0 p.u., dashed: P = 0.5 p.u., dotted: P = 0.7 p.u.).
v (rad)
0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.5
P (p.u.)
0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.5
U (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
Fig. 5.8 Step response of v for power-synchronization control at P = 0.0 p.u. with SCR = 1.0.
97
v (rad)
0.7 0.65 0.6 0 0.65 0.1 0.2 0.3 0.4 0.5
P (p.u.)
0.6 0.55 0.5 0 0.1 0.2 0.3 0.4 0.5
U (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
1.05 v (rad) 1 0.95 0.9 0 0.1 0.2 0.3 0.4 0.5
P (p.u.)
0.8 0.75 0.7 0.65 0 0.1 0.2 0.3 0.4 0.5
U (p.u.)
1 0.8 0
0.1
0.2 time (sec)
0.3
0.4
0.5
98
10 |JPI (j)|
d
10 |JPI (j)|
q
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JQI (j)|
d
10 |JQI (j)|
q
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.11 Bode plots of the transfer functions of the Jacobian transfer matrix for vector current control with P = 0.0 p.u. (solid: SCR =10.0 p.u., dashed: SCR =5.0 p.u., dotted: SCR =3.0 p.u.).
10 |JP(j)|
10 |JPV(j)|
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
10 |JU (j)|
f
10 |JU V(j)|
f
10
-2
10
0 2 4
-2
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 5.12 Bode plots of the transfer functions of the Jacobian transfer matrix for power-synchronization control with P = 0.0 p.u. (solid: SCR =10.0 p.u., dashed: SCR =5.0 p.u., dotted: SCR =3.0 p.u.).
99

current control controls the reactive power instead of controlling the alternating voltage, which is acceptable for strong ac-system connections. By comparing Fig. 5.11 and Fig. 5.12, it can be seen that the magnitude of the offdiagonal elements of the Jacobian transfer matrix for vector current control are generally less than those for power-synchronization control, and more importantly, the magnitudes of the diagonal elements of the Jacobian transfer matrix for vector current control are much less sensitive to SCR variations than those for power-synchronization control. In other words, vector current control is more robust than power-synchronization control if the connected ac system is guaranteed to be strong. Fig. 5.13 shows the time-simulation results of vector current control from PSCAD/ref EMTDC by applying 0.1 p.u. steps with iref d and iq respectively where the connected ac system has an SCR of 10. Vector current control shows excellent reference tracking and decoupling capabilities. It should be noted that the Jacobian transfer matrices include various control functions for both power-synchronization control and vector current control. The parameters of those control functions can also affect the above frequency-domain analysis and timesimulation results. However, the selection of control parameters will not affect the major conclusions drawn in this section since the difference between vector current control and power-synchronization control is more related to their fundamental control strategies. Therefore, such an analysis is not given in the thesis but typical values for those control parameters are selected. Among the two control methods compared in this section, power-synchronization control is the most suitable control system for VSC-HVDC links connected to weak ac systems. Consequently, only power-synchronization control will be applied for the remaining work of the thesis.
5.3 Multivariable feedback designs

As shown in Section 5.2, the magnitudes of the off-diagonal transfer functions of the Jacobian transfer matrix become larger for weak-ac-system connections. To achieve high control bandwidth and to decouple the interactions between the active-power control loop and the alternating-voltage control loop, multivariable feedback control design should be applied. Fig. 5.14 shows the general control block diagram for multivariable feedback designs, i.e., a controller transfer matrix K(s) with four controllers K11 (s), K12 (s), K21 (s), and K22 (s) is applied for active-power and alternating-voltage control. In this section, two multivariable feedback design approaches are investigated: Internal model control (IMC) and H control. For simplicity, the impedance-source system shown in Fig. 4.4 is assumed as the ac-network conguration. 100
5.3. Multivariable feedback designs
iref , ic (p.u.) d cd
0.1 0.05 0 0 0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4
iref icd 0.5

d c
P (p.u.)
0.5 iref
q
iref , ic (p.u.) q cq
0.1 0.05 0 -0.05 0 0.05 0 -0.05 -0.1 0
ic cq 0.1 0.2 0.3 0.4 0.5
Q (p.u.)
0.1
0.2 time (sec)
0.3
0.4
0.5
ref Fig. 5.13 Step response of iref d and iq for vector current control at P = 0.0 p.u. with ac-system SCR = 10.0.
K11 (s)
Pref + -
K12 (s)
+ +
U ref
K 21 (s) +K 22 (s)
P AC-Jacobian transfer V matrix U f V0 J ( s)
Fig. 5.14 Multlivariable feedback control of grid-connected VSCs using power-synchronization control
101

r
K (s)
J (s)
~ J ( s) + ~ y -
Fig. 5.15 Block diagram of the internal model control system.
5.3.1 Internal model control

The concept of IMC design is based on the idea that the transfer functions of the closedloop system can be directly shaped by explicitly including a process model in the controller [88]. Fig. 5.15 shows the the block diagram of the IMC design, where r = [Pref Uref ]T , u = [v V /V0 ]T , and y = [P Uf ]T . From Fig. 5.15, it can be easily observed that, if the process J (s) is stable and the process model J (s) is equal to J (s), the whole system is internally stable if and only if K (s) is stable. Thus, if K (s) is chosen as K (s ) = J 1 (s ) F (s ) the output response would be y = J (s) J1 (s) F (s) r = F (s) r. (5.5) (5.4)
Therefore, the bandwidth of the closed-loop system would be solely determined by the lter F (s), which is often dened as a rst-order lter with the desired bandwidth. However, for VSC-HVDC applications, the above philosophy cannot work directly, since the Jacobian transfer matrix J (s) contains an RHP transmission zero as shown in Chapter 4. A direct inverse of J (s) would end up with an unstable controller. In such a situation, a factorization technique needs to be applied [89]. Consequently, the process model can be factorized into two parts J (s ) = J p (s ) J n (s ) (5.6)
where Jn (s) contains the invertible elements and Jp (s) contains the non-invertible elements. An easy factorization method is to place the RHP transmission zero at the diagonal part of Jp (s) as ZRHP s 0 ZRHP +s J p (s ) = (5.7) ZRHP s 0 ZRHP +s where ZRHP is the RHP transmission zero of J (s). The invertible matrix Jn (s) can be solved by (5.6) and (5.7). A low-pass lter matrix F (s) is chosen to specify the closed102

K ' (s)
K (s) +
~ J ( s)
J (s)
Fig. 5.16 IMC in classical feedback-control structure.
loop bandwidth F (s ) =
p s+p
0
u s+u
(5.8)
where p and u are the desired bandwidths of the active-power controller and alternatingvoltage controller. In specifying the bandwidth p and u , the RHP transmission zero of J (s) has to be considered. As a rule of thumb, the bandwidth of the closed-loop system should be chosen at least lower than half of the location of the RHP zero [82]. Thus, the controller is designed as
1 K (s) = J n (s ) F (s ) .
(5.9)
1 If the model is assumed to be perfect, i.e., J (s) J n (s) = Jp (s), then the output response would be y = J p (s ) F (s ) r . (5.10)
Because both Jp (s) and F (s) are diagonal, the resulting closed-loop response is also diagonal. The IMC controller can easily be formulated into the classical feedback-control structure, as shown by the block diagram in Fig. 5.16. The controller K (s), therefore, can be written as K (s) = K (s) [I J (s) K (s)]1 (5.11) where I represents the identity matrix. Substituting (5.6) and (5.9) into (5.11) yields
1 1 K (s ) = J n (s) F (s) [I Jp (s) F (s)] .
(5.12)
Due to its feedback-control nature, IMC is able to compensate for load disturbances and model uncertainties. However, for controller design, a nominal model needs to be dened. The nominal model should be chosen in the center of the possible process models such that the model uncertainties are minimized within all the possible ac-network congurations and operating conditions. 103

For VSC-HVDC applications, this is not a simple task, since there are a few factors that affect the model uncertainties of the Jacobian transfer matrix. The following is just a simple rule to give an initial guidance. The nominal model has to be re-adjusted based on the results of the robustness tests as shown later. For power-synchronization control, the power-angle relationship has the dominant role in the controller design. From Fig. 5.6, it can be observed that the magnitude of JP (s) varies with the SCR of the ac system. The uncertainty of JP (s) should be given higher priority in the controller design since it affects the effective gain of the active-power controller. If the ac system is simplied as an impedance source and the ac capacitor at the lter bus is neglected, it is known from Table 4.1 that the static magnitude of JP (s) is inversely proportional to the inductance L, which is the sum of the ac-system inductance Ln and the phase-reactor inductance Lc . If the short-circuit ratios of the weakest and strongest ac-system scenarios are SCRmin and SCRmax respectively, it is reasonable to choose the short-circuit ratio SCRnom of the nominal ac system such that 1 1 = 1 2 1 Lc + SCRnom The solution to (5.13) is found to be SCRnom = 21 Lc SCRmin SCRmax + (SCRmin + SCRmax ) . 2 + 1 Lc (SCRmin + SCRmax ) (5.14) 1 1 + 1 1 1 Lc + SCRmin 1 Lc + SCR max . (5.13)
If there is no other preference, the operating point of the nominal model should be chosen with the full loading of the VSC-HVDC link. Since the process model is included in the IMC controller, the order of the controller is high if the ac-system conguration is complex. Thus, a simple process model should be chosen as the nominal model to reduce the order of the controller. For this particular application, the ac capacitor at the lter bus is neglected in the nominal model, but included in the robustness tests as a model uncertainty. Fig. 5.17 shows the step response of the active-power and alternating-voltage of the closed-loop system of the nominal model. The parameters and initial conditions of the nominal model are given in Table 5.1. As shown by the gure, the IMC controller successfully decouples the cross-coupling between the two control channels. However, the test result shows that the controller has poor robustness with power-direction uncertainty. The reason is that the transfer function JPV (s) is power-direction dependent (cf. Table 4.1). Since the power direction is a known information for a VSC-HVDC link, it is feasible to design two different controllers for the rectier and the inverter operation modes respectively. Fig. 5.18 and Fig. 5.19 show the robustness-test results of the IMC controllers regarding variations of the SCRs of the ac system and the power levels for rectier and inverter operation respectively. With an SCR = 1.0 p.u., the maximum loading is chosen as P = 0.86 p.u. instead of P = 1.0 p.u. 104
From: Pref To: P 1 P (p.u.) 0.5 0 0 0.1 0.2 time (sec) From: Pref To: Uf 1 Uf (p.u.) 0.5 0 0 0.1 0.2 time (sec) 0.3 Uf (p.u.) 1 0.5 0 0 0.3 P (p.u.) 1 0.5 0 0
From: Uref To: P
0.1 0.2 time (sec) From: Uref To: Uf
0.3
0.1 0.2 time (sec)
0.3
Fig. 5.17 Active-power and alternating-voltage step responses of the IMC controller with the nominal model. Control parameters: p = 100 rad/sec, u = 100 rad/sec.
Table 5.1 Parameters of the nominal model used for IMC and H control design. Per unit based on 350 MVA and 195 kV.
Parameters Rc 1 Lc Main-circuit parameters 1 Cf Rn 1 Ln E0 Initial conditions V0 u0 (v0 ) kv High-pass current control v
Values 0.01 p.u. 0.2 p.u. 0.17 p.u. 0.01 p.u. 0.667 p.u. 1.0 p.u. 1.08 p.u. 40.6 (51 ) 0.45 p.u. 40 rad/s
105
From: Pref To: P 1.5 P (p.u.) P (p.u.) 1 0.5 0 -0.5 0 0.1 0.2 time (sec) From: Pref To: Uf 1 0.5 0 0 0.1 0.2 time (sec) 0.3 1 0.5 0 0 0.3 1.5 1 0.5 0 -0.5 0
From: Uref To: P
0.3
Uf (p.u.)
Uf (p.u.)
0.1 0.2 time (sec)
0.3
Fig. 5.18 Robustness tests by active-power and alternating-voltage step responses with the IMC controller for the VSC-HVDC converter operating in inverter mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
From: Pref To: P 1.5 P (p.u.) P (p.u.) 1 0.5 0 -0.5 0 0.1 0.2 time (sec) From: Pref To: Uf 1 Uf (p.u.) 0.5 0 0 0.1 0.2 time (sec) 0.3 Uf (p.u.) 1 0.5 0 0 0.3 1.5 1 0.5 0 -0.5 0
From: Uref To: P
0.3
0.1 0.2 time (sec)
0.3
Fig. 5.19 Robustness tests by active-power and alternating-voltage step responses with the IMC controller for the VSC-HVDC converter operating in rectier mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
106
(s )
K (s)
J (s)
I+
(s )
+ +
Fig. 5.20 Additive and multiplicative uncertainties.
For the real system, the measuring signals are usually low-pass ltered, and the controller computation and PWM switching can also introduce a delay. Thus, a transfer matrix with the expression J (s) =
1 eTd s 1+ s
0
1 eTd s 1+ s
(5.15)
is cascade-connected to the Jacobian transfer matrix J (s) for the robustness tests. The parameters of J (s) are chosen as = 3 ms and Td = 0.5 ms. The design of the IMC controller is an iterative process. If the robustness is not satised, the controller has to be detuned, i.e., choosing lower values of p and u , to meet the robustness requirement. In the next subsection, another multivariable feedback design approach, i.e., H control, is introduced. With H control, the robustness requirement on the controller can be explicitly specied in the initial controller design stage.
5.3.2 H control
H control is another methodology for multivariable controller design. The main feature of H control is its explicit way in dealing with model uncertainties, i.e., an H controller can achieve closed-loop stability with satisfactory performance under process variations as well as in the presence of other uncertainties such as disturbances and errors in the sensors [82, 90]. Fig. 5.20 shows a standard multivariable feedback-control block diagram including process uncertainties. A (s) and M (s) represent the additive and multiplicative uncertainties respectively. The transfer matrices S(s), R(s) and T(s) are dened as S (s) = [I + J (s) K (s)]1 R (s) = K (s) [I + J (s) K (s)]1 T (s) = J (s) K (s) [I + J (s) K (s)]1 (5.16)
where S(s) and T(s) are known as sensitivity and complementary sensitivity functions respectively. The matrix R(s) does not have a name yet. 107

The maximum singular value of the sensitivity function [S (j )] determines the disturbance attenuation, because S (s) is, in fact, the closed-loop transfer function from load disturbances to process output. Thus, a disturbance attenuation performance specication can be written as 1 [S (j )] |W1 (j ) |. (5.17)
1 Allowing the weighting function |W1 (j ) | to depend on frequency enables to specify a different attenuation factor for each frequency . With H control, the performance can be achieved by nding the controller K (s) through solving the problem
||W1 (j ) S (j ) || <
(5.18)
where denotes the innite norm. The value is a constant which indicates the accuracy to which the optimal loop matches the desired loop shape. W1 (s) is a diagonal transfer matrix that has W1 (s) as the diagonal elements. The maximum singular value of the complementary sensitivity function [T (j )] is used to measure the stability margins with respect to M , as shown in Fig. 5.20. Assuming the additive uncertainty A (j ) = 0, taking [M (j )] to be the denition of the size of M (j ), the size of the smallest multiplicative destabilizing uncertainty M (j ) is 1 . (5.19) [M (j )] = [T (j )] The smaller is [T (j )], the greater will be the size of the smallest destabilizing multiplicative perturbation, and hence the greater will be the stability margin. A similar result is available for relating the stability margin with respect to the additive plant perturbations [A (j )] = 1 . [R (j )] (5.20)
As a consequence of (5.19) and (5.20), the stability margins of control systems are specied via singular-value inequalities such as
1 [R (j )] |W2 (j ) |
1 [T (j )] |W3 (j ) |.
(5.21)
It is common practice to lump the effects of all plant uncertainties into a single cti1 tious multiplicative perturbation M , i.e., allowing the weighting function |W3 (j ) | to depend on frequency to specify a different attenuation factor for each frequency . The stability margin can be achieved by nding the controller K (s) through solving the problem ||W3 (j ) T (j ) || < (5.22) where W3 (s) is also a diagonal transfer matrix that has W3 (s) as the diagonal elements. In order to guarantee closed-loop stability and at the same time to achieve desired control 108

performance under process uncertainties, the objectives of performance and robust stability can be simultaneously achieved by nding a controller K (s) that satises both (5.18) and (5.22). The solution to the optimal problem is often called H controller based on mixed performance and robustness objectives. The numerical methods for solving H optimization problem are usually complex. Thus, commercial softwares, such as MATLAB, are commonly used to ease such tasks. The major work for the designer is then to specify W1 (j ) and W3 (j ) to meet the design requirement of the control performance and robustness. Following the denition in (5.16), the sensitivity function S (s) and complementary sensitivity function T (s) have the relation S (s ) + T (s ) = I (5.23)
i.e., [S (j )] and [T (j )] cannot both be small at the same frequency. The relationship between S (s) and T (s) reects the inherent conict between control performance and robustness. Usually, this conict can be resolved by requiring [S (j )] to be small at low frequencies, and [T (j )] to be small at high frequencies, due to the fact that the control performance is more important in the low-frequency range, while measurement noise and process uncertainties are more often of high-frequency nature.3 With H design, the closed-loop system performance is basically dened by the weighting functions W1 (j ) and W3 (j ). Optimization algorithms are used to synthesize the controller with the bandwidth between the crossover frequency of W1 (j ) and W3 (j ). Similar to IMC design, the RHP transmission zero of the Jacobian transfer matrix has to be considered in specifying the bandwidth of the closed-loop system, i.e., the desired bandwidth should be lower than at least half of the location of the RHP zero. The basic principle for selecting weighting functions are to give W1 (j ) a low-pass property and W3 (j ) a high-pass property. It is necessary to ensure that the crossover frequency for the Bode plot of W1 (j ) is below the crossover frequency of W3 (j ), such that there is a gap for the desired loop shape to pass. Fig. 5.21 shows the Bode plots of W1 (j ) and W3 (j ), which have the following transfer functions + S s + AS S T s+ M W3 (s) = AT s + T W1 (s) =
s M
(5.24)
where M represents the desired bounds on ||S(j )|| and ||T(j )||, AS and AT are the desired disturbance attenuation inside bandwidth for S(j ) and T(j ), and S and T are the crossover frequencies of W1 (j ) and W3 (j ) respectively. Fig. 5.22 shows the active-power and alternating-voltage step responses of the linear closed-loop system of the nominal model where the nominal model is also chosen
Unfortunately, this is not fully true for VSC-HVDC applications, since the variations of SCRs of the ac system and power levels of the VSC-HVDC link create process uncertainties in the low-frequency range.
3
109
10
|W1 (j)|, |W3 (j)|
10
|W1 (j)|
|W3 (j)|
10
10
-1
10
10
10 (rad/sec)
10
10
Fig. 5.21 Bode plots (magnitude) of wighting functions W1 (j ) and W3 (j ). Parameters: M = 2.0, AS = 0.0005, AT = 0.03, S = 70 rad/sec, T = 130 rad/sec [solid: weighting function |W1 (j ) |, dashed: weighting function |W3 (j ) |].
From: Uref To: P
0.3
0.1 0.2 time (sec)
0.3
Fig. 5.22 Active-power and alternating-voltage step responses of the H controller with the nominal model.
110

as dened in Table 5.1. Similar to the IMC controller, the H controller also successfully decouples the cross-coupling between the two control channels. However, different from IMC, where a transparent algorithm was used, H control achieved the decoupling through optimization. Consequently, a slight cross-coupling can still be observed in Fig. 5.22. The parameters of the weighting functions W1 (s) and W3 (s) have been chosen in such a way that the H controller has the same response time as the IMC controller in order to compare the results from the two controllers. Fig. 5.23 and Fig. 5.24 show the robustness-test results of the H controller regarding variations of SCRs of the ac system and power levels of the VSC-HVDC link for rectier and inverter operations respectively. In these two sub-sections, two multivariable feedback design approaches, i.e., IMC and H control, have been applied for a VSC-HVDC link connected to a weak ac system. Time simulations are performed with the linear models to verify the robustness of the controllers regarding variations of SCRs of the ac system and power levels of the VSC-HVDC link. From the time-simulation results, no signicant difference is observed between the two controllers. In the next subsection, a comparison in the frequency domain is given.
5.3.3 Performance and robustness comparison

To show the advantages of the multivariable feedback designs proposed in the previous subsections, the diagonal controller proposed in Chapter 3 is used as a reference design, i.e., kp 0 s . (5.25) K (s) = 0 ksu Fig. 5.25 shows the active-power and alternating-voltage step responses of the VSCHVDC converter using the diagonal controller. The nominal model is chosen to be the same as that is used for the IMC and the H control design, i.e., with the parameters given in Table 5.1. The control parameters kp and ku are tuned so as to give similar response times as the other two controllers. Fig. 5.26 and Fig. 5.27 show the robustness tests for the VSC-HVDC converter operating in rectier mode and inverter mode respectively. Generally speaking, the timedomain responses of the diagonal controller are more oscillatory than those of the other two controllers since the diagonal controller disregards the cross-coupling between the two control loops. Fig. 5.28 shows the comparison of the above three controllers by plotting the maximum singular values of their sensitivity functions [S (j )] together with the magnitude 1 of the weighting function W1 (s) that was used to synthesize the H controller. The values of [S (j )] for the IMC and H controllers are almost identical, which is in 111
From: Pref To: P 1.5 P (p.u.) P (p.u.) 1 0.5 0 -0.5 0 0.1 0.2 time (sec) From: Pref To: Uf 1 0.5 0 0 0.1 0.2 time (sec) 0.3 1 0.5 0 0 0.3 1.5 1 0.5 0 -0.5 0
From: Uref To: P
0.3
Uf (p.u.)
Uf (p.u.)
0.1 0.2 time (sec)
0.3
Fig. 5.23 Robustness tests by active-power and alternating-voltage step responses with the H controller for the VSC-HVDC converter operating in inverter mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
From: Uref To: P
0.3
0.1 0.2 time (sec)
0.3
Fig. 5.24 Robustness tests by active-power and alternating-voltage step responses with the H controller for the VSC-HVDC converter operating in rectier mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
112
From: Uref To: P
0.3
0.1 0.2 time (sec)
0.3
Fig. 5.25 Active-power and alternating-voltage step responses of the diagonal controller with the nominal model. Control parameters: kp = 50 rad/s, ku = 60.
From: Uref To: P
0.6
0.2 0.4 time (sec)
0.6
Fig. 5.26 Robustness tests of the diagonal controller for the VSC-HVDC converter operating in rectier mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
113

From: Pref To: P 1.5 P (p.u.) 1 0.5 0 -0.5 0 0.2 0.4 time (sec) From: Pref To: Uf 1 Uf (p.u.) 0.5 0 0 0.2 0.4 time (sec) 0.6 Uf (p.u.) 1 0.5 0 0 0.2 0.4 time (sec) 0.6 0.6 P (p.u.) 1.5 1 0.5 0 -0.5 0 0.2 0.4 time (sec) From: Uref To: Uf 0.6 From: Uref To: P
Fig. 5.27 Robustness tests of the diagonal controller for the VSC-HVDC converter operating in inverter mode. Model variations: SCR = 1, 2, 5. P = 0.0, 0.5, 1.0 p.u.
agreement with the step responses shown in Fig. 5.17 and Fig. 5.22. However, the value of [S (j )] of the diagonal controller is not bounded by W 1 (s), which implies that the performance of the diagonal controller is worse than the performance requirement specied by W 1 (s), especially with the peak at around = 80 rad/s. This also generally agrees with the time-domain performance. The step response shown in Fig. 5.25 for the diagonal controller, is more oscillatory than the step responses shown in Fig. 5.17 and Fig. 5.22 for the IMC and H controllers. The oscillation mode of the diagonal controller is at around = 80 rad/s. Fig. 5.29 shows the comparison of the above three controllers by plotting the maximum singular values of their complementary sensitivity functions [T (j )] together 1 with the magnitude of the weighting function W3 (s) that was used to synthesize the H controller. The values of [T (j )] for the IMC and H controllers are almost identical, but the IMC controller, in fact, shows even slightly lower values at frequencies above = 400 rad/s. Of course, this difference cannot be reected by the robustness tests in the time domain since a = 3 ms ( = 333.3 rad/s) low-pass lter has been applied in the test model [cf. (5.15)]. The frequency-domain comparison conrms the similarity of robustness of the two controllers by the time-domain tests. In Fig. 5.29, it shows that the value of [T (j )] of the diagonal controller is not bounded by the weighting function 1 W3 (s), which implies that the robustness of the diagonal controller is worse than the 1 robustness requirement specied by W3 (s), mainly at the peak around = 80 rad/s. By comparing Fig. 5.28 with Fig. 5.29, both [S (j )] and [T (j )] for the diagonal controller have a high peak around = 80 rad/s, which seem to be contradictory to the 114
Diagonal controller
|W (j)|
-1 1
10
H controller [S(j)]
IMC controller 10
-1
10
-2
10
10
10 (rad/sec)
10
10
Fig. 5.28 Maximum singular values of the sensitivity functions for nominal performance comparison [solid: H controller, dashed: IMC controller, dotted: diagonal controller, thick 1 line: weighting function |W1 (j ) |].
|W-1 (j)| 3
Diagonal controller
10
[ T(j)]
H controller
10
-1
IMC controller
-2
10
10
10
10 (rad/sec)
10
10
Fig. 5.29 Maximum singular values of the complementary sensitivity functions for robustness comparison [solid: H controller, dashed: IMC controller, dotted: diagonal controller, 1 (j ) |]. thick line: weighting function |W3
115

constraint in (5.23). However, since S(s) and T(s) are transfer matrices, they also differ by the phases of the transfer functions. Thus, the real meaning of the constraint in (5.23) is that [S (j )] and [T (j )] cannot both be small at the same frequency, but they can certainly both be large at the same frequency. In fact, it can be mathematically proven that [S (j )] and [T (j )] differ at most by 1 [91], i.e., a very high [S (j )] always implies a very high [T (j )]. In practice, for a controller that yields very poor performance with the nominal model, it certainly implies that the controller would also have poor robustness for model variations! Remark 1: It is remarkable that the IMC controller solved by the simple algorithm achieves similar performance and robustness (or even slightly better robustness in the high-frequency range) as the more advanced optimization-based H control. Generally speaking, IMC design might yield poor controllers for processes that have poorly damped poles, since IMC simply inverses the process while the poorly damped poles, in fact, still remain in the system. However, with power-synchronization control, this problem is avoided, since the high-pass current control function HHP (s) has shifted all the resonant poles of the Jacobian transfer matrix towards the left-half plane (cf. Fig. 4.11). On the other hand, compared to H control, the order of the controller solved by IMC is lower. With the nominal model used in this section, IMC design yields a controller only of 4th order while H design yields a controller of 8th order. Remark 2: The sensitivity function S(s) is a good indicator of the performance of the closed-loop system. However, it is only valid for the performance of the nominal model. Meanwhile, the bounded complementary sensitivity function T(s) only ensures robust stability, i.e., the closed-loop system is only stable within the specied range of model uncertainties. Thus, these two functions cannot be used as indicators of the performance of the worst-case scenario, i.e., robust performance. To synthesize controllers that full robust performance, the structured singular value, i.e., the , needs to be applied [92]. However, controllers synthesized by can be conservative for this application since the variations of SCRs of the ac system and power levels of the VSC-HVDC link are real-number uncertainties, while is only tight for complex-number perturbations [93], e.g., model uncertainties in the high-frequency range. On the other hand, in this section, it has been demonstrated that the robustness tests performed in the time domain provides a good overview of the performance of the closed-loop system with all the possible scenarios. Accordingly, the frequency-domain approach and the time-domain approach should be used as complementary tools for controller design and evaluation. Remark 3: In this section, the controllers are designed and tested based on the simple impedance-source representation of the ac system. In reality, the topology of the ac system is, of course, much more complex. However, given the same SCR and the same operating point of the VSC-HVDC converter, such variations with the process model appear mainly in the high-frequency range. These types of uncertainties can easily be dealt 116
5.4. Direct-voltage control

with by feedback control, especially with the high-pass current control function applied. Therefore, for VSC-HVDC applications, the major challenges for robustness of the controller are the variations of SCRs of the ac system and power levels of the VSC-HVDC converter. One exception is that, if there is a series-compensated ac line feeding into the converter station, the series-compensated ac line should be included in the nominal model for the controller design instead of being treated as a model uncertainty, since the characteristics of a series-compensated ac line can be special as shown in Chapter 4. Another situation is that, if there are other large input devices in the vicinity of the converter stations and their dynamic interactions with the VSC-HVDC link are of concern, then an equivalent system is necessary to be developed to properly reect the topology of the ac system.
5.4 Direct-voltage control

For VSC-HVDC applications, at least one of the converter stations has to control the direct voltage, while the other converter station controls the active power. The active power is thus automatically balanced between the two converter stations. In this section, various aspects of direct-voltage control are discussed.
5.4.1 Controller design

The block diagram of the direct-voltage controller is shown in Fig. 5.30, where the multivariable power-synchronization controller is used as an inner loop. For the direct-voltage controller, a PI controller is proposed with the control law Pref = Kpd +
Fdc (s)
Kid s
2 2 (uref dc ) udc .
(5.26)
If the power-synchronization controller is assumed to be sufciently fast, i.e., Pref = PVSC , and the dc link is assumed to be a single capacitor, i.e., Gdc (s) = 2/(sCdc ) [cf. (4.75)], then the closed-loop system is expressed as u2 dc = s2 C 2(Kpds + Kid ) 2 (uref dc ) . + 2 K s + 2 K dc pd id (5.27)
The control parameters are selected as

2 Kpd = d Cdc and Kid = d Cdc /2
(5.28)
which places two real poles of the closed-loop system at s = d . It is generally difcult to achieve reference tracking and disturbance reduction by a single feedback controller. 117

Power-synchronization control inner loop
K 22 (s) U ref f1 + K 21 (s) + + V V0 U f 1 AC-Jacobian transfer PVSC1 matrix

J(s) P 1
ref P 1
K12 (s) +K11 (s)
v + +
DC-Jacobian PVSC2 transfer matrix G dc ( s)

2 udc1 2 udc2
Fdc1 (s)
ref 2 (udc1 )
+ K r (s)
Fig. 5.30 Control block diagram of the direct-voltage controller.

2 The solution is to use a two degree-of-freedom controller, where the reference (uref dc ) is connected to a prelter to improve the reference tracking while the controller Fdc (s) is only tuned for disturbance reduction. A convenient practical choice of prelter is the lead-lag network 1 + sT1 Kr ( s ) = (5.29) 1 + sT2 where T1 > T2 is chosen to speed up the response and T1 < T2 is chosen to slow down the response. The effect of the prelter is demonstrated in Fig. 5.31, which shows that the overshoot of the step response of the direct-voltage controller is removed by the prelter. As it has been shown in Chapter 4, for long overhead lines, the dc-line inductance and the dc capacitors may create a resonance peak in the low-frequency range. If the resonance appears within the bandwidth where the direct-voltage controller is active, an effective way to mitigate its impact is to add a notch lter. A notch lter commonly has the expression 2 s2 + 2 1 n s + n Fn (s) = 2 (5.30) 2 s + 2 2 n s + n
where the three adjustable parameters are 1 , 2 , and n . The ratio of 2 /1 sets the depth 118
1 (p.u.) u 0 0 0.05 0.1 time (sec) 0.15 0.2
Fig. 5.31 Step response of the direct voltage. The overshoot of the direct voltage is removed by a prelter (solid: without prelter, dashed: with prelter). Direct-voltage controller: d = 40 rad/sec. Prelter: T1 = 0.075 sec, T2 = 0.0925 sec.
of the notch, and n is the resonance frequency. Fig. 5.32 shows the effect of the notch lter by the open-loop transfer function of the direct-voltage control, which has the expression Hdc (s) = Fdc (s)Fn (s)Gdc11 (s). (5.31)
where the parameters of Gdc11 (s), i.e., Gdc (s) are given in Table 4.4. It should be noted that the notch lter may adversely affect the phase margin of the direct-voltage controller, as shown in Fig. 5.32.
5.4.2 DC-capacitance requirement

In Section 5.4.1, it has been shown that, by explicitly including the dc capacitance in the parameters of the direct-voltage controller, one can freely place the poles of the closedloop system. This implies that, as long as its size is known, the dc-capacitance does not affect the linear stability of the direct-voltage control. However, for disturbance reduction, the dc capacitance has an important role. In this section, the dc-capacitance requirement for weak-ac-system connections is discussed. The applied method follows the procedure introduced by [85]. To simplify the analysis, the dc link is assumed to be the single dc-capacitor model, i.e., Gdc (s) = 2/(sCdc ). If the power-synchronization loop has a bandwidth that is considerably higher than that of the direct-voltage control, the transfer function from the load disturbance to the error signal becomes Gpe (s) =
2 sCdc s+ 2d d Cdc s
dc
0.5
1+
2 sCdc
2s . Cdc (s + d )2
(5.32)
If the worst scenario is considered, i.e., the active-power output to the other converter station is changed stepwise from the maximum active power Pm to 0, e.g., the converter is suddenly blocked. The step response of the error signal edc (t) in the time domain 119

10
2
|Hdc(j)|
10
10
-2
10 100 arg Hdc(j) (deg.) 0 -100 -200
10
10
-300 1 10
10
10
(rad/sec)
Fig. 5.32 Bode plots of Hdc (s) to show the effect of the notch lter for reducing the dc resonance peak (solid: without notch lter, dashed: with notch lter). Direct-voltage controller: d = 40 rad/sec. Notch lter: n = 322 rad/sec, 1 = 0.2, 2 = 0.8.
becomes edc (t) = L1 Gpe (s) = L 1 The time derivative of edc (t) is dedc 2 Pm = (1 d t) ed t dt Cdc (5.34) Pm s = 2Pm d te t . Cdc (5.33)
2 Pm Cdc (s + d )2
which has a local maximum for t = 1/d . By substituting this into (5.33), the maximum error is found to be 2 Pm 1 edc,max = e . (5.35) d Cdc By considering udc,max the maximum direct-voltage allowed, then edc,max = u2 dc,max ref 2 (udc ) , the required dc capacitance is Cdc > 2Pm e1 . ref 2 [u2 dc,max (udc ) ]d (5.36)
A common expression for dc capacitance is using its time constant Tdc , which is dened as 2 Cdc0 UdcN Cdc Tdc = = (5.37) 2PdcN 2 120

where UdcN is the rated (base) direct voltage (kV) and PdcN is the rated (base) power (MW) of the VSC-HVDC link, and Cdc0 is the dc capacitance (F). Substituting (5.37) into (5.36) yields Pm e1 Tdc > 2 . (5.38) 2 [udc,max (uref dc ) ]d
It should be noted, for the dc-capacitance requirement, only per unit values of the direct voltage and the dc power are applicable for (5.38), while either per unit values or real values can be applied in (5.36). For weak-ac-system connections, the VSC-HVDC link needs to operate with large load angles, where the RHP transmission zero of the ac Jacobian transfer matrix J(s) moves closer to the origin. The RHP zero limits the bandwidth of the power-synchronization controller, and eventually limits the bandwidth of the direct-voltage control. A rule of thumb is that the bandwidth of the power-synchronization controller should be lower than half of the location of the RHP zero [82]. If the voltages of the ac source and the lter bus are assumed to be nominal, i.e., E0 1.0 p.u. and Uf 0 1.0 p.u., the location of the RHP zero can be simplied from (4.17) as s = 1 E0 cos u0 1 Uf 0 E0 cos u0 cos u0 . 1 cos u0 (5.39)
Based on the denition of SCR in (5.3), the well-known power-angle equation between the ac source and the lter bus can be expressed as P = E0 Uf 0 sin u0 SCR sin u0 . 1 Ln (5.40)
If it is further assumed that the direct-voltage control is four times slower than the powersynchronization loop, then d can be solved by (5.39) and (5.40). Accordingly, 1 d < 1 8 cos u0 1 = 1 1 cos u0 8
Pm 2 1 ( SCR )
Pm 2 1 ( SCR )
(5.41)
Another issue that needs to be taken into account is that, if the bandwidth of the directvoltage controller is chosen four times slower than the power-synchronization loop, the inner loop will affect the maximum voltage variation. Fig. 5.33 shows the ratio of bandwidth reduction by the inner loop, which is approximated by a rst-order lter with bandwidth p . Considering this effect, (5.38) is adjusted as Tdc > 1.3Pm e1 . ref 2 [u2 dc,max (udc ) ]d (5.42)
Given the maximum loading Pm and maximum allowed direct voltage udc,max , the relationship between dc-capacitance requirement and the SCR of the ac system can be established by (5.41) and (5.42). If the worst scenario is considered, the maximum loading 121

2.5
dc,max dc,dmax
2 1.5 1 1
/e
5 /
p d
10
Fig. 5.33 Bandwidth reduction by the inner loop.

30 25 (ms)
dc
20 15 10 5 1 1.2 1.4 SCR 1.6 1.8 2
Fig. 5.34 DC-capacitance requirements for weak-ac-system connections with uref dc = 1.0 p.u. (solid: udc,max = 1.2 p.u., dashed: udc,max = 1.3 p.u., dotted: udc,max = 1.4 p.u.).
should be chosen as Pm = 1.0 p.u. However, as it was mentioned before, for very weakac-system connections, it is recommended that the load angle shall not be above 60 to maintain a reasonable stability margin. For instance, if the SCR of the ac system is 1.0, then the maximum loading is Pm = SCR sin u0 = 0.86 p.u. Fig. 5.34 shows the plots of dc-capacitance requirements for ac systems with SCR 2.0 with different allowed udc,max .
5.5 Interconnection of two very weak ac systems

In this section, interconnection of two very weak ac systems by a VSC-HVDC link is investigated by one design example and followed by linear-model verication and ac-fault studies simulated by PSCAD/EMTDC. In the design example, a 350 MW VSC-HVDC link is applied to interconnect two very weak ac systems, where the ac system connected to the inverter side (station 1, power receiving end) has network strength SCR1 = 1.0, while the ac system connected to the rectier side (station 2, power sending end) has network strength SCR2 = 1.2. The detailed parameters of the converter stations are given in Appendix C. From Fig. 5.30, it can be easily observed that, if one converter station controls the direct voltage, while the other converter station controls the active power, the two 122
5.5. Interconnection of two very weak ac systems

U ref f1
ref P 1
DC-Jacobian PowerPower PVSC1 transfer PVSC2 synchronization synchronization matrix inner loop P inner loop 1 G dc ( s)
2 udc1 2 udc2
U ref f2 P2ref
Fdc1 (s) + K r1 (s)
Fdc2 (s) + K r2 (s)
P +
ord 1
Fpc (s)
+
ref 2 (udc1 )
+ (u )
ref 2 dc1
ref 2 (udc2 )
Fig. 5.35 Control block diagram of a VSC-HVDC link interconnecting two weak ac systems.
converter stations are, in fact, linearly independent. This implies that the stability of one converter station does not affect the stability of the other converter station. One might consequently conclude that there is nothing more special for a VSC-HVDC link connected to two weak systems than it is only connected to one weak system. This might be true if only linear effects are considered. However, the real system is non-linear. For VSCHVDC operations, the direct voltage has to be carefully maintained around its nominal value. For instance, a big direct-voltage drop might temporarily limit the capability of the alternating-voltage controller and negatively affect the linear stability of the closed-loop system. The proposed control structure for weak-ac-system interconnection is shown in Fig. 5.35, where the dashed block represents the power-synchronization control inner loop in Fig. 5.30. The basic idea of the design is that both of the two converter stations have direct-voltage controllers, while the active-power controlling station controls the active power by adding an additional contribution to the reference of the direct-voltage controller and its output shall be limited. With the proposed control structure the linear independence between the rectier station and the inverter station is lost. However, the bandwidth of the direct-voltage controllers is much higher than the bandwidth of the acsystem dynamics. This implies that the rewall effect of the VSC-HVDC link is still in force. In the following, the detailed procedure for controller design and parameter settings are given. The linear model is compared to the simulation results from PSCAD/EMTDC for each major design step. 1. Power-synchronization inner loop. Due to the very low SCR of the inverter ac system, the maximum allowed power is Pm1 = 0.86 p.u., which corresponds to approximately u10 = 60 load angle. The ac system at the rectier side has a slightly 123

higher SCR, but considering the losses of the converter station and of the dc cable, it has a maximum loading Pm2 = 0.91 p.u., which corresponds approximately to a load angle of u20 = 50 . The power-synchronization controller is used as the inner loop of the direct-voltage controller as previously shown in Fig. 5.30. For this application, the IMC controller investigated in Section 5.3 is applied to parameterize the power-synchronization controller. According to the discussion in Section 5.4.2, the bandwidth of the powersynchronization controller should be lower than half of the RHP-zero location of the ac Jacobian transfer matrix. By using (5.39), the bandwidths of the activepower controllers of power-synchronization control at the two converter stations can be calculated as p1 = 160 rad/s and p2 = 210 rad/s. The bandwidths of the alternating-voltage controllers, however, do not need to be very high. Thus, they have been chosen as u1 = u2 = 50 rad/s. 2. Direct-voltage controller. As mentioned in Section 5.4.1, the bandwidth of the direct-voltage controllers should be chosen at least four times slower than the activepower controller of the power-synchronization control. Alternatively, the bandwidth of the direct-voltage controllers can be calculated directly by substituting the SCR and maximum loading into (5.41), which yields d1 = 39 rad/s and d2 = 53 rad/s. However, the linear analysis shows that the direct-voltage controller at the rectier side (station 2) tends to have less phase margin for the same load angle than that at the inverter (station 1) side. Therefore, the bandwidths of both of the direct-voltage controllers have been chosen as d1 = d2 = 40 rad/s. Fig. 5.36 shows the step response of the direct voltages at the two converter stations with the comparison between the linear model and the nonlinear model in PSCAD/EMTDC. The maximum allowed direct voltage is 1.3 p.u. Based on the curves of the dccapacitance requirement in Fig. 5.34, the dc capacitor has been chosen to make the total dc capacitance Tdc = 15 ms. Fig. 5.37 shows the plots of disturbance reduction of the direct-voltage controllers at the two converter stations with the comparison between the linear model and the nonlinear model in PSCAD/EMTDC. The maximum direct-voltage peak in Fig. 5.37(b) is below the specication udc,max = 1.3 p.u. The discrepancies between the results of the linear models and the nonlinear simulations are due to that the results of the linear model are obtained at one operating point, while the disturbance reduction in PSCAD/EMTDC involves operating-point changes. The tests in PSCAD/EMTDC are performed by blocking the converter at the other station with maximum loading. 3. Active-power controller. In principle, either of the VSC-HVDC converter stations can be the power-controlling station. In this example, the inverter station (station 1) controls the active power by a proportional-integral (PI) controller with the control 124

1.15 udc1 (p.u.) 1.1 1.05 1 0.95 0 0.05 0.1 time (sec) 0.15 0.2
(a)
1.15 (p.u.) u
dc2
1.1 1.05 1 0.95 0 0.05 0.1 time (sec) 0.15 0.2
(b)
Fig. 5.36 Step response of the direct-voltage controllers (solid: nonlinear simulation, dashed: linear model). (a) Direct-voltage step at station 1. (b) Direct-voltage step at station 2.
1 udc1 (p.u.) 0.8 0.6 0 0.1 0.2 time (sec) 0.3 0.4 0.5
(a)
1.4 1.3 udc2 (p.u.) 1.2 1.1 1 0 0.1 0.2 time (sec) 0.3 0.4 0.5
(b)
Fig. 5.37 Disturbance reductions of the direct-voltage controllers with maximum load changes (solid: nonlinear simulation, dashed: linear model). (a) Converter block at station 2. (b) Converter block at station 1.
125

law
2 (uref d1 ) = Kpp (1 + Fpc (s)
Kip ord P1 . ) P1 s
(5.43)
The parameters of the active-power controller are tuned by the root-locus technique in two steps: Step 1: Start with tuning the proportional gain Kpp by applying P control, i.e., Fpc (s) = Kpp . Step 2: The integral gain Kip is tuned by applying the full PI controller. Fig. 5.38 shows the root-loci of the closed-loop system by applying the proportional controller. By varying Kpp from 0.0 to 1.0, three dominant pole pairs are affected. The pole pair p5,6 (p5 = 10.8 rad/s, p6 = 0.296 rad/s) is shifted towards the left-half plane, which can be viewed as the stabilizing effect of feedback control. However, both the pole pairs p1,2 (p1,2 = 63.7 j 232 rad/s) and p3,4 (p3 = 37.1 rad/s, p4 = 50 rad/s) are shifted towards the right-half plane. The frequency-domain analysis shows that p1,2 is related to the gain margin, while p3,4 is related to the phase margin of the active-power control. Kpp = 0.6 is chosen to get a balance of stability and response time, which places the three pole pairs at p1,2 = 53 j 227 rad/s, p3,4 = 27 j 24 rad/s, and p5,6 = 15.1 j 3 rad/s. Fig. 5.39 shows the root-loci of the closed-loop system by applying the PI-type active-power controller, where p0 is a pole introduced by the integral controller. By varying Kip from 0.0 to 60, p0 moves quickly towards the left-half plane. The pole pair p1,2 is rather insensitive to the variation of Kip , and p5,6 move towards two lefthalf plane zeros on the real axis. However, p3,4 is negatively affected. Frequencydomain analysis also shows that the phase margin of the active-power control is reduced by increased integral gain Kip . However, a larger integral gain is necessary in reducing the power-recovery time after ac-system faults and minimizing the steady-state error. Finally, Kip = 30 is chosen which places the three dominant pole pairs at p1,2 = 54 j 225 rad/s, p3,4 = 13.2 j 29.4 rad/s, p5 = 28 rad/s, and p6 = 10 rad/s. Fig. 5.40 shows the step response of the active-power control at low and high power levels respectively. The step response at the high power level corresponds to the operating point applied for the root-locus tuning. It should be noted that the output of the active-power controller should be limited to avoid too large direct-voltage variations. In this example, the limitation of the output of the active-power controller is chosen as 0.25, which corresponds to approximately +12% and 13% direct-voltage variations. 126
250 200 150 100 (rad/sec) 50 0 -50 -100 -150 -200 -250 -80 -70 p
2
p6 p
4
-60
-50
-40 -30 (rad/sec)
-20
-10
10
Fig. 5.38 Root-loci of the dominant poles of the closed-loop system by applying P-type activepower control. Kip = 0.0, variations of Kpp from 0 to 1.0.
250 200 150 100 (rad/sec) 50 0 -50 -100 -150 -200 -250 -80 -70 -60 p
2
-50
-40 -30 (rad/sec)
-20
-10
10
Fig. 5.39 Root-loci of the dominant poles of the closed-loop system by applying PI-type activepower control. Kpp = 0.6, variations of Kip from 0 to 60.
127
P1 (p.u.)
0.8 0.75 0.7 0
0.1
0.2 time (sec)
0.3
0.4
0.5
(a)
P1 (p.u.)
0.1 0.05 0 0
0.1
0.2 time (sec)
0.3
0.4
0.5
(b)
Fig. 5.40 Step response of the active-power controller at low and high power levels at converter station 1 (solid: nonlinear simulation, dashed: linear model). Active-power controller: Kpp = 0.6, Kip = 30. (a) Step from P1 = 0.7 p.u. to P1 = 0.8 p.u. (b) Step from P1 = 0.0 p.u. to P1 = 0.1 p.u.
The fault ride-through capability of the VSC-HVDC link is tested in PSCAD/EMTDC by applying three-phase ac-system faults at both of the converter stations. The VSC-HVDC link initially operates with the maximum loading, i.e., P1 = 0.86 p.u. The VSC-HVDC link is supposed to ride through ac-system faults without relying on telecommunications between the two converter stations. In Fig. 5.41, a three-phase ac fault with 0.2 sec duration is applied at the inverter station (station 1), i.e., the power-controlling station in this example, at 0.1 sec. One consequence of the ac fault is that the direct voltage increases to approximately 1.3 p.u. due to the loss of power output. The rectier station (station 2) brings down the direct voltage to its nominal value after an initial overshooting. Another consequence of the ac fault is the increase of the modulus of the valve current |ic1 |. After detecting the fault, the current limiter reduces the converter current to half of the maximum load current Imax (or any other desired value) except a very short current spike at the fault inception. Fig. 5.42 shows a three-phase ac fault with 0.2 sec duration applied at the rectier station, i.e., the direct-voltage controlling station in this example, at 0.1 sec. During the ac-system fault, the power-controlling station controls the direct voltage to a lower voltage level which is 13% less than the nominal value. The 13% is a result of the limitation of the active-power controller. Since the ac fault is applied at the rectier side, there is no over128

1 P1 (p.u.) 0.5 0 0 1.5 udc1 (p.u.) 1 0.5 0 0 2 | ic1| (p.u.) 1 0 0 0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
0.2
0.4 time(sec)
0.6
0.8
Fig. 5.41 Fault ride-through capability of the VSC-HVDC link during a three-phase ac-system fault at the inverter station (station 1). Upper plot: active power from the VSC-HVDC link at station 1. Middle plot: direct voltage at station 1. Lower plot: valve current at station 1.
1 P1 (p.u.) 0.5 0 0 1.5 udc1 (p.u.) 1 0.5 0 0 2 | ic1| (p.u.) 1 0 0 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
0.2
0.4 time (sec)
0.6
0.8
Fig. 5.42 Fault ride-through capability of the VSC-HVDC link during a three-phase ac-system fault at the rectier station (station 2). Upper plot: active power from the VSC-HVDC link at station 1. Middle plot: direct voltage at station 1. Lower plot: valve current at station 1.
129

current problem at the inverter station (station 1). A specic amount of active power ows in the reverse direction during the fault period. However, the current limitation controller at the rectier (station 2) limits the fault current to half of the maximum load current.
5.6 Summary
In this chapter, the control and modeling issues for VSC-HVDC links connected to highimpedance ac systems are investigated. Power-synchronization control and vector current control are compared for VSC-HVDC links connected to weak ac systems. It is concluded that, for weak-ac-system connections, the voltage variations at the lter bus are much less for power-synchronization control than for vector current control. Thus, powersynchronization control is the most suitable controller for VSC-HVDC links connected to weak ac systems. Two multivariable feedback designs, i.e., internal model control and H control, are applied and compared with the simple diagonal controller. The multivariable feedback design show clear advantages in control performance and robustness. The IMC controller is simple yet achieves similar results as the more advanced H controller. A two-degree-of-freedom direct-voltage control with a prelter to reduce the overshoot is proposed. For VSC-HVDC links connected to weak ac systems, a higher value of dc capacitance is required. A control structure for interconnecting two very weak ac systems is proposed. As an example, it is shown that a VSC-HVDC link using powersynchronization control enables a power transmission of 0.86 p.u. from a system with an SCR of 1.2 to a system with an SCR of 1.0.
130
Chapter 6 Control of VSC-HVDC Links Connected to Low-Inertia AC Systems

In this chapter, power-synchronization control is applied to control VSC-HVDC links connected to low-inertia ac systems, or island systems. In Section 6.1, general aspects of low-inertia ac systems are described. In Section 6.2, a frequency droop controller and an alternating-voltage droop controller are proposed. In Section 6.3, a typical island system which includes four common power components, i.e., a synchronous generator, an induction motor, some passive RLC loads, and a VSC-HVDC link is modeled by the Jacobian transfer matrix modeling concept. In addition, the linear model is applied for controlparameter tuning using root-locus techniques. In Section 6.4, simulation studies are performed with PSCAD/EMTDC to demonstrate the exibility of power-synchronization control for various network conditions. In Section 6.5, the Jacobian transfer matrices for a synchronous generator and an induction motor are developed, and their characteristics are analyzed. Finally, the subsynchronous characteristics of a VSC-HVDC converter are analyzed by the frequency-scanning method in Section 6.6. The major results of this chapter are summarized in Section 6.7. Some results of this chapter are included in [94].
6.1 General aspects of low-inertia ac systems

In this chapter, various aspects of a VSC-HVDC link connected to another type of weak ac systems, i.e., low-inertia or island systems, are investigated. Low-inertia systems are considered to have limited number of rotating machines in the system, or no rotating machine at all. Examples of such applications can be found when an HVDC link is powering an island system, or if it is connected to a windfarm. It may also be the case when an HVDC link initially is connected to a large ac system, but comes into island operation due to tripping of critical ac lines to the large system, or black start after a blackout. For the conventional LCC-HVDC system, the commutations of the converter valves depend on the ability of the ac system to maintain the required voltage and frequency. 131
Chapter 6. Control of VSC-HVDC Links Connected to Low-Inertia AC Systems

Therefore, the conventional LCC-HVDC system has a requirement on the minimum inertia of the connected ac system. Similar to the denition of SCR, it was proposed in [5] to use the effective inertia constant Hdc as a measure of the relative rotational inertia, which is dened as Hac Hdc = (6.1) PdcN where Hac (MW s) is the total rotational inertia of the ac system and PdcN is the rated dc power. For a conventional LCC-HVDC link connected to an island system, an effective inertia constant, Hdc , of at least 2.0 s to 3.0 s is required for satisfactory operation [32]. Synchronous condensers (SCs) have been the traditional means to increase the inertia of an island system [95]. Since SCs also increase the short-circuit capacity of the ac system, the weak-ac-system problem of the conventional LCC-HVDC system, either low SCR or low inertia, can generally be solved by installation of SCs. However, SCs can substantially increase the investment and maintenance costs of an HVDC project. In this respect, the VSC-HVDC technology has a clear advantage over the conventional LCC-HVDC technology. Different from the thyristor valves, where relatively stiff voltage and frequency of the ac system are pre-conditions for valve commutations, the VSC can produce its own alternating-voltage waveform independent of the ac system. Thus, a VSC-HVDC link can even be connected to a passive network with no other power source at all [96, 97]. However, control of VSC-HVDC links connected to an island system is technically different from connected to a large ac system for the following reasons: 1. Due to the relatively limited number of power generating units in a typical island system, the VSC-HVDC link should participate in frequency control of the island system instead of applying constant power control. 2. An island system with low inertia is often of small geographic extent and the generators and loads are found electrically close to the converter station. More interactions are expected between the VSC-HVDC converter and other input devices, e.g., the risks of voltage-control hunting and SSTI with local synchronous generators are higher. 3. In some situations, island operation is a consequence of ac-system failures by losing critical ac lines to a large ac system. The control system of the VSC-HVDC link is required to be able to handle the transition of operation modes seamlessly. In the following sections, the above issues related to low-inertia ac systems are investigated for a VSC-HVDC link using power-synchronization control. 132
6.2. Controller design

Pref + P
ref
+-
Kf 1 + sT f
Pref
kp s
+-
Island system
s 1 + sTm
Fig. 6.1 Frequency droop controller.
6.2 Controller design

In this chapter, it is assumed that the VSC-HVDC link is connected to a strong system at the other converter station, which controls the direct voltage of the dc link constant disregarding the power variations of the converter connected to the island system. Thus, only the controllers of the converter station connected to the island system are discussed. In Chapter 5, multivariable feedback designs have been applied for VSC-HVDC links connected to high-impedance ac systems. However, such designs are not considered necessary for low-inertia system connections, since the bandwidth requirement for the control system is generally not as critical for island operation. Thus, to simplify the control design discussions, the simple diagonal controller in (5.25) is applied. The tuning of the control parameters kp and ku will be further discussed in Section 6.3 using the root-locus techniques.
6.2.1 Frequency droop control

For a VSC-HVDC link connected to an island system, the VSC-HVDC converter shall not have constant power control but rather participate in frequency control of the island system. A frequency controller of the following form is proposed Pref = Kf 1 + Tf s [ref ] (6.2)
where ref is the reference and is the measured value of the angular frequency. The output of the frequency controller is added to the power reference of the active-power controller, as shown in Fig. 6.1. In the gure, the frequency measurement is taken by the derivative of the angle output of the active-power controller with a measuring time constant Tm , typically in the range of 10 20 ms. Kf and Tf are the gain and time constant of the frequency controller. Such a design gives the VSC-HVDC link a frequency droop characteristic. If there are any other power generating units in the island system, the load sharing is determined 133

Rtg + jX tg PCC
SG
ig
+
eg Rtv + jX tv P 1 , Q1
VSC
iv
+
uf -
P2 , Q2
P, Q
Fig. 6.2 Voltage droop control for parallel connected voltage-control units.
by the frequency droop of each power generating unit, which is dened as R = /P . For two power generating units with frequency droops R1 and R2 , the following relationship is established for the power outputs P1 and P2 P1 R2 = . P2 R1 (6.3)
Accordingly, the power generating unit with smaller frequency droop shares more loads in the island system. The frequency droop of the VSC-HVDC link with the proposed frequency controller in (6.2) can be expressed as =
1 kp
1 (Pref P ). + Kf
Rvsc
(6.4)
Thus, the frequency droop of the VSC-HVDC link depends on both Kf and kp . The time constant Tf should be chosen similar to the time constant of the turbines of the local generators in the island system.
6.2.2 Alternating-voltage droop control

If two or more alternating-voltage controlling units are connected to a common bus, the alternating-voltage controls should be coordinated to avoid voltage hunting between the units. Fig. 6.2 shows a VSC-HVDC link connected in parallel with a synchronous generator with step-up transformers. In Fig. 6.2, eg , uf are the terminal/lter-bus voltage vectors of the generator, the VSC-HVDC converter. P and Q are the active power and reactive power to the island system. P1 and Q1 are the active power and reactive power from the synchronous generator, while P2 and Q2 are the active power and reactive power from the VSC-HVDC link. Rtg and Xtg are the resistance and leakage reactance of the step-up transformer of the synchronous generator, while Rtv and Xtv are the resistance and leakage reactance of the converter transformer of the VSC-HVDC converter. The solution to solve the voltage-control coniction between the synchronous generator and the 134
6.3. Dynamic modeling and linear analysis of a typical island system

PCC
Ltg
SG
ig
+
eg Ltm Ltv
Large ac system VSC
IM
im
+
em Ll Rl Cl +
+
ev -
iv
iL u pcc -
Passive load
AC network
Fig. 6.3 A typical island system.
VSC-HVDC converter is using the so-called load compensation, which is a common solution for synchronous generators terminated at the same bus [32]. That is, instead of controlling the alternating voltage of the point-of-common-coupling (PCC) bus, the VSCHVDC converter and the synchronous generator both control voltages between their own terminal/lter-bus voltages and the voltage of the PCC to give droop characteristics to their alternating-voltage controls. With load compensation, the magnitudes of the resulting compensated voltages Vcg of the synchronous generator and Vcv of the VSC-HVDC converter are Vcg =|eg + kg (Rtg + jXtg )ig |
Vcv =|uf + kc (Rtv + jXtv )iv |
(6.5)
where the compensation ratios kg and kc are typically chosen between 50% and 80%. By adjusting the compensating ratios, the reactive-power sharings between the VSC-HVDC converter and the synchronous generator are re-distributed.
6.3 Dynamic modeling and linear analysis of a typical island system

Fig. 6.3 shows an island system introduced by [98], which includes four common components in a typical island system, i.e., a synchronous generator (SG), an induction motor (IM), some passive RLC loads, and the VSC-HVDC link. The induction-motor loads are 135

included in the studied system since they are quite common in industrial systems [99]. The space vectors eg , em and ev are the terminal voltage vectors of the SG, the IM, and the VSC-HVDC converter, while ig , im and iv are the corresponding injecting current vectors. Ltg , Ltm , and Ltv are the leakage inductances of the step-up transformers of the SG, the IM, and the converter transformer of the VSC-HVDC converter. Rl , Ll and Cl are the resistance, inductance, and capacitance of the passive loads. iL is the current vector owing through Ll , and upcc is the voltage vector of the PCC. An ac line is connected to a large ac system at the PCC bus. This connection is intentional added to demonstrate the mode shift of the VSC-HVDC link in Section 6.4.
6.3.1 Jacobian transfer matrix

In Chapter 4, the concept of Jacobian transfer matrix was proposed as a unied dynamic modeling technique for ac/dc systems. For the network congurations applied in Chapter 4, the ac Jacobian transfer matrices were derived directly from the state-space representation of the combined ac network and the main circuit of the VSC. Such a procedure is practical if only one input device is connected to a simple ac network. However, for a larger ac network connected to several input devices, it is necessary to develop the statespace representation of the ac network and the electrical part of each individual input device separately as several modules, while the ac Jacobian transfer matrix is derived by connecting all the modules together. Such an approach not only eases the task for model development, but more importantly, the developed model is easier to be modied if the topology of the ac network and/or the input devices are changed. In the following subsections, such an approach is applied for the model development of the island system in Fig. 6.3. All the three input devices, i.e., the SG, the IM, and the VSC-HVDC converter are modeled as current sources which inject currents to the ac network, while the ac network supplies the terminal voltage to each input device [100, 101]. Modeling of the ac network The ac network is dened as the electrical circuit inside the dashed box in Fig. 6.3. In Chapter 4, a constant-frequency stiff voltage source E was assumed, while the d component of the synchronous dq frame was chosen aligned with the real part of E. Such a constant-frequency stiff voltage source cannot be assumed in an island system. However, for model development this is not really a problem. In fact, any electrical node in the island system can principally be chosen as a reference point, as long as the reference frame is not disturbed by the dynamics of the system. For this particular application, the PCC of the ac-network is chosen as the reference point, a common synchronous R I reference frame is dened which has the R axis aligned with the real axis of the voltage vector upcc . In the dened synchronous R I reference frame, the dynamic equations of the ac 136

network can be expressed as diL = upcc j1 Ll iL dt dupcc upcc Cl = ig + im + iv iL j1 Cl upcc dt Rl dig eg = upcc + j1 Ltg ig + Ltg dt dim em = upcc + j1 Ltm im + Ltm dt div ev = upcc + j1 Ltv iv + Ltv . dt Ll
(6.6)
By linearization and writing in component form, (6.6) can be expressed in state-space form dxn = An xn + Bn un dt yn = Cn xn + Dn un + Dn1 where the matrices An and Bn are 0 1
1 Ll
dun dt
(6.7)
1 0 0 1 Ll An = 1 0 Cl1Rl 1 C l 1 0 C 1 Cl1Rl l and the matrices Cn and Dn1 are
0 , Bn = 1 Cl 0
0 0 0
1 Cl
0 0
1 Cl
0 0 0
1 Cl
0 0
1 Cl
1 Cl
0 0
(6.8)
0 0 0 0 0 0 Cn = , Dn 1 1 0 1 0 1 0 0 1 0 1 0 1
0 0 0 0 0 0
Ltg 0 0 0 0 0
0 Ltg 0 0 0 0
0 0 Ltm 0 0 0
0 0 0 Ltm 0 0
0 0 0 0 Ltv 0
Ltv
0 0 0 0
(6.9) 137

Ra
rg q
Ll L fd
+
esd
isd Lad
i1d
L1d
i fd
R fd R1d
+
e fd
Ra
rg d isq
Ll
+ esq -
i1q
Laq
L1q
i2 q
L2 q
R1q
R2 q
Fig. 6.4 Equivalent circuit of the synchronous generator.
and the matrix Dn is
and the inputs, outputs, and state variables are un = yn = xn = igR igI imR imI ivR ivI egR egI emR emI evR evI iLR iLI upccR upccI
T T T
1 Ltg 0 Dn = 0 0 0
1 Ltg 0 0 0 0 0
0 0 0 1 Ltm 0 0
0 0 1 Ltm 0 0 0
0 0 0 0 0 1 Ltv
0 0 0 0 1 Ltv 0
(6.10)
(6.11)
The subscript R represents the real component, while I represents the imaginary component in the common ac-network R I frame. Modeling of the electrical part of the synchronous generator Fig. 6.4 shows the equivalent circuit of a salient pole synchronous machine based on the two-axis theory where the three-phase windings of the synchronous machine are repre138

sented by two orthogonal windings along the d and q axes [32]. In a dq frame chosen along the rotor DQ axis, the dynamic equations of the stator and the rotor can be expressed as esd = esq = ef d = 0= 0= 0= dd g q r Ra isd dt dq g + d r Ra isq dt df d + Rf d if d dt d1d + R1d i1d dt d1q + R1q i1q dt d2q + R2q i2q dt
(6.12)
where esd and esq are the stator voltage components in the d and q directions respectively. The quantity ef d is the eld voltage. The quantities isd and isq are the stator current components in the d and q directions respectively. The quantity if d is the eld current. The quantities i1d , i1q , and i2q are the currents of the damping circuits (two damping circuits in q axis). Ra is the stator resistance. R1d , R1q , and R2q are the resistances of the damping circuits. Rf d is the resistance of the eld circuit (d axis). d and q are the ux linkages of the stator in the d and q directions. f d is the ux linkage of the eld circuit. 1d , 1q , and g 2q are the rotor ux linkages. r is the rotor angular speed. The ux linkages in (6.12) are expressed as d = (Lad + Ll )isd + Lad if d + Lad i1d
Lsd
q = (Laq + Ll )isq + Laq i1q + Laq i2q

Lsq
f d = Lad isd + (Lad + Lf d )if d + Lad i1d

Lf md
1d = Lad isd + Lad if d + (Lad + L1d )i1d

L1md
1q = Laq isq + (Laq + L1q )i1q + Laq i2q

L1mq
2q = Laq isq + Laq i1q + (Laq + L2q )i2q

L2mq
(6.13)
where Ll is the stator leakage inductance. Lad and Laq are the d and q axis mutual inductances. Lf d is the eld leakage inductance. L1d , L1q and L2q are the leakage inductances of the damping circuits. By linearization, the dynamic equations of the synchronous gen139

erator can be expressed in state-space form dxg = L1 R xg + L1 Bu ug + dt
Ag Bg
L1 B L1 Bf
Bg1
ug1
yg =
1 0 0 0 0 0 0 1 0 0 0 0
Cg
xg
(6.14)
and the matrix R is g g g Ra r 0 0 r 0 Lsq 0 Laq r 0 Laq g g g r Ra r 0 0 0 Lsd 0 Lad r 0 Lad 0 0 Rf d 0 0 0 R= 0 0 0 R1d 0 0 0 0 0 0 R1q 0 0 0 0 0 0 R2q and the matrices Bu , B , and Bf are Bu = B = 1 0 0 0 0 0 0 1 0 0 0 0
T
where the inductance matrix L is Lsd 0 Lad Lad 0 0 0 Lsq 0 0 Laq Laq Lad 0 Lf md Lad 0 0 L= Lad 0 Lad L1md 0 0 0 Laq 0 0 L1mq Laq 0 Laq 0 0 Laq L2mq
(6.15)
(6.16)
, Bf =
0 0
Rf d g r 0 Lad T
0 0 0
isq0 Lsq (isd0 Lsd if d0 Lad ) 0 0 0 0

T T g r ef d T
(6.17)
and the inputs, outputs, and state variables are ug = yg = esd esq isd isq , ug1 = , xg =
isd isq if d i1d i1q i2q
(6.18)
where the subscript 0 denotes the value of the operating point. In (6.18), the rotor speed g r is connected to the transfer function of the rotor, while the eld voltage ef d is 140

q I
es
esI esq
rg
d esd
esR
R
Fig. 6.5 AC-network R I reference frame vs. synchronous-generator dq frame.
connected to the excitation control. The dynamic equations described by (6.12) is only valid if the reference frame is chosen along the DQ axes of the rotor. In order to connect the synchronous generator to the ac network, the state-space description in (6.14) has to be transformed to the common ac-network R I frame, as shown in Fig. 6.5. The angle represents the electrical angle between the two reference frames. Accordingly, the terminal voltage of the synchronous generator in the two reference frames are related as
RI j edq s = es e
(6.19)
in linearized component form esd esq = cos 0 sin 0 esR esI + esR0 sin 0 + esI 0 cos 0 esR0 cos 0 esI 0 sin 0
PE 1
sin 0 cos 0
PE
(6.20) Similarly, the stator current of the synchronous generator in the two reference frames are related as dq j iRI (6.21) s = is e in linearized component form isR isI = cos 0 sin 0 sin 0
PI
isd isq
cos 0
isd0 sin 0 isq0 cos 0 isd0 cos 0 isq0 sin 0

PI 1
. (6.22)
Substituting (6.20) and (6.22) into (6.14), yields dxg = Ag xg + Bg u g + Bg1 u g1 dt y g = Cg xg + Dg1 u g1
(6.23) 141

Rs Ls is Lm Lr ir Rr
+
es
+ jrm r -
Fig. 6.6 Equivalent circuit of the induction motor.
where ug = u g1 =

esR esI
, yg=
T
isR isI
g r ef d
, Bg = Bg PE , Bg1 =
Bg PE 1 Bg1 (6.24)
C g = P I C g , Dg 1 =
PI 1 zeros (2, 2) .
That is, the input and output variables have been transformed to the common ac-network R I frame. After reference-frame transformation, ug1 has one additional input variable , which is connected to the rotor transfer function. It should be noted that the state variables of (6.23) are still in the synchronous-generator dq frame. Modeling of the electrical part of the induction motor Fig. 6.6 shows the equivalent circuit of a single cage induction motor based on the twoaxis theory [32]. Different from the synchronous generator, the two axes of the induction motor are symmetrical. Rs and Rr are the resistances of the stator and rotor respectively. Ls and Lr are the inductances of the stator and rotor respectively. Lm is the mutual inductance. The vectors is and ir are the stator and rotor current vectors respectively. r is the rotor ux. The vector es is the stator voltage vector. The reference direction of the stator current vector is chosen outwards to match the current direction dened in Fig. 6.3. In a synchronous reference frame, the dynamic equations of the stator and rotor can be expressed as es = ds + js 1 Rs is dt dr m 0= + jr (1 r ) + Rr ir dt
(6.25)
m where r is the rotor angular speed. s and r are the stator and rotor uxes, which are dened as
s = Ls is + Lm ir , r = Lr ir Lm is 142
(6.26)

where Ls = Ls + Lm and Lr = Lr + Lm . The dynamic equations of the induction motor thus can be written in state-space form dxm = L1 R xm + L1 Bu um + L1 B um1 dt
Am Bm Bm1
ym =
1 0 0 0 0 1 0 0
Cm
xm
(6.27)
where the matrices L and R are Ls 0 Lm 0 0 Ls 0 Lm L= Lm 0 Lr 0 R= 0 Rs 0

m (1 r 0 )Lm
Lm
Lr
1 Ls Rs
0 1 Lm Rr
m (1 r 0 )Lr
1 Ls
m (1 r 0 )Lm
and the matrices Bu and B are Bu = B = 1 0 0 0 0 1 0 0

T
m (1 r ) L r 0 Rr 0
1 Lm
(6.28)
0 0 (Lm isq0 + Lr irq0 ) (Lm isd0 Lr ird0 )

T T m , um1 = r
(6.29)
and the inputs, outputs, and state variables are um = ym = esd esq isd isq
, xm =
isd isq ird irq
(6.30)
m The angular speed r of the induction motor is connected to the transfer function of the rotor. For the induction motor, the common ac-network R I frame can be directly chosen as the reference dq frame. Therefore, there is no need for reference-frame transformation.
Modeling of the electrical part of the VSC-HVDC converter Fig. 6.7 shows the main circuit of the VSC-HVDC converter, where Lv represents a small ctitious inductance for the convenience of model development. In a synchronous reference frame, by taking into account HHP (s), the dynamic equations of the main circuit of 143

VSC
Lv iv Rc Lc V0 v 0 ic
+
ev
+
uf Cf
+
v
Fig. 6.7 Main circuit of the VSC-HVDC converter.
the VSC-HVDC converter can be expressed as [cf. (4.45)] dic = (V0 + V )ejv HHP (s) ic uf Rc ic j1 Lc ic dt duf Cf = ic iv j1 Cf uf dt div Lv = uf ev j1 Lv iv . dt Lc
(6.31)
The Laplace transform variable s of HHP (s) can be eliminated in the same way as shown in (4.47). By linearization and writing in component form, the dynamic equation of the VSC-HVDC converter can be written in state-space form dxv = Av xv + Bv uv + Bv1 uv1 dt yv = Cv xv where the Av matrix is Av = 144 v
Rc +kv Lc
(6.32)
1 v
Rc +kv Lc
1 1 0 L c
0
1 L c
0 0 0 0
1 C f
0 0 0 0 0
1 C f
1
Rc v Lc
0 1
1 v
Rc v Lc
v 0 0 Lc
0
v Lc
1 v
1 Cf
0 0 0 0
0 0
0
1 Cf
1 0 0
1 Lv
(6.33)
0 0 0
0 0 1 0 0 0 0
1 Lv
0 0 1
0 0
1 0

and the Bv , Bv1 , and Cv matrices are Bv =
1 0 0 0 0 0 0 L v
0
1 L v
0 0 0 0 0 0
v 0 V0 sin Lc V0 cos v0 Lc V0 cos v0 Lc V0 sin v0 Lc
Bv1 = Cv =
sin v0 v V0L c v V0 cos v0 Lc
v V0 cos v0 Lc v V0 sin v0 Lc
0 0 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1
(6.34)
and the inputs, outputs, and state variables are uv = xv = evd evq
T
, uv1 =
V V0
, yv =
ivd ivq
T
icd icq cd cq uf d uf q ivd ivq
(6.35)
The input v is connected to the active-power controller, while V /V0 is connected to the alternating-voltage controller. For the VSC-HVDC converter, the common ac-network R I frame can also be directly chosen as the synchronous dq frame to avoid referenceframe transformation. Jacobian transfer matrix of the island system Before connecting to the ac network, the state-space representations of the electrical parts of the three input devices, i.e., the synchronous generator, the induction motor and the VSC-HVDC converter are lumped into one state-space model as dxz = Az xz + Bz uz + Bz 1 udz dt yz = Cz xz + Dz 1 uz 1

(6.36)
T T T T T T T T T T where xz = [xT uT ym yv ] , and uz 1 = g xm xv ] , uz = [(ug ) m uv ] , yz = [(yg ) T T T T [(u g1 ) um1 uv1 ] . Az , Bz , Bz 1 , Cz , Dz 1 are the block diagonal matrices composed of the corresponding input-device matrices in (6.23), (6.27) and (6.32), i.e., Ag Bg Bg1 , Bz = , Bz 1 = Az = A B B m m m 1 Av Bv Bv1 Cg Dg 1 , Dz 1 = . Cz = C zeros (1 , 1) m Cv zeros (2, 2)
145

Note that each input and output of the state-space model of the input devices in (6.36) and the state-space model of the ac network in (6.7) are reciprocal, i.e., each output of one system is the input of the other system. Consequently, the state-space model of the combined electrical systems can be solved as dxE = dt N (Az + Bz Dn Cz ) NBz Cn Bn Cz
AE
An
xE
N (Bz Dn Dz 1 + Bz 1 ) Bn Dz 1
uE +
BE uE
NBz Dn1 Dz 1 zeros (4, 6)
duE dt
(6.37)
T T where N = (I Bz Dn1 Cz )1 , xE = [xT z xn ] , and uE = uz 1 . In (6.37), the derivative of the input duE /dt appears on the right side of the state-space equation. However, considering the structure of Dz 1 , the differentiation is, in fact, only effective on the rotor g g angle of the synchronous generator, which yields r = d/dt. Since r belongs to uE , (6.37) can still be t in standard state-space form. Assuming per unit quantities, the output variables of the electrical system can be expressed as
Teg = Im {is conj( )} = isq d isd q Eg =

2 e2 sd + esq
Tem = Im {is conj(s )} = isq sd isd sq Uf =

2 u2 f d + uf q
P = Re {uf conj(iv )} = uf d ivd + uf q ivq (6.38)
where Teg is the electrical torque of the synchronous generator. Eg is the terminal voltage of the synchronous generator. Tem is the electrical torque of the induction motor. P is the active power from the VSC-HVDC converter. Uf is the lter-bus voltage of the VSCHVDC converter. In (6.38), Teg , Tem , P and Uf are expressed by state variables, but Eg is expressed by esd and esq , which are neither state variables nor inputs. Therefore, the terminal voltages ( of all the input devices) are solved by (6.36) and (6.7), yields uz = Nt (Dn Cz + Dn1 Cz Az ) Nt Cn xE duE dt (6.39)
+ Nt (Dn Dz 1 + Dn1 Cz Bz 1 )uE + Nt Dn1 Dz 1
where Nt = (I Dn1 Cz Bz )1 . By having all the output variables expressed by state variables and inputs, the outputs of the electrical systems can be expressed in state-space form after linearizing (6.38) yE = CE xE + DE uE . (6.40) 146

The state-space representation (6.37) and (6.40) can be further written in input-output transfer matrix form yE = CE (sI AE )1 BE + DE uE
J(s)
(6.41)
where the Jacobian transfer matrix J(s) is the linear description of the electrical part of the island system. That is, J(s) is a 5x6 transfer matrix which has uE = as the inputs and yE = as the outputs. Teg Eg Tem P Uf
T g m r ef d r v V V0 T
(6.42)
(6.43)
6.3.2 Integrated linear model

Fig. 6.8 shows the integrated linear model of the island system. The central part is the ac Jacobian transfer matrix J(s), i.e., the controlled process of the island system. The non-electrical parts, i.e., the controllers, are described in the following:
Excitation system of the synchronous generator The input signals of the exciter are the alternating-voltage reference Uref and the compensated terminal voltage Vcg . The output signal is the eld voltage ef d . KA and TA are the gain and time constant of the exciter.
Rotor dynamics of the synchronous generator Considering a single mass rotor, the dynamics of the rotor are determined by the equation of motion: g 1 dr g = (T g Teg KD r ). (6.44) dt 2H m By applying the Laplace transform, the corresponding control law can be expressed as
g r =
1 (T g Teg ). 2Hs + KD m
HRotor (s)
(6.45)
147
Induction motor
m Tm
+ -
1 2Hs + K D Rotor
U ref + -
ku s
AC-voltage control
V V0
Tem
rm
e fd
KA 1 + sTA Exciter
+ -
ref Eg
V =| u f + kc jX tv i v |
Load compensation
v c
U f P
E g AC-Jacobian transfer matrix J(s) Teg

1 s
Vcg =| e g + k g jX tg i g |
Load compensation
Pref
kp s Powersynchronization loop
Kf 1 + sT f Frequency controller
rg
1 2Hs + K D
Pref
s 1 + sTm Frequency measurement

+
+ Rotor T g m
1 R 1 1 + sTG Governor
1 1 + sTCH Turbine
Pref
VSC-HVDC Synchronous generator
Fig. 6.8 A complete linear model of the example island system.
148

Turbine and governing systems of the synchronous generator In the thesis, a steam turbine model is assumed. A simplied steam turbine model can be represented as 1 + sFHP TRH g Tm = Y (6.46) (1 + sTCH )(1 + sTRH ) with the typical values FHP = 0.3, TRH = 7.0 s, TCH = 0.3 s [32]. For a non-reheat steam turbine model [32], the transfer function is
g Tm =
1 Y. 1 + sTCH
(6.47)
The steam ow in the turbine is controlled by the governing system. The governor senses the frequency error and adjusts the steam into the turbine. The governor has the speed g deviation r and the load-reference set-point Pref as input variables. The output of the governor is the steam gate Y , which supplies the input to the steam turbine. The frequency droop R determines the power-sharing proportion of the synchronous generator. Rotor dynamics of the induction motor The rotor dynamics of the induction motor are similar to the synchronous generator. The input signals are the mechanical torque from the load and the electrical torque of the motor. Similar to the synchronous generator, the rotor dynamics of the induction motor can be expressed as 1 m r = (T m Tem ). (6.48) 2Hs + KD m It should be noted that the generator convention is applied in (6.48). In other words, if the m mechanical load is TL in motor operation, then it gives Tm = TL . Active-power and alternating-voltage control of the VSC-HVDC link The control laws are given by (5.25). Frequency and alternating-voltage droop control of the VSC-HVDC link The control laws are given by (6.2) and (6.5). The frequency measurement is obtained by differentiating angle output of the active-power controller with a measuring time constant Tm , typically in the range of 10 20 ms. It should be noted that the load compensation of the VSC-HVDC converter and the synchronous generator in Fig. 6.8 are in their nonlinear form. It requires linearization if their effects are to be considered in the linear model. The technical data of the VSC-HVDC link, the synchronous generator and the induction motor are given in Appendix C. The rest parameters of the linear model for the island system are given in Table 6.1. 149

Table 6.1 Parameters of the linear model for the island system. Loads: per unit based on 350 MVA and 400 kV. VSC-HVDC: per unit based on 350 MVA and 195 kV.
Parameters Loads Synchronous generator Rl , 1 Ll , 1 Cl KA , TA R, TG , TCH , kp , ku kv , v VSC-HVDC Kf , Tf , Tm
Values 1.75 p.u., 16.8 p.u., 0.01 p.u. 200, 0.02 s 0.067, 0.02 s, 0.3 s 100 rad/s, 60 0.3 p.u., 40 rad/s 3.72, 0.3 s, 0.02 s
6.3.3 Linear analysis

As it was discussed in Chapter 5, power-synchronization control of VSC-HVDC link is fundamentally of MIMO nature. However, with the simple diagonal controller proposed for island operation, the control parameters can be selected based on single-input singleoutput (SISO) technique by tuning one loop at a time. Accordingly, the tuning process is performed in two steps: Step 1: Starting with the integral gain of the alternating-voltage control ku = 0.0, plot the root-loci of the dominant poles by varying the integral gain kp of the activepower controller. Find the suitable parameter range of kp . Step 2: Using kp chosen at step 1, plot the root-loci of the dominant poles by varying ku to nd a suitable parameter range of ku . Fig. 6.9 shows the root-loci of the dominant poles of the closed-loop system by varying kp from 0 to 140 rad/s. Two pairs of dominant poles of the open-loop system are affected by varying kp . The critical poles are the pole pair p1,2 (p1,2 = 127 j 298), which move towards the right-half plane with increased kp . The pole pair p1,2 imposes an upper limitation on kp . Thus, kp should be chosen between 80 100 rad/s to get a trade-off between dynamic-response time and stability margin. The pole pair at p3,4 (p3,4 = 36.4 j 17.1) becomes more damped by increasing kp . Fig. 6.10 shows the root-loci of the dominant poles of the closed-loop system by varying ku with kp = 100 rad/s. The pole p0 is a pole at the origin which is introduced by the integrator. By increasing ku from 0 to 120, p0 moves towards the left-half plane. The two dominant pole pairs are also affected by variations of ku . The pole pair p3,4 moves towards the left-half plane, however, the pole pair at p1,2 slightly moves towards the right-half plane instead. Therefore, the above tuning steps should be iterative to nd a compromised combination of kp and ku to place the two dominant pole pairs. Finally, the 150
300 p 200 100 0 -100 -200 p -300 -150 -100 (rad/sec) -50 0
2 1
(rad/sec)
Fig. 6.9 Root-loci of the dominant poles of the island system. ku = 0.0, variations of kp from 0 rad/s to 140 rad/s.
300 p 200 100 0 -100 -200 -300 -150 -100 (rad/sec) -50 0
1
p3 p
0
(rad/sec)
p4 p
Fig. 6.10 Root-loci of the dominant poles of the island system. kp = 100.0 rad/s, variations of ku from 0 to 120.
151

control parameters are chosen with kp = 100 rad/s and ku = 60, which end up with the dominant pole pairs at p1,2 = 77.8 j 269 rad/s and p3,4 = 74.3 j 56 rad/s.
6.4 Simulation studies

To demonstrate the dynamic performance of the VSC-HVDC link, the island system in Fig. 6.3 is simulated with PSCAD/EMTDC. The other end of the VSC-HVDC link is assumed to be connected to a strong ac system. One of the features of power-synchronization control is its exibility with different network conditions, i.e., the proposed control structure can deal with various possible operating conditions, as well as handle the transitions between the operation modes seamlessly. In this section, this feature of power-synchronization control is demonstrated. Initially, the island system is connected to a large ac system with constant frequency. In this operation mode, the VSC-HVDC link is possible to apply constant power control. The initial condition is as following. The total loads in the island system is 320 MW whereas 120 MW from induction-motor loads (4 units) and 200 MW from RLC passive loads. The synchronous generator supplies 100 MW, while the VSC-HVDC link supplies 150 MW with constant power control. The rest of the loads are balanced by the large system, i.e., 70 MW power transmitted to the island system. Fig. 6.11 shows the simulation results when the line connection to the large ac system is lost at 0.1 sec. By detecting this event (by transient frequency measurement or the state of the ac breaker of the tripped ac line), the VSC-HVDC link enables the frequency droop control. The power that is lost from the large ac system is shared by the VSC-HVDC link and the local synchronous generator. The proportion of the power sharing is determined by the frequency droops of the synchronous generator and the VSCHVDC link where it is equally shared in this case. After disconnection from the large ac system, the frequency of the island system deviates slightly from the nominal value due to the frequency droop characteristics of the VSC-HVDC link and the synchronous generator. Fig. 6.12 is a continuation simulation from Fig. 6.11. The local synchronous generator is also tripped at 0.1 sec (new time frame). The island system then becomes a passive network with induction-motor loads and RLC passive loads. The VSC-HVDC link supplies all the loads in the island system. Due to the frequency-droop characteristics of the VSC-HVDC link, the frequency of the island system deviates much from the nominal value. At 1.5 sec, the power order of the VSC-HVDC link is adjusted to 0.9 p.u. to match the loads in the island system. Consequently, the frequency in the island system is back to the nominal value. Fig. 6.13 and Fig. 6.14 demonstrate the fault ride-through capability of the VSCHVDC link in two different operating conditions, i.e., large-ac-system connection and island operation. With both of the operating conditions, the VSC-HVDC link manages 152
6.4. Simulation studies
200 P dc (MW) 150 100 0 200 P g (MW) 150 100 50 0 1.1 f1 (p.u.) 1 0.9 0 0.5 1 1.5 2 2.5 3
0.5
1.5
2.5
0.5
1.5 time (sec)
2.5
Fig. 6.11 Trip of the ac line to the large ac system (entering island operation). Upper plot: active power from the VSC-HVDC link. Middle plot: active power from the synchronous generator. Lower plot: network frequency.
400 P dc (MW) 300 200 100 0 200 P g (MW) 100 0 0 1.1 f1 (p.u.) 1 0.9 0.8 0 0.5 1 1.5 time (sec) 2 2.5 3 0.5 1 1.5 2 2.5 3
0.5
1.5
2.5
Fig. 6.12 Trip of the local synchronous generator (entering passive network operation). At 1.5 sec, the power order of the VSC-HVDC link is adjusted to 0.9 p.u.
153
(p.u.) U
pcc
1 0 -1 0 1 0.2 0.4 0.6 0.8 1
P dc (p.u.)
0.5 0 0 2 0.2 0.4 0.6 0.8 1
| i | (p.u.)
1 0 0
0.2
0.4 time(sec)
0.6
0.8
Fig. 6.13 Fault ride-through capability of the VSC-HVDC with large-ac-system connection. Upper plot: alternating voltage at the PCC. Middle plot: active power from the VSCHVDC. Lower plot: valve current of the VSC-HVDC.
(p.u.) U
pcc
1 0 -1 0 1 0.2 0.4 0.6 0.8 1
P dc (p.u.)
0.5 0 0 2 0.2 0.4 0.6 0.8 1
| i | (p.u.)
1 0 0
0.2
0.4 time(sec)
0.6
0.8
Fig. 6.14 Fault ride-through capability of the VSC-HVDC in island operation.
154
6.5. Jacobian transfer matrix for other input devices

E g 0 g 0 SG
ig Ln Rn
E000
+
eg
+
E
Fig. 6.15 A synchronous generator connected to an impedance ac source.
recovery from the severe three-phase ac-system faults applied at the PCC. After detecting the faults, the current limiter reduces the fault currents to half of the maximum load current Imax or any other desired values. The above simulation studies demonstrate the exibility of power-synchronization control for various operating conditions. It should be noted that black start of the island system is just the reverse procedure of the above. However, the synchronous generator needs to be synchronized to the VSC-HVDC link before it can be re-connected to the island system.
6.5 Jacobian transfer matrix for other input devices

In Chapter 4, the poles and transmission zeros of the ac Jacobian transfer matrices are analyzed for grid-connected VSCs connected to several simplied ac-network congurations. It has been found that the poles and transmission zeros of the Jacobian transfer matrix are informative in understanding the basic characteristics of the ac system. The major ndings are that, the resonant (complex) poles of the Jacobian transfer matrices generally correspond to the resonances in the ac system (except the poles at the grid frequency), while the operating-point-dependent zeros have a close relationship with angle stability and voltage stability phenomena in ac systems. By moving to the origin, the zeros impose a fundamental limitation on power transmission in the ac network. In addition, the RHP zero of the Jacobian transfer matrix imposes a fundamental limitation on the achievable bandwidth of the VSC. In this section, the characteristics of the Jacobian transfer matrix are further analyzed for two additional conventional input devices in power systems, i.e., the synchronous generator and the induction motor. The goal of the analysis is to gain more insights of the Jacobian transfer matrix modeling concept.
6.5.1 Synchronous generator

Fig. 6.15 shows a synchronous generator connected to an impedance ac source, which is the classical one-machine innite-bus (OMIB) system. Rn and Ln are the resistance and inductance of the ac system. The bold letter symbols E and eg represent the voltage vectors of the ac source and the terminal voltage of the generator. E0 and Eg0 are 155
Tmg +
1 2Hs + K D Rotor
rg Jacobian transfer matrix e fd J ( s)
Teg
ref Eg +
KA 1 + TA s Exciter
E g
Fig. 6.16 Closed-loop system of the synchronous generator.
Table 6.2 lists the poles and the transmission zeros of J(s) for the synchronous generator at no-load operation, and Fig. 6.17 shows the pole-zero map of J(s) by varying the load angle of the synchronous generator. 156
their corresponding voltage magnitudes. The ac source is used as the voltage reference, and the phase angle of eg in the steady-state operating point is g0 . The quantity ig is the current vector with the reference direction from the synchronous generator to the ac system. The technical data of the synchronous generator are given in Appendix C (no step-up transformer connected). The resistance and reactance of the ac system are chosen as Rn = 0.05 p.u. and 1 Ln = 1.0 p.u. (per unit based on 150 MVA and 20 kV). The reason for using such a long line with a comparably low voltage is to facilitate studies on the zeros of the Jacobian transfer matrix. By applying the Jacobian transfer matrix modeling concept, Fig. 6.16 shows the closed-loop system of the synchronous generator connected to the ac system. For simplicity, the governing system of the synchronous generator is neglected, i.e., a constant mechanical torque is assumed. As it was shown in Fig. 6.8, the synchronous generator g has three input variables to the Jacobian transfer matrix, i.e, the rotor speed r , the rotor angle , and the eld voltage ef d . This is rather inconvenient for analysis since g r and are not independent. Therefore, in Fig. 6.16, the boundary between the Jag cobian transfer matrix and the controllers are re-dened by only having r and ef d g as inputs, while the integrator between r and is formed into the Jacobian transfer matrix, i.e., the modied Jacobian transfer matrix has the expression J T e (s ) JTe Efd (s) J T e (s ) + s JTe (s) . J (s ) = (6.49) J ( s ) E J (s ) + g JEg Efd (s) Eg s
JEg (s)
Table 6.2 Locations of poles and transmission zeros of the Jacobian transfer matrix J(s) for the synchronous generator with g0 = 0 (no load operation).
Poles p1,2 p3 p4 p5 p6 p7
rad/sec 2.56 j 314.1 40.4 25.6
Zeros z1,2 z3 z4 z5 z6
rad/sec 0.002 j 314.1 243.7 24.3 0.9 INF
2.13 0.44 0
300 200 100 z 0

3
z1
(rad/sec)
z -z
4
p3-p7 -100 -200 -300 -200 p2 0 z g0=90
g0=45
g0=27
g0=17
200
400 600 (rad/sec)
800
1000
1200
Fig. 6.17 Loci of the transmission zeros of the Jacobian transfer matrix for the synchronous generator with increased load angles.
157

It is interesting to note that, similar to that for the grid-connected VSCs, the Jacobian transfer matrix for the synchronous generator also has a resonant pole pair at the grid frequency (p1,2 ). However, a zero pair z1,2 appears very close to the locations of the resonant pole pair. The zero pair z1,2 is affected only slightly by the load-angle variations, i.e., the resonant poles at the grid frequency are canceled by the resonant zeros. Although the pole-zero cancelation is basically not affected by operating-point variations, the balance can be broken by series-compensated ac lines. As it was shown in Section 4.2.3, a series capacitor splits the resonant pole pair at the grid frequency into two other resonant pole pairs, one at the subsynchronous frequency and the other at the supersynchronous frequency [cf. (4.56)]. The SSR problem of the synchronous generator is related to the pole pair at the subsynchronous frequency. The integrating process between the rotor speed and the rotor angle has the advantage to reduce the peaks of various resonances in the ac system. However, the integrator also reduces the phase margin of the rotor loop. The small-signal stability or low-frequency oscillations in power systems are largely due to the poor phase margin of the rotor loop of the synchronous generators, which is worsened by fast excitation control. It is believed by the author that the PSS solution can, in some sense, be viewed upon as a multivariable feedback-control design which aims to decouple the interaction between the rotor loop and the excitation control. Similar to the VSC, the Jacobian transfer matrix for the synchronous generator also has but one operating-point-dependent zero z6 , which moves from innity towards the origin with increased load angles. By g0 = 90 , z6 is very close to the origin. However, the RHP zero z6 usually does not create any problem for the synchronous generator, since a synchronous generator normally operates with a load angle lower than 40 due to the transient stability limitation [64]. The RHP zero z6 at the operating point g0 = 40 will not impose any bandwidth limitation on the synchronous generator that usually has a bandwidth lower than 10 rad/s. Fig. 6.18 shows the four transfer functions of the Jacobian transfer matrix, JTe (s), JTe Efd (s), JEg (s), and JEg Efd (s) overlapped with the frequency-scanning results from PSCAD/EMTDC. The synchronous generator is assumed to operate with a load angle g0 = 30 . Due to limitations of the applied frequency-scanning technique in PSCAD/EMTDC, only the results with frequencies higher than 6.28 rad/s (1 Hz) are shown. The transfer functions of JTe (s), JTe Efd (s) show the pole-zero cancelation phenomena of the Jacobian transfer matrix at the grid frequency which are also conrmed by the frequency-scanning results from PSCAD/EMTDC. From the magnitude and phase plot of JTe (s), it can be observed that the integrator between the rotor speed and the rotor angle gives a negative slope to the magnitude, which suppresses the high-frequency resonances. However, it also introduces a 90 phase shift which limits the phase margin of the rotor loop.
158

10 |JT (j)|
5
10 |JT E (j)|
10
e fd
10 -2 10
ARG JT (j) (deg.)
-5
10
-5 -2 0 2
10
10
10 ARG J E (j) (deg.) T e fd 0
10
10
0 -50 -100 -150 -2 10 10 (rad/sec)

0
-100
10
-200 -2 10
10 (rad/sec)
10
(a) JTe (s).

10 |JE (j)|
5
(b) JTe Efd (s).

10 |JE E (j)|
0
10
g fd
10 -2 10
ARG JE (j) (deg.)
-5
10
-5 -2 0 2
10
10
10 ARG J E (j) (deg.) E g fd 0
10
10
-200
-300
-100
-400 -2 10
10 (rad/sec)
10
-200 -2 10
10 (rad/sec)
10
(c) JEg (s).
(d) JEg Efd (s).
Fig. 6.18 Bode plots of the transfer functions of the Jacobian transfer matrix for the synchronous generator (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
6.5.2 Induction motor

Fig. 6.19 shows an induction motor connected to a constant-frequency stiff voltage source. The quantity im is the stator-current vector of the induction motor. The reference direction of the stator current vector is chosen outwards to be consistent with other input devices. The technical data of the induction motor are given in Appendix C (no step-up transformer connected). By applying the Jacobian transfer matrix modeling concept, Fig. 6.20 shows the closed-loop system of the induction motor connected to the ac system. The induction m motor only has the rotor speed r connected to the Jacobian transfer matrix, while the m electrical torque Te is the output. Thus, the induction motor is modeled as a single-input159

E000 IM
im
+
E
Fig. 6.19 An induction motor connected to a stiff voltage source.

m Tm +
1 2Hs + K D Rotor
rm
Jacobian Tem transfer matrix J ( s)
Fig. 6.20 Closed-loop system of the induction motor.
single-output (SISO) feedback control system, where the rotor is treated as the controller, similar to the synchronous generator. In practice, there are also rotor-speed or squaredrotor-speed dependent contribution to the mechanical torque Tm , which can be modeled as part of the controller. However, since the major subject of this section is about the Jacobian transfer matrix, a constant mechanical torque is assumed for simplicity. Accordingly, the Jacobian transfer matrix has the expression J (s ) = J T e (s ) (6.50)
i.e., the Jacobian transfer matrix is a 1x1 matrix with JTe (s) as the only element. Table 6.3 lists the poles and zeros of the Jacobian transfer matrix for the induction motor at no-load operation. Fig. 6.21 shows the pole-zero map by increasing the loading of the induction motor.
Table 6.3 Locations of poles and zeros of the Jacobian transfer matrix of the induction motor at no-load operation (slip = 0.0).
Poles p1,2 p3 p4
rad/sec 9.4 j 314.16 13.3 13.3
Zeros z1,2 z3
rad/sec 5.1 j 314.1 13.3
Similar to the synchronous generator, the resonant pole pair p1,2 at the grid frequency is also canceled by the resonant zero pair z1,2 . This implies that the induction motor might also suffer from SSR problems if a series-compensated ac line is connected in the vicinity. Of course, it is exceptional rare for induction motors to have sufciently long rotor shafts to excite such oscillation modes. 160
300 200 100 p 0 p4 -100 -200 -300 -15 slip=0

3
p1
z1
(rad/sec)
slip=0.02
slip=0.032
slip=0.042
-10
-5 (rad/sec)
Fig. 6.21 Locus of z3 with increased slips.
3 z3=0 rad/s 2.5 z3=-5.6 rad/s z3=-10.3 rad/s
2 Te (p.u.)
1.5
0.5 z =-13.3 rad/s

3
0 0 0.2 0.4 0.6 1-slip 0.8 1 1.2 1.4
Fig. 6.22 Locus of z3 on the torque-slip curve of the induction motor .
161
10 |JT (j)|
10
-2
10 -1 10 50 (j) (deg.) 0 -50 -100 -150 -1 10
-4
10
10
10
10
ARG J
10
10 (rad/sec)
10
10
Fig. 6.23 Bode plot of JTe (s) (solid: linear model, dashed: frequency-scanning result from PSCAD/EMTDC).
With slip = 0.0, two poles p3 and p4 , as well as a real LHP zero z3 have exactly the same location. By increasing the slip of the induction motor, p3 and p4 move slightly apart along the imaginary-axis direction. However, z3 moves towards the origin, and it reaches the origin with slip = 0.042. Fig. 6.22 shows the well-known torque-slip curve of the induction motor. It is interesting to note that z3 reaches the origin exactly at the pull-out slip of the induction motor! Fig. 6.23 shows the Bode plot of the transfer function JTe (s) overlapped with the frequency scanning result from PSCAD/EMTDC with frequencies higher than 6.28 rad/s. The pole-zero cancelation phenomenon of the Jacobian transfer matrix at the grid frequency can be observed. The frequency-scanning result from PSCAD/EMTDC conrms the same phenomenon.
6.6 Subsynchronous characteristics

In the previous sections, the rotor of the synchronous generator was assumed to be made up of a single mass. With such a representation of the rotor, the bandwidth of a synchronous generator is usually in the frequency range of 0.2 to 2 Hz. In reality, however, the rotor of a steam-turbine generator has a complex mechanical structure consisting of several predominant masses connected by shafts of nite stiffness. Therefore, when the generator is perturbed, torsional oscillations occur between sections of the turbine-generator 162
6.6. Subsynchronous characteristics
Rn
Ln
Rl 400 kV PCC PCC Ltg 20 kV SG 150 MVA 350 MW (a) (b) Ltv 195 kV
400 kV
Ltg 20 kV SG 150 MVA
Ltv 195 kV
VSC
VSC 350 MW
Fig. 6.24 AC-network congurations for investigation of the subsynchronous characteristics of a VSC-HVDC link. (a) Large ac-system connection. (b) Island operation.
rotor. These torsional oscillation modes, typically in the subsynchronous frequency range, could interact with the electrical system in an adverse manner. There are predominantly two types of subsynchronous torsional interactions (SSTIs) of concern: 1. Subsynchronous resonances with series-compensated transmission lines. 2. Torsional interactions with large power-electronic devices, such as HVDC systems and FACTS devices. In this section, the subsynchronous characteristics of a VSC-HVDC link located in the vicinity of a steam-turbine synchronous generator are analyzed. The two ac-network congurations to be investigated are shown in Fig. 6.24. With the ac-network conguration shown in Fig. 6.24(a), a VSC-HVDC link and a synchronous generator are connected through an ac transmission line to a constant-frequency stiff voltage source E, which represents a large ac system. With the ac-network conguration shown in Fig. 6.24(b), a VSC-HVDC link and a synchronous generator are connected in an island system. The resistive load Rl is used to adjust the load ow in the island system. 163
Tmg + Teg
Gm ( s )
rg
Ge ( s )
Fig. 6.25 Closed-loop system formed by the linearized electrical and mechanical dynamics.
The technical data of the synchronous generator and the VSC-HVDC link are given in Appendix C.
6.6.1 Frequency-scanning method

As it was mentioned in Chapter 4, there are predominantly two methods for SSTI analysis: eigenvalue analysis and frequency scanning. The eigenvalue-analysis method is a rigorous tool. However, it requires detailed representation of the mechanical system of the synchronous generator. The frequency-scanning method is based on the complex-torque theory [81]. With the closed-loop system shown in Fig. 6.25, the transfer function between g r and Teg can be expressed as Ge (j ) = De (j ) jKe (j ) (6.51)
where De and Ke are dened as the damping torque and the synchronizing torque. Without the electrical system, the damping torque of the mechanical system is always positive due to friction. Thus, if De (j ) > 0, unstable subsynchronous oscillations (SSOs) cannot occur. It is a common practice to use the electrical damping curve of the synchronous generator to evaluate the subsynchronous characteristics of an HVDC link [102, 103]. If the characteristics of the HVDC link are taken into consideration a modied electrical damping curve De (j ) is obtained. Now, if De (j ) is larger than the original electrical damping curve De (j ) without the HVDC link, i.e.,
De (j ) > De (j )
(6.52)
the negative inuence of the HVDC link on SSOs can be eliminated. As a screening tool to evaluate the subsynchronous characteristics of an HVDC link, the frequency-scanning method has the following merits: 1. There is no need to model the mechanical system of the synchronous generator. 2. It can be performed by both frequency-domain analysis and time-domain simulations. 164

Ge ( s)
U ref +
ku s
V V0
U f
Teg rg AC-Jacobian transfer matrix
AVC
Pref + kp v P
J ( s) PVSC1 DC-Jacobian PVSC2 transfer matrix G dc ( s)

2 udc1 2 udc2
PSL
( d Cdc +
2 d Cdc
2s
) +
ref 2 (udc1 )
DVC
Fig. 6.26 Jacobian transfer matrix formulation of the large ac-system connection conguration in Fig. 6.24(a).
6.6.2 Large ac-system connection

System denition Fig. 6.26 shows the Jacobian transfer matrix formulation of the large ac-system connection conguration in Fig. 6.24(a). The transfer function Ge (s) is dened as the transfer g function between r to Teg with all the feedback loops of the VSC-HVDC link closed (inside the dashed box). In principle, Ge (s) should also take into account the effects of the exciter and possibly the power-system stabilizer (PSS) of the synchronous generator. However, since the goal of the investigation is to study the subsynchronous characteris tics of the VSC-HVDC link, where the difference between De (j ) and De (j ) is of more concern, the excitation system of the synchronous generator is neglected. The VSC-HVDC converter has two possible control modes: active-power control and direct-voltage control. As it was discussed in Chapter 5, if the direct voltage is controlled by only one converter station, the two converter stations are linearly independent. Consequently, in either of the operation modes, it is enough to have the control system 165

of one converter station modeled. Also for the sake of simplicity, the dc-link circuit is assumed to be a single capacitor, i.e., Gdc (s) = 2/(sCdc). Since the dc capacitance Cdc is explicitly included into the control parameters of the direct-voltage controller, the value of Cdc has no impact on Ge (s). Table 6.4 gives the base control parameters selected for the VSC-HVDC converter. The controllers of the VSC-HVDC converter have been tuned relatively conservative to operate the VSC-HVDC link for wider ac-network conditions. The impact of the control parameters on the subsynchronous characteristics of the VSC-HVDC converter will be further discussed after presenting the initial results based on the control parameters given in Table 6.4.
Table 6.4 Control parameters of the VSC-HVDC converter with large ac-system connection. Per unit based on 350 MVA and 195 kV.
Controller Power-synchronization loop Alternating-voltage control Direct-voltage control High-pass current control
Parameters kp ku d kv , v
Values 60 rad/s 60 40 rad/s 0.2 p.u., 40 rad/s
For analyzing SSTI between a synchronous generator and a nearby HVDC link, the unit interaction factor (UIF) is commonly used [104, 105], which is dened as UIF = MWdc MVAg 1 SCg SCtot
2
(6.53)
where MWdc is the rating of the dc system, MVAg is the rating of the generator, SCtot is the short-circuit capacity at the PCC including the synchronous generator unit, and SCg is the short-circuit capacity at the PCC excluding the generator. Based on the parameters of the synchronous generator and its step-up transformer in Appendix C, the short-circuit capacity contribution from the synchronous generator at the PCC is MVAg /(Ld + Ltg ) = 545.3 MVA. By adjusting the parameters of the ac line to the large system, ac-network congurations with different UIF values are obtained. As shown by Table 6.5, the UIF value is directly related to the SCR of the ac system1, i.e., the UIF values are higher with weak-ac-system connections. The resistances of the ac lines are chosen such that the ac lines have 85 impedance angles. In the following analysis, the synchronous generator is assumed to supply Pg = 0.1 p.u. power, while the VSC-HVDC link transmits 90% of the rated power, i.e., Pdc = 0.9 p.u. for inverter and rectier operation respectively.
1
The synchronous generator is not included in the calculation of the SCR.
166

Model validation Fig. 6.27 shows the model validation by comparing the frequency-scanning results from the linear model and the time-domain simulation. The ac-network is chosen to have UIF = 0.3. The VSC-HVDC converter operates as an inverter in power control mode. The electrical damping curves are obtained by both the linear model and the time simulations in PSCAD/EMTDC. The agreement between the linear model and the time simulation is good. Results from the linear analysis In the following, the electrical damping curves of the synchronous generator are plotted using the linear model to show the subsynchronous characteristics of the VSC-HVDC converter for ac-network congurations with the four UIF values dened in Table. 6.5. The VSC-HVDC converter operates in the following four modes: 1. Rectier operation, power control mode. 2. Inverter operation, power control mode. 3. Rectier operation, direct-voltage control mode. 4. Inverter operation, direct-voltage control mode. Figs. 6.28-6.31 show the electrical damping curves of the synchronous generator affected by the VSC-HVDC converter. The gures clearly show that the VSC-HVDC converter can have a signicant impact on the electrical damping of the generator, and the impact increases with higher UIF values (weaker ac systems). In addition, the impact also depends on the operation mode of the VSC-HVDC converter. The following can be observed from the linear results: 1. In rectier operation, power control mode, the impact of the VSC-HVDC converter is positive in the entire subsynchronous frequency range. The reason could be that the VSC-HVDC converter resembles the behavior of a resistive load in this operation mode. 2. In inverter operation, power control mode, the impact of the VSC-HVDC converter is positive in the lower subsynchronous frequency range. However, it gives negative damping contributions in the higher subsynchronous frequency range for approximately UIF > 0.1. 3. With direct-voltage control, for either rectier or inverter operation, the VSCHVDC converter tends to give negative damping in the lower subsynchronous frequency range for approximately UIF > 0.1. However, the negative contribution is rarely above 10 Hz (62.8 rad/s) even with the highest UIF = 0.7. 167
Table 6.5 Parameters of the ac line for different UIF values. Per unit based on 350 MVA and 400 kV.
UIF 0.02 0.1 0.3 0.7
Rn 0.0056 p.u. 0.0147 p.u. 0.0315 p.u. 0.0665 p.u.
1 Ln 0.0655 p.u. 0.168 p.u. 0.3581 p.u. 0.7746 p.u.
SCR 15.2 5.9 2.8 1.3
5 4 3 D (j) 2 1 0 -1 0
Linear model Time simulation
5 4 3 D (j) 2 1 0 -1 0
50
100
150 200 (rad/sec)
250
300
50
100
150 200 (rad/sec)
250
300
(a) Without VSC-HVDC.
(b) With VSC-HVDC.
Fig. 6.27 Model validation with UIF = 0.3 network conguration (solid: linear models, dashed: time simulations). (a) Electrical damping curves of the generator (Pg = 0.1 p.u.) without the VSC-HVDC link in the system. (b) Electrical damping curves of the generator (Pg = 0.1 p.u.) with the VSC-HVDC link in the system (inverter operation, power control mode, Pdc = 0.9 p.u.).
168

5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300
Without HVDC With HVDC
(a) UIF = 0.02.

5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 Without HVDC With HVDC De (j) 5 4 3 2 1 0 -1 0
(b) UIF = 0.1.

100 200 (rad/sec)
300
(c) UIF = 0.3.
(d) UIF = 0.7.
Fig. 6.28 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the VSC-HVDC converter (rectier operation, power control mode, Pdc = 0.9 p.u.).
5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 Without HVDC With HVDC De (j) 5 4 3 2 1 0 -1 0 100 200 (rad/sec) 300 Without HVDC With HVDC
(a) UIF = 0.02.

(b) UIF = 0.1.

100 200 (rad/sec)
300
(c) UIF = 0.3.
(d) UIF = 0.7.
Fig. 6.29 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the VSC-HVDC converter (inverter operation, power control mode, Pdc = 0.9 p.u.).
169

5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300
(a) UIF = 0.02.

(b) UIF = 0.1.

100 200 (rad/sec)
300
(c) UIF = 0.3.
(d) UIF = 0.7.
Fig. 6.30 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the VSC-HVDC converter (rectier operation, direct-voltage control mode, Pdc = 0.9 p.u.).
5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 Without HVDC With HVDC De (j) 5 4 3 2 1 0 -1 0 100 200 (rad/sec) 300 Without HVDC With HVDC
(a) UIF = 0.02.

5 4 De (j) 3 2 1 0 -1 0 100 200 (rad/sec) 300 -2 0 Without HVDC With HVDC De (j) 4 2 0
(b) UIF = 0.1.

100 200 (rad/sec)
300
(c) UIF = 0.3.
(d) UIF = 0.7.
Fig. 6.31 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the VSC-HVDC converter (inverter operation, direct-voltage control mode, Pdc = 0.9 p.u.).
170

Impact of the control parameters of the VSC-HVDC converter Besides the operation modes, the control parameters of the VSC-HVDC converter can also have a strong impact on the electrical damping of the generator. The most critical parameters are the gain kv of the high-pass current control HHP (s) and the gain kp of the active-power control. To evaluate the effect of HHP (s) on the electrical damping, an ideal operation mode of the VSC-HVDC converter is studied. The active-power control and the alternatingvoltage control loops are open, i.e., the VSC-HVDC converter operates as a constantfrequency stiff voltage source. The results from the linear analysis in Fig. 6.32(a) show that the high-pass current control HHP (s), in fact, gives negative damping to the generator. This seems to be contradictory to the role of HHP (s), which is introduced for the purpose of damping in the rst place. To understand this phenomenon, the damping-torque concept of the synchronous generator has to be interpreted by feedback-control theory. As shown by Fig. 6.26, the electrical damping curve De is essentially the real part of the transfer function JTe of the Jacobian transfer matrix with the controllers of the VSC-HVDC converter closed. Assuming constant mechanical torque input, the stability margin of the rotor loop of the generator can be evaluated by its open-loop transfer function HGen , which is expressed as HGen (s) = Ge (s)HRotor (s) (6.54)
where HRotor (s) is the transfer function of the rotor, which is expressed in (6.45) for a single-mass rotor. The concept of damping torque (either electrical or mechanical) has a close relationship with the stability margin of the rotor loop of the synchronous generator. The higher the damping torque, the more stability margin the synchronous generator has. However, such a stability margin is manifested by two aspects: phase margin and gain margin. The reduction of either of the stability margins will result in reduced damping torque. Based on this interpretation of the damping-torque concept, it would be easier to understand the negative damping effect of HHP(s). As shown in Fig. 6.33(a), with HHP (s) the gain of HGen (s) becomes steeper in the higher frequency range (approximately 15 250 rad/s), which is useful for suppressing other resonances in the ac system. However, the reduction of gain HGen (s) also results in the reduction of the phase of HGen . For this particular case, however, the phase margin of HGen (s) is more critical for the stability of the generator. Consequently, HHP(s) yields negative damping to the synchronous generator since it reduces the phase margin of the synchronous generator in the subsynchronous frequency range. On the other hand, as shown in Fig. 6.33(b), a higher gain kp of the power-synchronization control loop of the VSC-HVDC converter makes the slope of HGen less steep in the higher frequency range (approximately 13 280 rad/s), which substantially increases the phase margin of HGen . This effect is reected as positive damping torque in Fig. 6.32(b). 171
5 4 3 D (j) 2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250
k =0.0
v v v
5 4 3 D (j) 2 1 0 -1
k =0.0
p p p
k =0.2 k =0.5
k =60 k =100
300
-2 0
50
100
150 200 (rad/sec)
250
300
(a) kp = 0.0, ku = 0.0, variations of kv .
(b) kv = 0.2 p.u., ku = 0.0, variations of kp .
Fig. 6.32 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the control parameters of the VSC-HVDC converter (inverter operation, power control mode, Pdc = 0.9 p.u.). UIF = 0.7.
10
(j)|
kv=0.5 10
-2
(j)|
kv=0.0
10
kp=0.0 kp=100
-2
Gen
Gen
10
|H
10
-4
|H
1 2
10
-4
10
10
10
10
(j) (deg.)
-140 -160 -180 -200 10

1
(j) (deg.) arg H

Gen
2
-140 -160 -180 -200 10

1
arg H
Gen
10
10
(rad/sec)
(rad/sec)
(a) kp = 0.0, ku = 0.0, variations of kv .
(b) kv = 0.2 p.u., ku = 0.0, variations of kp .
Fig. 6.33 Phase margin of the generator (Pg = 0.1 p.u.) affected by the control parameters of the VSC-HVDC converter (inverter operation, power control mode, Pdc = 0.9 p.u.). UIF = 0.7.
172

5 4 3 D (j) 2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250 300 k =0.0
u u u
5 4 3 D (j) 2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250
k =0.0
u u u
k =60 k =100
k =60 k =100
300
(a) Rectier operation, power control.
(b) Inverter operation, power control.
Fig. 6.34 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the alternatingvoltage control of the VSC-HVDC converter (power control mode, Pdc = 0.9 p.u.). Variations of ku . UIF = 0.7.
The alternating-voltage controller of the VSC-HVDC converter also affects the electrical damping of the synchronous generator. As shown in Fig. 6.34, the alternatingvoltage controller has the effect to bring the synchronous generator and the VSC-HVDC converter electrically closer for both rectier and inverter operation. Consequently, with higher gain ku , the impact of the VSC-HVDC converter on the electrical damping curve of the synchronous generator increases, either positive or negative.
6.6.3 Island operation

System denition Fig. 6.35 shows the Jacobian transfer matrix formulation of the island system shown g in Fig. 6.24(b), where Ge (s) is the transfer function between r to Teg with all the feedback-control loops of the VSC-HVDC converter closed (inside the dashed box). Table 6.6 shows the base control parameters selected for the VSC-HVDC converter. In island operation, more interactions between the synchronous generator and the VSC-HVDC converter in the subsynchronous frequency range are expected. According to the denition in (6.53), the UIF with the island-operation conguration in Fig. 6.24(b) is UIF = MWdc = 2.33 MVAg (6.55)
simply because SCg = 0, i.e., the short-circuit power of the ac system is zero if the synchronous generator is not included. 173
Ge ( s )
U ref + -
ku s
V V0
U f
Teg rg AC-Jacobian transfer matrix J ( s)
AVC
Pref + -
P
v
kp s
PSL
s 1 + sTm Pref Kf 1 + sT f Frequency controller
Frequency measurement
Fig. 6.35 Jacobian transfer matrix formulation of the island system shown in Fig. 6.24(b).
Table 6.6 Control parameters of the VSC-HVDC converter in island operation. Per unit based on 350 MVA and 195 kV.
Controller Active-power control Alternating-voltage control Frequency control High-pass current control
Parameters kp ku Kf , Tf , Tm , kv , v
Values 60 rad/s 60 3.72, 0.3 s, 0.02 s 0.2 p.u., 40 rad/s
174

Rectier operation To operate the VSC-HVDC converter as a rectier, the synchronous generator generates 85% of its rated power, i.e., Pg = 0.85 p.u. (150 MVA power base). The generated power is then transmitted by the VSC-HVDC link to the ac system connected to the other converter station, i.e., the VSC-HVDC converter operates with Pdc = 0.36 p.u. (350 MVA power base). No load is connected in the island system, i.e., Rl = 0.0. Fig. 6.36 shows the electrical damping curves of the generator. In Fig. 6.36(a), the HVDC converter is represented by a constant-frequency stiff voltage source. Similar to the large ac-system connection conguration, the VSC-HVDC converter operating in rectier mode in the island system contributes with only positive damping to the generator in the entire subsynchronous frequency range. The damping effects are, however, much higher due to the high UIF value. The frequency-scanning results from PSCAD/EMTDC (dashed curve) conrm the same effect. The two plots produced by the linear model and the time simulation match fairly well. Inverter operation To operate the VSC-HVDC converter as an inverter, the synchronous generator generates 10% of its rated power, i.e., Pg = 0.1 p.u., while the VSC-HVDC link operates with Pdc = 0.9 p.u. Resistive loads (330 MW) are connected in the island system to consume the power from the generator and the VSC-HVDC link. Fig. 6.37 shows the electrical damping curves of the generator. In Fig. 6.37(a), the HVDC converter is represented by a constant-frequency stiff voltage source. Compared to Fig. 6.36(a), the electrical damping of the synchronous generator in Fig. 6.37(a) is apparently higher, which is due to the damping effects contributed by the resistive loads connected in the island system. As shown in Fig. 6.37(b), the inuence on the damping curve of the generator by the VSC-HVDC converter operating in inverter mode is also similar to the large ac-system connection conguration but with a higher effect due to the high UIF value. The VSC-HVDC converter contributes with positive damping in the lower frequency range but with negative damping in the higher frequency range. The frequencyscanning results from PSCAD/EMTDC (dashed curve) conrm the same effect. The two plots in Fig. 6.37 produced by the linear model and the time simulation also match fairly well. The inuence of the control parameters kp and kv of the VSC-HVDC converter on the electrical damping is also similar to the large ac-system connection conguration. As shown in Fig. 6.38, the high-pass current control contributes with negative damping, while the active-power control contributes with positive damping. Therefore, from subsynchronous characteristics point of view, a high gain kp of the active-power control is favorable even if it is not necessary from a power-control point of view. The control parameters of the frequency controller of the VSC-HVDC converter 175
5 4 3 D (j) 2 1 0 -1 0
5 4 3 D (j) 2 1 0 -1 0
50
100
150 200 (rad/sec)
250
300
50
100
150 200 (rad/sec)
250
300
(a) With constant-frequency stiff voltage source.
(b) With VSC-HVDC converter.
Fig. 6.36 Electrical damping curves of the generator (Pg = 0.85 p.u.) affected by the VSCHVDC converter (rectier operation, Pdc = 0.36 p.u.).
5 4 3 D (j)
5 4 3 D (j) 2 1 0 -1
2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250 300
-2 0
50
100
150 200 (rad/sec)
250
300
(a) With constant-frequency stiff voltage source.
(b) With VSC-HVDC converter.
Fig. 6.37 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the VSC-HVDC converter [inverter operation, Pdc = 0.9 p.u., Rl = 1.06 p.u. (330 MW resistive loads)].
176

5 4 3 D (j) 2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250 300 k =0.0
v v v
5 4 3 D (j) 2 1 0 -1 -2 0 50 100 150 200 (rad/sec) 250
k =0.0
p p p
k =0.2 k =0.5
k =60 k =100
300
(a) kp = 0.0 rad/s, ku = 0.0, variation of kv .
(b) kv = 0.2 p.u., ku = 0.0, variation of kp .
Fig. 6.38 Electrical damping curves of the generator (Pg = 0.1 p.u.) affected by the control parameters of the VSC-HVDC converter (inverter operation, power control mode, Pdc = 0.9 p.u.).
might also affect its subsynchronous characteristics. However, since a large time constant Tf is usually applied, such an inuence mainly appears in the low-frequency range where no risk of SSTI is expected. Therefore, no further discussion of the sensitivity of the subsynchronous characteristics of the VSC-HVDC converter to those control parameters is given in the thesis.
6.6.4 Summary of the subsynchronous characteristics

In this section, the subsynchronous characteristics of a VSC-HVDC converter using powersynchronization control are investigated by the frequency-scanning method. Two types of ac-network congurations, i.e., large ac-system connection and island operation, are investigated. The following conclusions are drawn for the large ac-system connection: 1. A VSC-HVDC converter can have considerable impact on the electrical damping curve of a synchronous generator for network congurations with UIF values greater than 0.1. The impact increases with higher UIF values. 2. The VSC-HVDC converter might have different subsynchronous characteristics depending on its operation mode. The safest mode is when the converter operates as a rectier in the power control mode. In this operation mode, the VSC contributes with only positive damping in the entire subsynchronous frequency range, similar to resistive loads. By operating as an inverter, the VSC contributes with positive damping in the lower frequency range. However, it contributes with slightly negative damping in the higher frequency range. In the direct-voltage control mode, in either rectier or inverter operation, the VSC might contribute with negative damp177

ing in the lower frequency range. However, this frequency range is rarely above 62.8 rad/s (10 Hz) even with the highest UIF value (UIF = 0.7). 3. The control parameters of the VSC-HVDC converter might also affect the subsynchronous characteristics. Surprisingly, the high-pass current control HHP (s), which is introduced for the purpose of damping, negatively affects the electrical damping of a synchronous generator. Further frequency-domain analysis shows that the reason is that HHP (s) reduces the phase margin of the synchronous generator. On the other hand, the active-power control of the VSC-HVDC converter provides positive damping. In other words, the negative-damping frequency range in the inverter operation mode can be reduced by selecting a higher gain of the active-power control. The alternating-voltage control has the effect of shortening the electrical distance between the VSC-HVDC converter and the synchronous generator. That is, with higher gain of the alternating-voltage control, higher subsynchronous impact from the VSC-HVDC converter is expected, either positive or negative. In island operation, stronger interactions between the VSC-HVDC converter and the synchronous generator in the subsynchronous frequency range are expected. The impact of operation modes and control parameters on the subsynchronous characteristics of the VSC-HVDC converter is similar to large ac-system connections, but with higher magnitudes. From an SSTI point of view, a higher gain of active-power control is more favorable if the stability margin of the VSC-HVDC link can be ensured. With the two types of investigated network congurations, the only possible risk of SSTI is when the VSC-HVDC converter operates as an inverter with network congurations having high UIF values, where the VSC-HVDC converter contributes with negative damping in the higher subsynchronous frequency range. The negative-damping frequency range, however, can be reduced by proper tuning of the control parameters. Thus, a subsynchronous damping controller (SSDC) for a VSC-HVDC converter using power-synchronization control is not considered necessary.
6.7 Summary
In this chapter, the control and modeling issues for VSC-HVDC links connected to lowinertia ac systems are investigated. For VSC-HVDC links connected to low-inertia ac systems, a frequency droop controller and an alternating-voltage droop controller are normally required to coordinate frequency and voltage control with local power/voltage controlling units. The Jacobian transfer matrix modeling concept is applied to model a typical island system which includes a synchronous generator, an induction motor, a VSC-HVDC link and some RLC loads. The linear model has been used for tuning the control parameters of the VSC-HVDC link. The simulation studies show that power-synchronization control is exible for various operating modes related to island operation, while the mode 178
6.7. Summary
shifts are handled seamlessly. The subsynchronous characteristics of a VSC-HVDC link using power-synchronization control are analyzed by the frequency-scanning method. A VSC-HVDC converter might have different subsynchronous characteristics depending on its operation mode. For some particular operation modes, a VSC-HVDC converter might contribute with negative damping. However, the negative-damping frequency range can be reduced by proper tuning of the control parameters.
179
180
Chapter 7 Conclusions and Future Work

This chapter presents the conclusions of the thesis and suggests future work.
7.1 Conclusions
PWM-based VSC-HVDC systems show many advantages compared to the thyristorbased LCC-HVDC system. One prominent feature is that the VSC-HVDC system has the potential to be connected to very weak ac systems where the LCC-HVDC system has difculties. In this thesis, the modeling and control issues for VSC-HVDC links connected to weak ac systems are investigated. In order to fully utilize the potential of the VSC-HVDC system for weak-ac-system connections, a novel control method, i.e., power-synchronization control, for grid-connected VSCs is proposed. A grid-connected VSC using power-synchronization control basically resembles the dynamic behavior of a synchronous machine. However, due to the technical requirements for a VSC-HVDC link and various limitations of VSC valves, additional control functions are required to deal with various practical issues during operation. Such control functions include: A high-pass current control function to damp various resonances in ac systems. A current limitation function to ride through ac-system faults. A bumpless-transfer scheme for switching the synchronization input of the VSC, and an anti-windup scheme for alternating-voltage control. A negative-sequence current controller to mitigate unbalance valve currents in the steady state or during unbalanced ac-system faults. By the comparison performed in the thesis, it is shown that power-synchronization control is superior to the traditional vector current control for VSC-HVDC links connected to weak ac systems. 181
Chapter 7. Conclusions and Future Work

The Jacobian transfer matrix is proposed as a unied modeling technique for dynamic modeling of ac/dc systems. With the ac Jacobian transfer matrix concept, a synchronous ac system is viewed upon as one multivariable feedback-control system where the feedback controllers and the controlled process, i.e., the Jacobian transfer matrix, are explicitly dened. Thus, the stability of a power system is uniquely dened as the stability of the closed-loop system formed by the controllers and the Jacobian transfer matrix. One interesting nding is that the transmission zeros of the ac Jacobian transfer matrix reect several power-system stability phenomena. By using a simplied ac-network conguration, the thesis shows that the transmission zeros of the ac Jacobian transfer matrix have close relationships with angle and voltage stability dened by the classical power-system theory. Two additional common power-system components, i.e., the synchronous generator and induction motor, are also modeled by the Jacobian transfer matrix concept. It is discovered that the transmission zeros of the ac Jacobian transfer matrices are useful to interpret some classical power-system concepts, such as the synchronizing torque for the synchronous generator and the pull-out slip for the induction motor, from a feedback-control point of view. The complex poles of the ac Jacobian transfer matrix represent the various resonances in the ac system. The high-pass current control of power-synchronization control and the inner current controller of vector current control are all shown to have general damping effects on the resonances regardless of the frequency of the resonance. The similar concept is extended to model dc systems constructed by multiple VSC terminals. By using a -link model, it is proven that the dc Jacobian transfer matrix is an inherently unstable process, where the dc resistance has a destabilizing effect. Accordingly, the feedback controller, i.e., the direct-voltage controller, has to be applied at least at one of the converter stations to stabilize the dc system. Similar to the complex poles of the ac Jacobian transfer matrix, the complex poles of the dc Jacobian transfer matrix also represent various resonances in the dc system. Long overhead dc-transmission line may create resonance peaks in the low frequency range where the control system of the VSC-HVDC link is active. A notch lter in the direct-voltage control loop is proposed to reduce the resonance peak. For VSC-HVDC links using power-synchronization control, the SCR of the ac system is no longer a critical limiting factor, i.e., VSC-HVDC links are feasible to be connected to ac systems with any short-circuit capacity. In addition, the VSC-HVDC link contributes with short-circuit capacity to the ac system at the PCC, however, without increasing the short-circuit current thanks to its current limiting capability during ac-system faults. The major consideration is the RHP zero of the ac Jacobian transfer matrix, which moves closer to the origin with larger load angles. The RHP zero imposes a fundamental limitation on the achievable bandwidth of the VSC, which implies that the VSC-HVDC link shall not operate with a too large load angle for maintaining a reasonable stability margin. For weak-ac-system connections, higher values of dc capacitance are also nec182
7.2. Future work

essary to reduce the direct-voltage variations. A control structure for interconnecting two very weak ac systems is proposed in the thesis. This control structure enables a power transmission of 0.86 p.u. from a system with an SCR of 1.2 to a system with an SCR of 1.0. For VSC-HVDC links connected to low-inertia ac systems, a frequency droop controller and an alternating-voltage droop controller are normally required to coordinate frequency and voltage control with local power/voltage controlling units. In the thesis, the Jacobian transfer matrix modeling concept is applied to model a typical island system which includes a synchronous generator, an induction motor, a VSC-HVDC link and some RLC loads. The linear model has been used for tuning the control parameters of the VSC-HVDC link. The simulation studies show that power-synchronization control is exible for various operation modes related to island operation, while the mode shifts are handled seamlessly. The subsynchronous characteristics of a VSC-HVDC link using power-synchronization control are analyzed by the frequency-scanning method. Two ac-network congurations, i.e., large ac-system connection and island operation, are investigated. A VSCHVDC converter might have different subsynchronous characteristics depending on its operation mode. For most of the operation modes, the VSC-HVDC converter contributes with positive damping to a synchronous generator located in the vicinity. The only possible risk of SSTI is when the VSC-HVDC converter operates as an inverter in a high-UIF system. The negative-damping frequency range, however, can be reduced by proper tuning of the control parameters. Thus, a subsynchronous damping controller for a VSC-HVDC link using power-synchronization control is not considered necessary.
7.2 Future work

The following is a list of possible future work: Power-synchronization control is a feasible control system for any grid-connected VSC. Another possible application is control of STATCOMs with energy storage. By using power-synchronization control, the STATCOM is able to operate with various challenging operating conditions. By using power-synchronization control, the SCR of the ac system is no longer a limiting factor for VSC-HVDC links connected to weak ac systems. However, the variation of SCRs, i.e., model uncertainty, is still a challenge. Besides the robust control methods proposed in the thesis, other possible solutions to deal with model uncertainties are 1. adaptive control. 2. online detection of the SCR of the ac system. The losses of the VSC valves have been neglected in the linear models in the thesis. However, it was shown by frequency-scanning results that the representation of the 183
Chapter 7. Conclusions and Future Work

valve losses in linear models has an importance if dc resonances are of concern. To properly represent the valve losses, the linear model should take into account the topology of the converter as well as the applied PWM technique, which are of nonlinear nature. The structure proposed for interconnecting two very weak ac systems has a possibility to be simplied. A possible simplication is to remove the power-synchronization control loop but to use the direct-voltage control to supply the synchronization input to the VSC instead. Such a modication has the potential to improve the response time of the VSC-HVDC link and reduce the dc-capacitance requirement. The modeling concept of the ac Jacobian transfer matrix is useful for studying the interactions of high power-electronics devices located in the vicinity, e.g., multiinfeed HVDC links, interactions between HVDC links and FACTS devices, etc. The modeling concept of the dc Jacobian transfer matrix is useful for studying the potential future dc systems. In this thesis, only a -link dc model with two VSC terminals is analyzed. Larger dc systems with more VSC terminals should be analyzed to gain a thorough understanding of the property of a dc system. In analyzing the subsynchronous characteristics of a VSC-HVDC converter, two ac-network congurations, i.e., large ac-system connection and island operation, are investigated. Another scenario to be investigated is when a series-compensated ac line is connected in the vicinity.
184
References
[1] VSC transmission, Cigr e Working Group B4.37-269, Tech. Rep., 2005. [2] G. Asplund, K. Eriksson, and H. Jiang, DC transmission based on voltage source converters, in Cigr e Colloquium on HVDC and FACTS, South Africa, 1997. [3] S. Johansson, G. Asplund, E. Jansson, and R. Rudervall, Power system stability benets with VSC DC-transmission systems, in Cigr e Conference B4-204, Paris, France, 2004. [4] L. Zhang, L. Harnefors, and P. Rey, Power system reliability and transfer capability improvement by VSC-HVDC (HVDC light), in Cigr e Regional Meeting, Tallin, Estonia, 2007. [5] IEEE guide for planning DC links terminating at AC locations having low shortcircuit capacities, IEEE Std 1204-1997, Tech. Rep., 1997. [6] B. Franken and G. Andersson, Analysis of HVDC converters connected to weak systems, IEEE Trans. Power Syst., vol. 5, no. 1, pp. 235242, February 1990. [7] A. Gavrilovic, AC/DC system strength as indicated by short circuit ratios, in AC/DC Power Transmission International Conference, London, 1991. [8] O. B. Nayak, A. M. Gole, D. G. Chapman, and J. B. Davies, Dynamic performance of static and synchronous compensators at an HVDC inverter bus in a very weak ac system, IEEE Trans. Power Syst., vol. 9, no. 3, pp. 13501358, August 1994. [9] D. Jovcic, N. Pahalawaththa, and M. Zavahir, Investigation of the use of inverter control strategy instead of synchronous condensers at inverter terminal of an HVDC system, IEEE Trans. Power Delivery, vol. 15, no. 2, pp. 704709, April 2000. [10] D. Jovcic, L. A. Lamont, and L. Xu, VSC transmission model for analytical studies, in IEEE Power Engineering Society General Meeting, Toronto, Canada, 2003. 185
References
[11] M. Durrant, H. Werner, and K. Abbott, Model of a VSC HVDC terminal attached to a weak ac system, in IEEE Conference on Control Applications, Istanbul, 2003. [12] P. Fischer, Modelling and control of a line-commutated HVDC transmission system interacting with a VSC STATCOM, Ph.D. dissertation, Royal Institute of Technology, Stockholm, Sweden, 2007. [13] L. Harnefors, M. Bongiornos, and S. Lundberg, Input-admittance calculation and shaping for controlled voltage-source converters, IEEE Trans. Ind. Electron., vol. 54, no. 6, pp. 33233334, December 2007. [14] On voltage and power stability in ac/dc systems, Cigr e Working Group 14.05222, Tech. Rep., 2003. [15] P. Kovacs, Transient Phenomena in Electrical Machines. Elsevier, 1984. [16] P. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of Electric Machinery and Drive Systems, 2nd Edition. Piscataway, NJ: IEEE Press, 2002. [17] M. P. Kazmierkowski, R. Krishnan, and F. Blaabjerg, Control in Power Electronics. Academic Press., 2002. [18] J. Holtz, The representation of ac machines dynamics by complex signal ow graphs, IEEE Trans. Ind. Electron., vol. 42, no. 3, pp. 263271, June 1995. [19] D. N. Zmood, D. G. Holmes, and G. H. Bode, Frequency-domain analysis of three phase linear current regulators, IEEE Trans. Ind. Applicat., vol. 37, no. 2, pp. 601 610, March/April 2001. [20] L. Harnefors, Modeling of three-phase dynamic systems using complex transfer functions and transfer matrices, IEEE Trans. Ind. Electron., vol. 54, no. 4, pp. 22392248, August 2007. [21] P. M. Anderson, B. L. Agrawal, and J. E. V. Ness, Subsynchronous resonance in power systems. Piscataway, NJ: IEEE Press, 1989. [22] K. R. Padiyar, Analysis of subsynchronous resonance in power systems. States of America: Kluwer Academic Publishers, 1999. [23] E. W. Kimbark, Direct Current Transmission, Vol. I. United
New York: Wiley, 1971. London:
[24] J. Arrillaga, High Voltage Direct Current Transmission, 2nd Edition. The Institution of Electrical Engineers, 1998. [25] B. M. Weedy and B. J. Cory, Electric Power Systems, 4th Edition. England: John Wiley & Sons Ltd., 1998. 186
West Sussex,
References
[26] J. Setr eus and L. Bertling, Introduction to HVDC technology for reliable electrical power systems, in The 10th International Conference on Probablistic Methods Applied to Power Systems, Rincon, Puerto Rico, 2008. [27] S. Corsi, A. Danelli, and M. Pozzi, Emergency-stability controls through HVDC links, in IEEE Power Engineering Society Summer Meeting, Chicago, USA, 2002. [28] T. Smed and G. Andersson, Utilizing HVDC to damp power oscillations, IEEE Trans. Power Delivery, vol. 8, no. 2, pp. 620627, April 1993. [29] W. Breuer, M. Lemes, and D. Retzmann, Perspectives of HVDC and FACTS for system interconnection and grid enhancement, in Cigr e Workshop: Brazil-ChinaIndia Summit Meeting on HVDC and Hybrid Systems, Rio de Janeiro, Brazil, 2006. [30] G. Andersson, P. Donalek, R. Farmer, and N. Hatziargyriou, Causes of the 2003 major grid blackouts in North America and Europe, and recommended means to improve system dynamic performance, IEEE Trans. Power Syst., vol. 20, no. 2, pp. 19221928, November 2005. [31] M. Bahrman and B. Johnson, The ABCs of HVDC transmission technologies, IEEE Power Energy Mag., vol. 5, no. 2, pp. 3244, March/April 2007. [32] P. Kundur, Power System Stability and Control. 1994. New York: McGraw-Hill, inc.,
[33] C. V. Thio, J. B. Davies, and K. L. Kent, Commutation failures in HVDC transmission systems, IEEE Trans. Power Delivery, vol. 11, no. 2, pp. 946957, April 1996. [34] L. Zhang and M. H. Bollen, Characteristics of voltage dips in power systems, IEEE Trans. Power Delivery, vol. 15, no. 2, pp. 827832, April 2000. [35] L. Zhang and L. D ofn as, A novel method to mitigate commutation failures in HVDC systems, in PowerCon 2002, Kunming, China, 2002. [36] Capacitor commutated converters (CCC) HVDC interconnections: digital modeling and benchmark circuit, Cigr e Working Group B4.34-352, Tech. Rep., 2008. [37] N. Flourentzou, V. G. Agelidis, and G. D. Demetriades, VSC-based HVDC power transmission systems: An overview, IEEE Trans. Power Electron., vol. 24, no. 3, pp. 592602, March 2009. [38] A. Lindberg, PWM and control of two and three level high power voltage source converters, Royal Institute of Technology, Stockholm, Sweden, Licentiate Thesis, 1995. 187
References
[39] S. Allebrod, R. Hamerski, and R. Marquardt, New transformerless, scalable modular multilevel converters for HVDC-transmission, in IEEE Power Electronics Specialists Conference - PESC 2008, Rhodes, Greece, 2008. [40] J. Rodr guez, J.-S. Lai, and F. Z. Peng, Multilevel inverters: a survey of topologies, controls, and applications, IEEE Trans. Ind. Electron., vol. 49, no. 5, pp. 724738, August 2002. [41] A. Lesnicar and R. Marquardt, A new modular voltage source inverter topology, in 10th European Conference on Power Electronics and Applications-EPE 2003, Toulouse, France, 2003. [42] M. Hagiwara and H. Akagi, Control and experiment of pulsewidth-modulated modular multilevel converters, IEEE Trans. Power Electron., vol. 24, no. 7, pp. 17371746, July 2009. [43] M. Glinka and R. Marquardt, A new AC/DC multilevel converter family, IEEE Trans. Power Electron., vol. 52, no. 3, pp. 662669, June 2005. [44] It is time to connect, http://www.abb.com/hvdc, accessed at February 2010, Tech. Rep. [45] L. Zhang, L. Harnefors, and H.-P. Nee, Power-synchronization control of gridconnected voltage-source converters, IEEE Trans. Power Syst., vol. 25, no. 2, pp. 809820, May 2010. [46] B. T. Ooi and X. Wang, Voltage angle lock loop control of the boosted type PWM converter for HVDC application, IEEE Trans. Power Electron., vol. 5, no. 2, pp. 229235, April 1990. [47] G. Joos, L. Mor an, and P. Ziogas, Performance analysis of a PWM inverter VAR compensator, IEEE Trans. Power Electron., vol. 6, no. 3, pp. 380391, July 1991. [48] J. Svensson, Voltage angle control of a voltage source inverter, application to a grid-connected wind turbine, in 6th European Conference on Power Electronics and Applications, Sevilla, 1995. [49] , Simulation of power angle controlled voltage source converter using a linear quadratic method in a wind energy application, in 5th Workshop on Computers in Power Electronics, IEEE, 1996. [50] B. K. Bose, Power Electronics and Variable Frequency Drives. IEEE Press, 1996. 188 Piscataway, NJ:
References
[51] L. Harnefors and H.-P. Nee, Model-based current control of ac machines using the internal model control method, IEEE Trans. Ind. Applicat., vol. 34, no. 1, pp. 133141, January/February 1998. [52] L. Harnefors, On analysis, control and estimation of variable-speed drives, Ph.D. dissertation, Royal Institute of Technology, Stockholm, Sweden, 1997. [53] H. Xie, On power-system benets, main-circuit design, and control of STATCOMS with energy storage, Ph.D. dissertation, Royal Institute of Technology, Stockholm, Sweden, 2009. [54] J. W. Choi and S. K. Sul, Fast current controller in three-phase AC/DC boost converter using d-q axis crosscoupling, IEEE Trans. Power Electron., vol. 13, no. 1, pp. 179185, January 1998. [55] M. P. Kazmierkowski and L. Malesani, Current control techniques for three-phase voltage-source PWM converters: A survey, IEEE Trans. Ind. Electron., vol. 45, no. 5, pp. 691703, October 1998. [56] T. Kawabata, T. Miyashita, and Y. Yamamoto, Dead beat control of three phase PWM inverter, IEEE Trans. Power Electron., vol. 5, no. 1, pp. 2128, January 1990. [57] L. Malesani, P. Mattavelli, and S. Buso, Robust deat-beat current control for PWM rectiers and active lters, IEEE Trans. Ind. Applicat., vol. 35, no. 3, pp. 613620, May/June 1999. [58] C. B. Jacobina, M. B. R. Corr ea, R. F. Pinheiro, E. R. C. Silva, and A. M. N. Lima, Modeling and control of unbalanced three-phase systems containing PWM converters, IEEE Trans. Ind. Applicat., vol. 37, no. 6, pp. 18071815, November/December 2001. [59] H.-S. Song and K. Nam, Dual current control scheme for PWM converter under unbalanced input voltage conditions, IEEE Trans. Ind. Electron., vol. 46, no. 5, pp. 953959, October 1999. [60] M. Bongiorno, On control of grid-connected voltage source converters: mitigation of voltage dips and subsynchronous resonances, Ph.D. dissertation, Chalmers University of Technology, Gothenburg, Sweden, 2007. [61] Y. Jiang, Active and reactive power control for transmission systems with voltage source converters, Ph.D. dissertation, Royal Institute of Technology, Stockholm, Sweden, 1997. 189
References
[62] L. Zhang, H.-P. Nee, and L. Harnefors, Analysis of stability limitations of a VSCHVDC link using power-synchronization control, Submitted to IEEE Trans. on Power Syst. [63] L. Zhang, L. Harnefors, and H.-P. Nee, Interconnection of two very weak ac systems by VSC-HVDC links using power-synchronization control, Accepted for publication in IEEE Trans. on Power Syst. [64] P. M. Anderson and A. A. Fouad, Power System Control and Stability, 2nd Edition. New York: IEEE Press, 2003. [65] T. V. Cutsem and C. Vournas, Voltage Stability of Electric Power Systems. Springer, inc., 1998. [66] C. Taylor, Power System Voltage Stability. New York: McGraw-Hill, inc., 1994. [67] I. J. T. F. on Stability Terms and Denitions, Denition and classication of power system stability, IEEE Trans. Power Syst., vol. 19, no. 2, pp. 13871401, May 2004. [68] F. P. Demello and C. Concordia, Concept of synchronous machine stability as affected by excitation control, IEEE Trans. Power App. Syst., vol. PAS-88, no. 4, pp. 189202, April 1969. [69] M. Klein, C. J. Rogers, and P. Kundur, A fundamental study of inter-area oscillations in power systems, IEEE Trans. Power Syst., vol. 6, no. 3, pp. 914921, August 1991. [70] E. V. Larsen and D. A. Swann, Applying power system stabilizers Part I: general concepts, IEEE Trans. Power App. Syst., no. 6, pp. 30173024, June 1981. [71] M. Noroozian and G. Andersson, Damping of inter-area and local modes by use of controllable components, IEEE Trans. Power Delivery, vol. 10, no. 4, pp. 2007 2012, October 1995. [72] D. J. Hill, Nonlinear dynamic load models with recovery for voltage stability studies, IEEE Trans. Power Syst., vol. 8, no. 1, pp. 166176, February 1993. [73] T. V. Cutsem, Voltage instability: phenomena, countermeasures, and analysis methods, Proceedings of the IEEE, vol. 88, no. 2, pp. 208227, February 2000. [74] B. Gao, G. K. Morison, and P. Kundur, Voltage stability evaluation using modal analysis, IEEE Trans. Power Syst., vol. 7, no. 4, pp. 15291542, November 1992. [75] M. K. Pal, Discussion of voltage stability evaluation using modal analysis, IEEE Trans. Power Syst., vol. 7, no. 4, pp. 15291542, November 1992. 190
References
[76] D. Lee, Voltage and power stability of HVDC systems, Ph.D. dissertation, Royal Institute of Technology, Stockholm, Sweden, 1997. [77] M. Bahrman, E. V. Larsen, R. J. Piwko, and H. S. Patel, Experience with HVDCturbine-generator torsional interaction at Square Butte, IEEE Trans. Power App. Syst., vol. PAS-99, pp. 966975, May/June 1980. [78] I. S. R. W. G. of the System Dynamic Performance Subcommitte Power System Engineering Committe, Terms, denitions and symbols for subsynchronous oscillations, IEEE Trans. Power App. Syst., vol. PAS-104, no. 6, pp. 13261334, June 1985. [79] I. S. R. W. G. of the System Dynamic Performance Subcommitte, Readers guide to subsynchronous resonance, IEEE Trans. Power Syst., vol. 7, no. 1, pp. 1326 1334, February 1992. [80] L. Harnefors, Analysis of subsynchronous torsional interaction with power electronic converters, IEEE Trans. Power Syst., vol. 22, no. 1, pp. 305313, February 2007. [81] I. M. Canay, A novel approach to the torsional interaction and electrical damping of the subsynchronous machine. Part I: Theory, IEEE Trans. Power App. Syst., vol. PAS-101, no. 10, pp. 36303638, October 1982. [82] S. Skogestad and I. Postlethwaite, Multivariable Feedback Control: Analysis and Design, 2nd Edition. West Sussex, England: John Wiley & Sons Ltd., 2005. [83] A. J. Laub and B. C. Moore, Calculation of transmission zeros using QZ techniques, Automatica, vol. 14, no. 6, pp. 557566, November 1978. [84] C. B. Moler and G. W. Stewart, An algorithm for generalized matrix eigenvalue problems, SIAM Journal on Numerical Analysis, vol. 10, no. 2, pp. 241256, April 1973. [85] R. Ottersten, On control of back-to-back converters and sensorless induction machine drives, Ph.D. dissertation, Chalmers University of Technology, Gothenburg, Sweden, 2003. [86] L. Zhang and H.-P. Nee, Multivariable feedback design of VSC-HVDC connected to weak ac systems, in PowerTech 2009, Bucharest, Romania, 2009. [87] J. D. Ainsworth, A. Gavrilovic, and H. L. Thanawala, Static and synchronous compensations for HVDC transmission converters connected to weak ac systems, in Cigr e General Session, France, 1980. 191
References
[88] M. Morari and E. Zariou, Robust Process Control. sey: Prentice-Hall, Inc., 1989. [89] B. W. Bequette, Process Control. 2003. Englewood Cliffs, New Jer-
Upper Saddle River, NJ: Prentice-Hall, Inc.,
[90] H. Kwakernaak, Robust control and H -optimizationtutorial paper, Automatica, vol. 29, no. 2, pp. 255273, 1993. [91] K. Zhou, J. Doyle, and K. Glover, Robust and Optimal Control. Englewood Cliffs, New Jersey: Prentice Hall, 1996. [92] J. C. Doyel, Analysis of feedback systems with structured uncertainties, IEE Proceedings, vol. 129, no. 6, pp. 242250, November 1982. [93] A. Packard and J. C. Doyel, The complex structured singular value, Automatica, vol. 29, no. 1, pp. 71109, 1993. [94] L. Zhang, L. Harnefors, and H.-P. Nee, Modeling and control of VSC-HVDC links connected to island systems, in IEEE Power and Energy Society General Meeting, Minneapolis, USA, 2010. [95] C. V. Thio and J. B. Davies, New synchronous compensators for the nelson river HVDC system-planning requirements and specications, IEEE Trans. Power Delivery, vol. 6, no. 2, pp. 922928, April 1991. [96] G. Zhang, Z. Xu, and H. Liu, Supply passive networks with VSC-HVDC, in IEEE Power Engineering Society Summer Meeting, Columbia, Canada, 2001. [97] C. K. Sao and P. W. Lehn, Intentional islanded operation of converter fed microgrids, in IEEE Power Engineering Society General Meeting, Minneapolis, Minnesota, USA, 2006. [98] C. Du, E. Agneholm, and G. Olsson, Comparison of different frequency controllers for a VSC-HVDC supplied system, IEEE Trans. Power Delivery, vol. 23, no. 4, pp. 22242232, October 2008. [99] , VSC-HVDC system for industrial plants with onsite generators, IEEE Trans. Power Delivery, vol. 24, no. 3, pp. 13591366, July 2009. [100] M. Parniani and M. R. Iravani, Computer analysis of small-signal stability of power systems including network dynamics, IEE Proc.-Gener. Transm. Distrib., vol. 142, no. 6, pp. 613617, 1995. 192
References
[101] D.-J. Kim, Y.-H. Moon, and H.-K. Nam, SSR small-signal stability analysis program of power systems and its application to IEEE benchmark systems, in PowerTech 2007, Lausanne, 2007. [102] Y. H afner, H. Duch en, K. Lind en, and M. Hyttinen, Improvement of subsynchronous torsional damping using VSC-HVDC, in PowerCon 2002, Kunming, China, 2002. [103] N. Prabhu and K. R. Padiyar, Investigation of subsynchronous resonance with VSC-based HVDC transmission systems, IEEE Trans. Power Delivery, vol. 24, no. 1, pp. 433440, January 2009. [104] HVDC system control for damping of subsynchronous oscillations, EPRI Research Project 1425-1, Tech. Rep., 1982. [105] R. J. Piwko and E. V. Larsen, HVDC system control for damping of subsynchronous oscillations, IEEE Trans. Power App. Syst., vol. PAS-101, no. 7, pp. 22032211, July 1982.
193
References
194
Appendices
195
Appendix A Fundamentals of the Phasor and the Space-Vector Theory

This appendix gives a brief review of the two different approaches for ac-circuit analysis, i.e., the phasor approach and the space-vector approach in Section A.1 and Section A.2. In the equations, phasors are denoted by capital letters with overline, while space vectors are denoted by bold letters. In order to avoid confusions with scaling, per unit values are assumed in the equations for both of the theory. In Section A.3, the implementations of and dq transformations for controller designs are described.
A.1 Fundamentals of the phasor theory

The phasor theory is the major mathematical tool for power systems. It has been used widely in power-ow programs, short-circuit calculations, and stability analysis. In this section, the fundamentals of the theory are briey reviewed . With the simplied three-phase ac system shown Fig. 4.5, the impedance Z of the ac line is dened as Z = R + jX (A.1)
where X = 1 L is dened as the reactance of the ac line, with 1 = 2f1 , where f1 is the fundamental frequency of the ac system. In the phasor theory, Ohms law still holds for three-phase ac systems if they are can be calculated as symmetrical, i.e., the current phasor I = V E = V E . I Z R + jX (A.2)
The phasor relations in (A.2) can also be conveniently visualized by the phasor diagram as shown in Fig. A.1. 197
Chapter A. Fundamentals of the Phasor and the Space-Vector Theory

Imaginary
jL I I
RI Real
Fig. A.1 Phasor diagram of the electrical circuit in Fig. 4.5.
The complex power is dened as = P + jQ = V I S V cos + jV sin E = (V cos + jV sin ) jX 2 EV V EV cos = sin + j X X
(A.3)
where the resistance R is neglected in (A.3) for simplicity. From (A.3), the active and reactive powers P and Q can be identied as EV sin X V 2 EV cos Q= . X P = (A.4) (A.5)
A.2 Fundamentals of the space-vector theory

In contrast to the phasor approach, where only fundamental frequency effects are treated, the space-vector approach deals with instantaneous values in the electrical circuit. The basic idea of the space-vector theory comes from the fact that one complex vector y can be used to represent three-phase variables ya , yb , yc , if any zero-sequence components are disregarded. As shown in Fig. A.2(a), the complex vector y is dened as ys = 2 2 4 ya + yb ej 3 + yc ej 3 3 (A.6)
where superscript s denotes the stationary reference frame. At any instant, the rotating vector ys has two corresponding variables y and y on the (i.e., real and imaginary) axes. To transform the rotating vector to a non-rotating 198
A.2. Fundamentals of the space-vector theory
b q y y y y a y d yd
1t
yq y
(a)
(b)
Fig. A.2 Reference frame transformations. (a) From three-phase variables to a rotating vector in the stationary frame. (b) From the stationary frame to the synchronous dq frame.
one, a rotating dq reference frame is introduced, as shown in Fig. A.2(b). The dq frame rotates in the same direction as ys with a speed that equals the angular frequency 1 of the ac system. Consequently, the vector y in the dq frame is related to ys in the stationary frame by ys = ej1 t y. (A.7) In the following, the simple ac system in Fig. 4.5 is analyzed by the space-vector approach. The equations for the dynamics of the ac system in Fig. 4.5 can be described by Kirchhoffs voltage law in the stationary frame as L dis = vs Es Ris dt (A.8)
where vs , Es , and is denote the space vectors of the voltage of node V, the voltage of node E, and the current of the ac line, respectively. To transform the electrical variables from the stationary frame to the dq frame, the space vectors are written as Es = Eej1 t , is = iej1 t , vs = vej1 t . Substituting (A.9) into (A.8) yields the dynamic equation in the dq frame L di = v E Ri j1 Li. dt (A.10) (A.9)
By comparing (A.10) to (A.2), it can be observed that there is a relationship between phasors and space vectors. If the same reference node is chosen, e.g., node E, the d and q components of the space vectors correspond to the real and imaginary parts of the phasors di in the steady state ( dt = 0). For the simple ac system in Fig. 4.5, ac capacitors are not included. However, it can easily be shown that the same statement holds even with ac capacitors included. 199

Assuming per unit scaling [20], the instantaneous active power P and reactive power Q are dened in the dq frame as P = Re {vi } = vd id + vq iq
(A.11) (A.12)
Q = Im {vi } = vq id vd iq .
In the steady state, the active and reactive powers dened in (A.11) and (A.12) in the dq frame equal those dened in (A.4) and (A.5) by the phasor approach.
A.3 Implementation of and dq transformations

In the control block diagram of vector current control (Fig. 3.5) and power-synchronization control (Fig. 3.10), and dq transformations are applied for controller implementations. In this section, the mathematical expressions for those blocks are described.
A.3.1 abc- transformation

Following (A.6), the vector ys in the stationary frame is expressed as ys = y + jy =
4 2 2 ya + yb ej 3 + yc ej 3 . 3
(A.13)
Since the zero-sequence quantities in the three-phase variables are neglected, the following relation is established y a + y b + y c = 0. (A.14)
Based on (A.13) and (A.14), the components can be derived from the three-phase variables y = ya 1 y = (y b y c ). 3
(A.15)
Based on (A.15) and (A.14), the transformation from the components to the threephase variables can also be derived ya = y 1 3 yb = y + y 2 2 1 3 yc = y y . 2 2
(A.16)
200
A.3. Implementation of and dq transformations
A.3.2 -dq transformation

According to (A.7), the vector y = yd + jyq in the synchronous dq frame is related to ys = y + jy in the stationary frame by y = ejt ys (A.17)
where t is the synchronization variable (from PSL or PLL) of the VSC. Eq. (A.17) can be written in component form as yd = y cos t + y sin t yq = y sin t + y cos t. (A.18)
Similarly, the vector ys = y + jy in the stationary frame is related to y = yd + jyq in the synchronous dq frame by ys = ejt y (A.19) which can also be written in component form as y = yd cos t yq sin t
y = yd sin t + yq cos t.
(A.20)
201
202
Appendix B Jacobian Transfer Matrix

In this appendix, some mathematical derivations and proofs related to Jacobian transfer matrix are given. In Section B.1, the transfer functions of the ac Jacobian transfer matrix for VSCs connected to an impedance source neglecting the ac capacitor at the lter bus are derived. In Section B.2, a mathematical proof of the instability of the dc Jacobian transfer matrix for a -link dc circuit is given.
B.1 Derivation of the transfer functions in Table 4.1

B.1.1 Transfer function JP (s)
In a synchronous reference frame, by neglecting the ac capacitor Cf , the dynamic equation of the main circuit in Fig. 4.4 can be written in the dq frame as di L = v E Ri j1 Li (B.1) dt where R = Rn + Rc , L = Ln + Lc , and i = ic = in . Eq. (B.1) can also be written in component form as did = V cos v E0 Rid + 1 Liq L dt diq L = V sin v Riq 1 Lid . (B.2) dt Let the voltage magnitude of the VSC be kept constant, i.e., V = V0 . If the operating points in (B.2) are denoted with subscript 0 and the deviations around the operating points are denoted with the prex , (B.2) can be linearized using v = v0 + v , id = id0 + id , iq = iq0 + iq where cos v and sin v can be linearized as cos (v0 + v ) cos v0 sin v0 v . sin (v0 + v ) sin v0 + cos v0 v (B.4) 203 (B.3)
Chapter B. Jacobian Transfer Matrix

Substituting (B.3) and (B.4) into (B.2), and keeping only the deviation parts yields the linearized form of (B.2) L did = V0 sin v0 v Rid + 1 Liq dt diq L = V0 cos v0 v Riq 1 Lid . dt
(B.5)
Moreover, by applying the Laplace transform to (B.5), the transfer functions of id vs. v , and iq vs. v are obtained id = V0 1 L cos v0 (sL + R) sin v0 v (sL + R)2 + (1 L)2 1 L sin v0 + (sL + R) cos v0 iq = V0 v . (sL + R)2 + (1 L)2
(B.6)
According to the denition in (A.11), the instantaneous active power P from the lter bus to the ac system is P = Re {uf i } . (B.7) Linearizing (B.7) yields the expression of the active-power deviation as
P = Re {i 0 uf + uf 0 i }
(B.8)
or in component form P = id0 iq0

T
uf d uf q
uf d0 uf q 0
id iq
(B.9)
The current vector i0 = id0 + jiq0 at the operation point can be derived by Uf 0 eju0 E0 i0 = Rn + j1 Ln which yields Uf 0 sin u0 E0 Uf 0 cos u0 , iq0 = . (B.11) 1 Ln 1 Ln The resistance Rn in (B.10) is neglected in (B.11) to simplify the expression. The voltage vector of the lter bus at the operating point, uf 0 = uf d0 + juf q0 , can be expressed as id0 = uf d0 = Uf 0 cos u0 , uf q0 = Uf 0 sin u0 . By linearization of the expression Ln 204 di = uf E Rn i j1 Ln i dt (B.13) (B.12) (B.10)
B.1. Derivation of the transfer functions in Table 4.1

and subdivision into d and q components, the voltage-deviation components uf d and uf q can be written as uf d = sLg id 1 Lg iq
uf q = sLg iq + 1 Lg id .
(B.14)
The resistance Rn is also neglected in (B.14) to simplify the expression. By substituting (B.6), (B.11), (B.12), and (B.14) into (B.9), the dynamic relation between P and v is found to be a0 s2 + a1 s + a2 v (B.15) P = (sL + R)2 + (1 L)2
J P ( s )
where L (k 3 k 1 ) 1 R a1 = (k3 k1 ) Lk2 1 a2 = 1 Lk3 Rk4 . a0 = The k parameters used in (B.16) are dened as k1 = V0 Uf 0 cos (v0 u0 ) k2 = V0 Uf 0 sin (v0 u0 ) k3 = E0 V0 cos v0 k4 = E0 V0 sin v0 .
(B.16)
(B.17)
B.1.2 Transfer function JQ (s)

According to the denition in (A.12), the instantaneous reactive power Q from the lter bus to the ac system is Q = Im {uf i } . (B.18) Linearizing (B.18) yields the expression of the reactive-power deviation. Thus,
Q = Im {i 0 uf + uf 0 i }
(B.19)
or in component form Q = id0 iq0

T
uf q uf d
uf d0 uf q 0
iq id
(B.20)
205

By substituting (B.6), (B.11), (B.12), and (B.14) into (B.20), the dynamic relation between Q and v is found to be Q = a0 s2 + a1 s + a2 v (sL + R)2 + (1 L)2
JQ ( s)
(B.21)
where a0 = L (k 4 k 2 ) 1 R a1 = (k4 k2 ) Lk1 1 a2 = 1 L (k4 2k2 ) R (k3 2k1 ) .
(B.22)
B.1.3 Transfer function JUf (s)

The voltage magnitude at the lter bus can be expressed as Uf = which can be linearized by using Uf = Uf 0 + Uf , uf d = uf d0 + uf d , uf q = uf q0 + uf q . (B.24)
2 u2 f d + uf q
(B.23)
Substituting (B.24) into (B.23) and only keeping the deviation parts yields the linearized form of (B.23) as uf d0 uf q 0 uf d + uf q . (B.25) Uf = Uf 0 Uf 0 By substituting (B.6) into (B.14), and further substituting (B.14) and (B.12) into (B.25), the dynamic relation between Uf and v is found to be a0 s2 + a1 s + a2 Uf = v (sL + R)2 + (1 L)2
J Uf ( s)
(B.26)
where a0 = LLn k2 /Uf 0
a1 = RLn k2 /Uf 0
2 a2 = 1 LLn k2 /Uf 0 RLn k1 /Uf 0 .
(B.27)
If the resistance R in (B.27) is neglected, JUf (s) can be simplied as J U f (s ) = 206 Ln V0 sin (v0 u0 ) . L (B.28)
B.1. Derivation of the transfer functions in Table 4.1
B.1.4 Transfer function JPV (s)

To derive the transfer functions related to the voltage-magnitude deviation V at the VSC, let v be kept constant, i.e., v = v0 . Eq. (B.2) can be linearized using V = V0 + V, id = id0 + id , iq = iq0 + iq . (B.29)
Substituting (B.29) into (B.2), and keeping only the deviation parts yields the linearized form of (B.2) as L did = V cos v0 Rid + 1 Liq dt diq L = V sin v0 Riq 1 Lid . dt
(B.30)
Moreover, by applying the Laplace transform to (B.30), the following transfer functions are obtained for id and iq vs. V id = 1 L sin v0 + (sL + R) cos v0 V (sL + R)2 + (1 L)2 1 L cos v0 + (sL + R) sin v0 V. iq = (sL + R)2 + (1 L)2
(B.31)
By substituting (B.11), (B.12), (B.14) and (B.31) into (B.9), the dynamic relation between P and V /V0 is found to be P = a0 s2 + a1 s + a2 V 2 2 (sL + R) + (1 L) V0
JPV (s)
(B.32)
where L (k 4 k 2 ) 1 R a1 = (k4 k2 ) + Lk1 1 a2 = 1 Lk4 + Rk3 . a0 =
(B.33)
B.1.5 Transfer function JQV (s)

By substituting (B.11), (B.12), (B.14) and (B.31) into (B.20), the dynamic relation between Q and V /V0 is found to be Q = a0 s2 + a1 s + a2 V 2 2 (sL + R) + (1 L) V0
JQV (s)
(B.34)
207

where L (k 1 k 3 ) 1 R a1 = (k1 k3 ) Lk2 1 a2 = 1 L (2k1 k3 ) + R (k4 2k2 ) . a0 =
(B.35)
B.1.6 Transfer function JUf V (s)

By substituting (B.31) into (B.14), and further substituting (B.14) and (B.12) into (B.25), the dynamic relation between Uf and V /V0 is found to be Uf = a0 s2 + a1 s + a2 V 2 2 (sL + R) + (1 L) V0
JUf V (s)
(B.36)
where a0 = LLn k1 /Uf 0 a1 = RLn k1 /Uf 0

2 a2 = 1 LLn k1 /Uf 0 RLn k2 /Uf 0 .
(B.37)
If the resistance R in (B.37) is neglected, JUf V (s) can be simplied as J U f V (s ) = Ln V0 cos (v0 u0 ) . L (B.38)
B.2 Proof of the instability of the dc Jacobian transfer matrix

In the following, a mathematical proof of the instability of the dc Jacobian transfer matrix Gdc (s) for operating points that Pdc10 = 0 and Pdc20 = 0 is given. Gdc (s) is a linear model of the dc -link in Fig. 4.25(b). The poles of Gdc (s) are the eigenvalues of the A matrix in (4.78), which has the characteristic equation det(I A) = 3 + a1 2 + a2 2 + a3 = 0 208 (B.39)
B.2. Proof of the instability of the dc Jacobian transfer matrix

with the coefcients Rdc Ldc Pdc10 Pdc20 Pdc10 Rdc Pdc20 Rdc 1 1 a2 = 2 + + + + udc10 u2 u2 u2 Ldc Cdc1 Ldc Cdc2 dc20 Cdc1 Cdc2 dc10 Ldc Cdc1 dc20 Ldc Cdc2 2 2 Pdc10 udc20 + Pdc20 udc10 + Pdc10 Pdc20 Rdc a3 = . (B.40) 2 Udc10 u2 dc20 Ldc Cdc1 Cdc2 a1 = Pdc10 2 udc10 Cdc1 + + To prove the instability of Gdc (s), it is sufcient that any one of the three coefcients in (B.40) is negative. The following is a proof of a3 < 0. (B.41) Pdc20 2 udc20 Cdc2
It is obvious that the denominator of a3 in (B.40) is positive. Thus, (B.41) is identical to

2 Pdc10 u2 dc20 + Pdc20 udc10 + Pdc10 Pdc20 Rdc < 0.
(B.42)
From the main circuit of the -link model in Fig. 4.25(b), the following equality is established Pdc10 Pdc20 = idc0 = . (B.43) udc10 udc20 For VSC-HVDC applications, it is apparent that udc10 and udc20 have the same polarity. Consequently, it follows from (B.43) that Pdc10 <0 Pdc20 or Pdc10 Pdc20 < 0. (B.45) Based on the inequality in (B.45), by dividing Pd10 Pd20 at both sides, (B.42) can be rewritten as u2 u2 dc20 + dc10 + Rdc > 0. (B.46) Pdc20 Pdc10 Substituting Pdc10 = udc10 idc0 Pdc20 = udc20 idc0 into (B.46), yields 2Rdc > 0 (B.48) which apparently holds. In other words, for other operating points when Pdc10 = 0 and Pdc20 = 0, Gdc (s) is unstable as long as the dc-transmission line is not lossless. 209 udc20 = udc10 Rdc idc0 (B.47) (B.44)
210
Appendix C Technical Data of the Test System

This appendix gives the technical data of the VSC-HVDC system, the synchronous generator, and the induction motor that are used in the thesis. A 50 Hz ac system is assumed.
C.1 The VSC-HVDC link
Table C.1 Technical data of the VSC-HVDC system. AC: per unit based on 350 MVA and 195 kV. DC: per unit based on 350 MW and 150 kV.
Rated power PdcN Rated ac voltage (Line-to-line) UacN Nominal ac system frequency f1 Phase-reactor reactance 1 Lc Phase-reactor resistance Rc Converter topology PWM switching frequency fsw Maximum valve current Imax Rated direct voltage UdcN DC capacitance Cdc Converter transformer rating Converter transformer ratio Transformer leakage inductance Ltv
350 MW 195 kV 50 Hz 0.2 p.u. 0.01 p.u. two-level 1650 Hz 1.08 p.u. 150 kV 0.015 p.u. 380 MVA 195 kV/400 kV 12% 211
Chapter C. Technical Data of the Test System
C.2 The synchronous generator
Table C.2 Technical data of the synchronous generator. Per unit based on 150 MVA and 20 kV.
Rated power MVAg Rated terminal voltage (Line-to-line) UacN Nominal ac system frequency f1 Stator resistance Ra Stator leakage inductance Ll Synchronous inductance Ld , Lq Transient inductance Ld , Lq Subtransient inductance Ld , Lq

150 MVA 20 kV 50 Hz 0.0045 p.u. 0.13 p.u. 1.79 p.u., 1.71 p.u. 0.169 p.u., 0.228 p.u. 0.135 p.u., 0.2 p.u. 4.3 sec, 0.85 sec
Transient time constant Td0 , Tq0 Subtransient time constant Td0 , Tq0 Inertia constant H Damping coefcient KD Step-up transformer rating Step-up transformer ratio Transformer leakage inductance Ltg
0.032 sec, 0.05 sec 6.175 sec 0.005 150 MVA 20 kV/400 kV 14%
The parameters of the equivalent circuit can be calculated by the technical data given in Table C.2: 1. The mutual inductances Lad and Laq are Lad = Ld Ll = 1.66 p.u. Laq = Lq Ll = 1.58 p.u.
(C.1)
2. The leakage inductance of the eld circuit Lf d is Lf d Lad (Ld Ll ) = = 0.04 p.u. Lad (Ld Ll )
(C.2)
212
C.3. The induction motor

3. The leakage inductances of the damping circuits L1d , L1q and L2q are L1d L1q L2q Lad Lf d (Ld Ll ) = = 0.0057 p.u. Lad Lf d (Lad + Lf d )(Ld Ll ) Laq (Lq Ll ) = = 0.1045 p.u. Laq (Lq Ll )

Laq L1q (Lq Ll ) = = 0.245 p.u. Laq L1q (Laq + L1q )(L q Ll ) Lad + Lf d = 0.0013 p.u. Td0 1
(C.3)
4. The resistance of the eld circuit Rf d is Rf d = (C.4)
5. The resistances of the damping circuits R1d , R1q and R2q are R1d = R1q R2q Lad Lf d 1 L1d + Lad + Lf d Td0 1 Laq + L1q = = 0.0063 p.u. Tq0 1 1 Laq L1q = L2q + Laq + L1q Tq0 1
= 0.0044 p.u.
= 0.0218 p.u.
(C.5)
C.3 The induction motor

Table C.3 Technical data of the induction motor. Per unit based on 30 MVA and 12 kV.
Rated power MVAm Rated terminal voltage (line-to-line) UacN Stator resistance Rs Stator leakage inductance Ls Rotor resistance Rr Rotor leakage inductance Lr Magnetizing inductance Lm Inertia constant J = 2H Damping coefcient KD Step-up transformer rating Step-up transformer ratio Transformer leakage inductance Ltm
30 MVA 12 kV 0.0034 p.u. 0.09 p.u. 0.007 p.u. 0.08 p.u. 1.9 p.u. 3.7267 sec 0.005 30 MVA 12 kV/400 kV 8% 213
Chapter C. Technical Data of the Test System
214
Appendix D List of Symbols and Abbreviations

Symbols A AS AT a B C Cdc Cdc1 , Cdc2 Cf Cn D De e E ef d eg em ev f1 fsw FPLL system matrix desired attenuation for the sensitivity function desired attenuation for the complementary sensitivity function coefcient input matrix output matrix capacitance of the single dc capacitor dc-link representation dc capacitances of the -link representation capacitance of the ac capacitor connected to the lter-bus of the VSC-HVDC system capacitance of a series capacitor feedthrough matrix damping torque control error signal voltage vector of a constant-frequency stiff voltage source eld voltage of the synchronous generator terminal voltage vector of the synchronous generator terminal voltage vector of the induction motor terminal voltage vector of the VSC-HVDC converter nominal frequency of the ac system PWM switching frequency PLL controller 215
Chapter D. List of Symbols and Abbreviations

Fn F H Hac Hdc HHP HLP I i j J, J J Jp Jn K, K kc Ke ku kv kg kp Kr L Lc Ldc P PacN PdcN p Q R Rdc r 216 notch lter for the direct-voltage controller low-pass lter transfer matrix inertia constant of the synchronous generator or the induction motor inertia of the ac system relative rotational inertia of a dc link high-pass lter low-pass lter identity matrix current vector 1
Jacobian transfer matrix and transfer function process model of J non-invertible part of J invertible part of J controller transfer matrix and transfer function load compensation ratio of the VSC synchronizing torque gain of alternating-voltage control of the VSC gain of high-pass current control of the VSC load compensation ratio of the synchronous generator gain of the power-synchronization loop of the VSC prelter of direct-voltage control inductance phase-reactor inductance of the VSC dc inductance of the -link representation active power ac power base value dc power base value pole reactive power resistance dc resistance of the -link representation reference signal
S s Td Tdc Tp Te Tm T u UacN UdcN uf v W1 W3 X y z ZRHP L 1 r n S T c
sensitivity function Laplace operator time delay time constant of the dc capacitor period time electrical torque mechanical torque complementary sensitivity function input vector alternating-voltage base value direct-voltage base value lter-bus voltage vector VSC voltage vector sensitivity weighting function complementary sensitivity weighting function reactance output vector zero RHP zero bandwidth load angle of the synchronous generator parameter of the anti-windup scheme Laplace transform structured singular value fundamental angular frequency rotor angular frequency resonance frequency of Fn crossover frequency of W1 crossover frequency of W3 power factor angle ux a state variable to represent HHP (s) in state-space form singular value
217

u v Subscripts
a,b,c ac c d dc f g I i l m max min n N nom p pcc PLL R ref s t q v 0
phase angle load angle of the VSC at the lter bus load angle of the VSC at the VSC bus time constant of a low-pass lter parameter of Fn
phase quantities alternating current quantity VSC quantity d component in the dq frame direct current quantity lter-bus values of the VSC synchronous generator quantity imaginary component in the R I frame integral control parameter load quantity induction motor quantity maximum value minimum value ac-system quantity base value nominal value proportional control parameter PCC quantity PLL quantity real component in the R I frame reference value stator quantity transformer quantity q component in the dq frame VSC quantity operating point value
218
innity norm
Superscripts
c g m P PLL Q ref s U
quantity in the converter dq frame synchronous generator quantity induction motor quantity control parameter related to active-power control control parameter related to PLL control parameter related to reactive-power control reference value quantity in the synchronous -frame control parameter related to alternating-voltage control
Abbreviations ac APC AVC CCC conj CSC dc det emf FACTS GTO HP HVDC IGBT Im IM IMC LCC LHP alternating current active-power control alternating-voltage control capacitor-commutated converter conjugate current-source converter direct current determinant electromotive force exible ac transmission systems gate turn off high pass high-voltage direct-current insulated-gate bipolar transistor imaginary induction motor internal model control line-commutated converter left-half plane
219

LP max min MIMO MPC M2C nom NPC OMIB PCC PI PLL PSL PSS p.u. PWM Re RHP RLC RPC SC SCR SG SISO SM SSDC SSR SSTI STATCOM TOV UIF VSC low pass maximum minimum multi-input multi-output maximum power curve modular-multilevel converter nominal neutral-point clamped one-machine innite-bus point-of-common-coupling proportional-integral phase-locked loop power-synchronization loop power-system stabilizer per unit pulse-width modulation real right-half plane resistive-inductive-capacitive reactive-power control synchronous condenser short-circuit ratio synchronous generator single-input single-output synchronous machine subsynchronous damping controller subsynchronous resonance subsynchronous torsional interaction static synchronous compensator transient over voltage unit interaction factor voltage-source converter
220

FULLTEXT01

Uploaded by

Copyright:

Available Formats

FULLTEXT01

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

FULLTEXT01

Uploaded by

Copyright:

Available Formats

Modeling and Control of VSC-HVDC Links Connected to Weak AC Systems

TRITAEE 2010:022 ISSN 1653-5146 ISRN KTH-EE10/22SE ISBN 978-91-7415-640-9

AT X. This document was prepared using L E

To Yibin, Karin, Vivianne and my parents and my sister Lixia

4.3 4.4 4.5 4.6

1.2. Project objectives and outline of the thesis

1.2 Project objectives and outline of the thesis

1.3 Scientic contributions of the thesis

1.3. Scientic contributions of the thesis

1.4 List of publications

1.4. List of publications

Chapter 2 High-Voltage Direct-Current Transmission

2.1 DC versus AC transmission

Chapter 2. High-Voltage Direct-Current Transmission

2.2. HVDC transmission using line-commutated current-source converters

Fig. 2.1 Graetz bridge for LCC-HVDC system.

2.2 HVDC transmission using line-commutated currentsource converters

Chapter 2. High-Voltage Direct-Current Transmission

-1 0 2 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

2.2. HVDC transmission using line-commutated current-source converters

Chapter 2. High-Voltage Direct-Current Transmission

Fig. 2.3 Two-level voltage-source converter.

2.3 HVDC transmission using forced-commutated voltagesource converters

2.3. HVDC transmission using forced-commutated voltage-source converters

Fig. 2.4 Three-level neutral-point-clamped voltage-source converter.

Chapter 2. High-Voltage Direct-Current Transmission

Negative arm (b)

2.3. HVDC transmission using forced-commutated voltage-source converters

Chapter 2. High-Voltage Direct-Current Transmission

Converter current limitation

Fig. 2.7 PQ diagram for a typical VSC-HVDC converter.

2.3. HVDC transmission using forced-commutated voltage-source converters

Chapter 2. High-Voltage Direct-Current Transmission

Chapter 3 Control Methods for VSC-HVDC Systems

3.1 Power-angle control

Chapter 3. Control Methods for VSC-HVDC Systems

Voltage reference control

3.1. Power-angle control

= sin 1 t cos t cos 1 t sin t

A PI controller can be used to minimize the error e, i.e., PLL =

Chapter 3. Control Methods for VSC-HVDC Systems

ref vc = (V0 + V ) cos(t + v 240 )

3.2. Vector current control

3.2 Vector current control

Chapter 3. Control Methods for VSC-HVDC Systems

Fig. 3.4 Converter dq frame of vector current control.

3.2. Vector current control

Linearizing (3.20) yields the following expressions P = ic cd0 ic cq 0 ic d0 ic q0

If the dynamics of the lter-bus voltage are neglected, it follows that

Chapter 3. Control Methods for VSC-HVDC Systems

3.2. Vector current control

RPC/AVC APC PLL

ref ref v vd Inner v ref ref vq current dq controller

ref va ref vb vcref

Fig. 3.6 PLL for vector current control.

Chapter 3. Control Methods for VSC-HVDC Systems

0.4 time (sec)

3.3 Power-synchronization control