FULLTEXT01
FULLTEXT01
FULLTEXT01
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
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
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
Chapter 1. Introduction
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.
Chapter 1. Introduction
Chapter 1. Introduction
V1 Ua Ub Uc
V3
V5
+ + +
L L L
Ia Ib Ic
udc
V4
V6
V2
udc (p.u.)
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
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
Ua Ub Uc
+ + +
L L L
Ia Ib Ic
udc
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.
Ua Ub Uc
+ + +
L L L
Ia Ib Ic
udc
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 B
Positive arm
A A B B ... B
Ua
+
A B A B ... A B
udc
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
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)
0.5
1.0
Active power
Under-voltage limitation
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
19
20
(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
Ln P, Q
Lc
Cf
+
uf
ic + v
Uf U ref + Q Qref +
P Pref + -
PLL
t
ref va ref vb vcref
RPC/AVC
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
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)
(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)
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
+
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 )
(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
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.
(3.12)
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
uf
E
VSC-d
Grid-d
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)
(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)
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)
(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
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
+
uf
ic + v
Uf U ref + -
Q Qref + P Pref + -
ref iq
uf
ref id
ic
abc
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
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
0.2 i ref
d
0.4
0.6
0.8
ref
i ref
q
ref
0.4
0.6
0.8
0.2
0.6
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.
(a) (b)
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.
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
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
k u V s AVC
VSC Lc
+
uf
ic
+
v
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.
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
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
+ +
ref vd
Uf
E
s
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
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
0.2
0.4
0.6
0.8
PSL
PLL
0.2
0.4
0.6
0.8
ref
Limited vd 0.4
ref
0.8
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
0.6
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
(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.
dq t
ref v
abc
(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
(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
ref ref , iq (p.u.) id
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
0.6
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
ref ref , iq (p.u.) id
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
0.6
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
K5
K2
HVDC system
K3
Induction motor
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.
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 )
51
Pref +
kp s PSL
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.
E000
Ln
Rn
P, Q in
U f 0 u 0 Rc
Lc
V0v 0 ic
+
E
+
Cf
+
v
uf
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
50
100 (deg.)
150
200
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)
4X 2 P 2 + Q2 0. E . 2 cos
(4.4)
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)
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
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
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)
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)
(4.14)
L 1 R 1
1 Lk3 Rk4
L 1 R 1
1 Lk4 + Rk3
L 1 R 1
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
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)
Infinity border
1 border
u0
1.5
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
V'
jL I '
U'f
Instant V j L I
Uf
I' I
Slow
E Real
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)
60
(4.27)
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)
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
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
(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
(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=
(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 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
= (V0 + V )ejv HHP (s)ic . Thus, the dynamic equation in (4.33) is modied as Lc
(4.45)
(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
p5 p3 p1
z2
p2
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.
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
-2
10
10
10
-2
10
10
10
10
10 (rad/sec)
10
10
10 (rad/sec)
10
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
PV
ARG J
10 (rad/sec)
10
ARG J
PV
10 (rad/sec)
10
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
-2
10
10
10
-2
10
10
10
10 (rad/sec)
10
10 (rad/sec)
10
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
-2
10
10
10 0
10
10
-2
10 0
10
10
10 (rad/sec)
10
10 (rad/sec)
10
Fig. 4.15 Bode plots of JUf V (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
69
VSC
Lc
Rn
in
U f 0u 0 Rc
V0 v 0
+
E
+
uf ic
+
v
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
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
Ln V0 (uf d0 LUf 0
id iq und unq
(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)
(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
z3 p1 p z2 u0=54 u0=41 z4
(rad/sec)
u0=41
u0=54
u0=32
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
|HPSL(j)|
10
10
-2
10
-4
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.
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
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
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
(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
(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)
(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 =
(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
ref vd ref vq
sin u0
cos u0
uc fd uc fq
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
(4.72)
p3
p1
z2
p2
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.
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
|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
PI
ARG J
-200
10
10 (rad/sec)
10
ARG J
PI
-100
10 (rad/sec)
10
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
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
VSC2
Pdc2 Pdc3
VSC3
Pac3
VSC1
Pac1
Pdc1
Pac4
VSC4
Pdc4
VSC5
Pac5
r + K dc ( s ) +
K ac ( s )
-
Pac = Pdc J ( s)
y
G dc ( s )
u dc
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= C= u=
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
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
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
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
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(s)
U f
Pdc2
G dc ( s )
2 udc2
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
10 |JU (j)|
1 2 3
-5
10
10
10
10
10
-5
10
10
10
10
10 (rad/sec)
10
10
10 (rad/sec)
10
Fig. 4.29 Bode plots of Jud (s) (solid: linear models, dashed: frequency-scanning results from PSCAD/EMTDC).
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
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
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
93
iref, ic (p.u.)
iref d ic
cd
cd
0.5
P (p.u.)
Uf (p.u.)
1 0.8 0
0.1
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
0.3
0.4
0.5
Fig. 5.4 Step response of id for vector current control at P = 0.5 p.u. with SCR = 1.0.
94
iref, ic (p.u.)
id ic
ref
cd
cd
0.5
P (p.u.)
Uf (p.u.)
1 0.8 0
0.1
0.3
0.4
0.5
Fig. 5.5 Step response of id for vector current control at P = 0.7 p.u. with SCR = 1.0.
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
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)
P (p.u.)
U (p.u.)
1 0.8 0
0.1
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)
P (p.u.)
U (p.u.)
1 0.8 0
0.1
0.3
0.4
0.5
Fig. 5.9 Step response of v for power-synchronization control at P = 0.5 p.u. with SCR = 1.0.
P (p.u.)
U (p.u.)
1 0.8 0
0.1
0.3
0.4
0.5
Fig. 5.10 Step response of v for power-synchronization control at P = 0.7 p.u. with SCR = 1.0.
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
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
P (p.u.)
0.5 iref
q
iref , ic (p.u.) q cq
Q (p.u.)
0.1
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)
Fig. 5.14 Multlivariable feedback control of grid-connected VSCs using power-synchronization control
101
K (s)
J (s)
~ J ( s) + ~ y -
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) +
~ J ( s)
J (s)
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
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
0.3
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.
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
0.3
Uf (p.u.)
Uf (p.u.)
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
0.3
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 )
+ +
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
||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
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
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: 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
0.3
0.3
Fig. 5.22 Active-power and alternating-voltage step responses of the H controller with the nominal model.
110
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
0.3
Uf (p.u.)
Uf (p.u.)
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: 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
0.3
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: 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
0.3
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: Pref To: P 1.5 P (p.u.) 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 0.6 1.5 1 0.5 0 -0.5 0
0.6
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
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
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)
(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
ref P 1
v + +
Fdc1 (s)
ref 2 (udc1 )
+ K r (s)
where the three adjustable parameters are 1 , 2 , and n . The ratio of 2 /1 sets the depth 118
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.
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
|Hdc(j)|
10
10
-2
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
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 1.5 1 1
/e
5 /
p d
10
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 .
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
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
(a)
1.15 (p.u.) u
dc2
(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
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
-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
-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.)
0.1
0.3
0.4
0.5
(a)
P1 (p.u.)
0.1 0.05 0 0
0.1
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
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
0.6
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
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
ref
+-
Kf 1 + sT f
Pref
kp s
+-
Island system
s 1 + sTm
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
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.
SG
ig
+
eg Ltm Ltv
IM
im
+
em Ll Rl Cl +
+
ev -
iv
iL u pcc -
Passive load
AC network
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 |
(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.
(6.6)
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)
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
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
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
(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
(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
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
(6.17)
and the inputs, outputs, and state variables are ug = yg = esd esq isd isq , ug1 = , xg =
(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
esI esq
rg
d esd
esR
R
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
. (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
+
es
+ jrm r -
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)
ym =
1 0 0 0 0 1 0 0
Cm
xm
(6.27)
Lm
Lr
1 Ls Rs
0 1 Lm Rr
m (1 r 0 )Lr
1 Ls
m (1 r 0 )Lm
m (1 r ) L r 0 Rr 0
1 Lm
(6.28)
(6.29)
and the inputs, outputs, and state variables are um = ym = esd esq isd isq
, xm =
(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
+
ev
+
uf Cf
+
v
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
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 =
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
(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
An
xE
N (Bz Dn Dz 1 + Bz 1 ) Bn Dz 1
uE +
BE uE
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
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)
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
(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)
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
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
+ Rotor T g m
1 R 1 1 + sTG Governor
1 1 + sTCH Turbine
Pref
VSC-HVDC Synchronous generator
148
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
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
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
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
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
P dc (p.u.)
| 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
P dc (p.u.)
| i | (p.u.)
1 0 0
0.2
0.4 time(sec)
0.6
0.8
154
E000
+
eg
+
E
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.
Tmg +
1 2Hs + K D Rotor
Teg
ref Eg +
KA 1 + TA s Exciter
E g
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
Zeros z1,2 z3 z4 z5 z6
2.13 0.44 0
z1
(rad/sec)
z -z
4
g0=45
g0=27
g0=17
200
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
158
10 |JT E (j)|
10
e fd
10 -2 10
ARG JT (j) (deg.)
-5
10
-5 -2 0 2
10
10
10
10
-100
10
-200 -2 10
10 (rad/sec)
10
10
g fd
10 -2 10
ARG JE (j) (deg.)
-5
10
-5 -2 0 2
10
10
10
10
-200
-300
-100
-400 -2 10
10 (rad/sec)
10
-200 -2 10
10 (rad/sec)
10
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).
+
E
1 2Hs + K D Rotor
rm
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
Zeros z1,2 z3
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
p1
z1
(rad/sec)
slip=0.02
slip=0.032
slip=0.042
-10
-5 (rad/sec)
2 Te (p.u.)
1.5
161
10 |JT (j)|
10
-2
-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.
Rn
Ln
Rl 400 kV PCC PCC Ltg 20 kV SG 150 MVA 350 MW (a) (b) Ltv 195 kV
400 kV
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.
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
U ref +
ku s
V V0
U f
AVC
Pref + kp v P
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).
Controller Power-synchronization loop Alternating-voltage control Direct-voltage control High-pass current control
Parameters kp ku d kv , v
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
166
Table 6.5 Parameters of the ac line for different UIF values. Per unit based on 350 MVA and 400 kV.
5 4 3 D (j) 2 1 0 -1 0
5 4 3 D (j) 2 1 0 -1 0
50
100
250
300
50
100
250
300
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
300
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
300
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
300
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
300
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
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
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
250
300
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.)
arg H
Gen
10
10
(rad/sec)
(rad/sec)
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
k =0.0
u u u
k =60 k =100
k =60 k =100
300
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.
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
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
174
5 4 3 D (j) 2 1 0 -1 0
5 4 3 D (j) 2 1 0 -1 0
50
100
250
300
50
100
250
300
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 0
50
100
250
300
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
k =0.0
p p p
k =0.2 k =0.5
k =60 k =100
300
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.
6.7 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
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
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
[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-
[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
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
jL I I
RI Real
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)
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
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
(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.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
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
(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)
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)
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.19)
uf q uf d
uf d0 uf q 0
iq id
(B.20)
205
(B.21)
(B.22)
(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)
a1 = RLn k2 /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)
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)
(B.33)
(B.34)
207
(B.35)
(B.36)
(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.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
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
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
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)
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)
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
214
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
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
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
220