Abstract

The increasing focus on large scale integration of new renewable energy

sources like wind power and wave power introduces the need for energy sto-
rage. Superconducting Magnetic Energy Storage (SMES) is a promising
alternative for active power compensation. Having high efficiency, very fast
response time and high power capability it is ideal for levelling fast fluctua-
tions.
This thesis investigates the feasibility of a current source converter as
a power conditioning system for SMES applications. The current source
converter is compared with the voltage source converter solution from the
project thesis. A control system is developed for the converter. The modu-
lation technique is also investigated.
The SMES is connected in shunt with an induction generator, and is
facing a stiff network. The objective of the SMES is to compensate for power
fluctuations from the induction generator due to variations in wind speed.
The converter is controlled by a PI-regulator and a current compensation
technique deduced from abc-theory. Simulations on the system are carried
out using the software PSIM. The simulations have proved that the SMES
works as both an active and reactive power compensator and smoothes power
delivery to the grid. The converter does however not seem like an optimum
solution at the moment. High harmonic distortion of the output currents is
the main reason for this. However this system might be interesting for low
power applications like wave power.

I
Preface

This report concludes my master thesis at Norwegian University of Science

and Technology (NTNU).
I would like to thank my supervisor Marta Molinas, ph.D student Alejandro
Garces for assisting me in my work. I would also like to thank my good friends
Nathalie, Kristian and Mari for helping me and making the hard days at the
office easier.

Knut Erik Nielsen

Trondheim, Norway

III
Contents

1 Introduction 1

2 Wind energy 3

3 Superconducting Magnetic Energy Storage 9

3.1 The principles of superconducting magnetic energy storage . . 9
3.2 Experimental testing of a superconducting coil . . . . . . . . . 11

4 The System Topology 17

4.1 Per unit representation of the system . . . . . . . . . . . . . . 18
4.2 The converter . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.1 Sizing of the converter . . . . . . . . . . . . . . . . . . 23
4.2.2 Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.3 Switching frequency selection . . . . . . . . . . . . . . 26

5 Control and modulation 29

5.1 Control system . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 ABC theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3 Space vector modulation . . . . . . . . . . . . . . . . . . . . . 33
5.4 Principle of operation . . . . . . . . . . . . . . . . . . . . . . . 38

6 Simulation and results 41

6.1 Simulation setup . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.2.1 Linearly changing wind speed . . . . . . . . . . . . . . 43
6.2.2 Comparison with VSC . . . . . . . . . . . . . . . . . . 53
6.3 High switching frequencies . . . . . . . . . . . . . . . . . . . . 60

7 Conclusion and further work 61

7.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.2 Further work . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

V
Bibliography 63

Appendices

A Key parameters of the model FBC A-1

B Component parameters A-3

C Different switching frequencies A-5

D Simulation model from PSIM A-7

E Space vector modulation A-11

F Data sheets A-17

Chapter 1

Introduction

The ever increasing demand for energy is one of the biggest challenges in the
world today. The combination of the threat from climate changes, increased
environmental focus and peak oil forces the introduction of new renewable
energy sources. There is no single renewable energy source which points itself
out to be the only solution [1]. The different energy sources have to be combi-
ned. The most promising new renewables for large scale use are onshore and
offshore wind energy, wave energy, solar energy and tidal energy. There are
several issues regarding large scale introduction of renewable energy sources
like these. One of the issues is the quality of supply. More specific it is not
possible to control the power output from a wind park or a wave power plant
like it is for hydropower. These energy sources will deliver or not deliver
energy independent of the demand. There can also be variations within a
relatively short time span. These variations can occur due to wind gusts,
clouds shadowing for the sun or other random events. These phenomena
present challenges which have to be solved before the new renewables can
provide the base load. A solution to this problem is the concept of energy
storage. There are several different types of energy storage. Fig. 1.1 shows
different concepts and their properties regarding output power and output
duration. These can mainly be divided into two groups. One type is the
energy storage solutions having the ability to store large amounts of energy
but have slow reacting times, in the other end are devices which have poor
energy storage capacity but react faster. Pumped hydro has large output po-
wer and it has the ability to supply power for a relatively long time. However
it has a long start-up time and is not suitable for fast fluctuations as the afore
mentioned. For these types of disturbances the SMES (Superconducting Ma-
gnetic Energy Storage) is most suitable having a very fast reaction time and
relatively large power output.
The main task of this master thesis is comparison of power conditioning

1
2 1. Introduction

Figure 1.1: Range of output power and output duration for different energy
storage concepts [2].

systems which connects a SMES system to the power grid. Both current
source and voltage source typologies will be evaluated. The purpose of the
SMES is to level out power fluctuations from a generator which supplies
an interconnected power network. The system response is studied using
simulation software. Loss characteristics are also taken into consideration.
Chapter 2

Wind energy

Wind energy is one of the most promising renewable energy sources available.
The worldwide production is increasing every year, and the installed capacity
reached 159 GW in 2009. Despite the financial crisis an increase rate of
31.7% occurred from 2008. This was the highest increase rate since 2001,
and a decline is not anticipated. The worldwide capacity is expected to
exceed 200 GW in 2010. The total electricity generated from the installed
wind power plants installed by 2009 will add up to 340 TWh. This equals 2%
of the world’s electricity production. The target in the European Union of
having a share of 20% renewables in the power system by 2020 will contribute
to further growth of wind power. In 2008 a report on the topic of increasing
the wind energy percentage to 20% by 2030 was released by U.S. Department
of Energy [3]. According to this report the installed capacity would have to
increase from 11.6 GW in 2006 to 305 GW in 2030. In 2009 the installed
capacity was 35 GW meaning that the building rate would have to increase
in order to reach the target.
A large share of wind power will result in a number of challenges in the
power system [4]. One of the two most basic ones is the challenge of how to
keep the voltage on an acceptable level for all customers at all time. The other
is how the fluctuating nature of wind power can meet the power demand in
the system. Wind energy is already integrated in some power systems with
a medium penetration level (18% - 31%) in some regions in Denmark and
Germany. However these regions are connected to strong networks which are
able to manage the power fluctuations from wind energy. High penetration of
wind power makes demands such as flexible sources in the network. Nuclear
power for instance is regarded as a very inflexible, and should be operated
on constant power to cover base loads. Other thermal power sources are
also desirable to operate on constant power because of the narrow maximum
efficiency band. Hydropower is a flexible source which is possible to operate

3
4 2. Wind energy

on a start and stop basis. However hydropower is a limited resource which

not every country is benefited by having. It is obvious that some measures
need to be taken. The various nature of wind power generation can be seen
by the following equation, which gives the power in the wind in watts:
1
P = ρAV 3 (2.1)
2
ρ is the average air density, A is the area swept by the turbine blades, and
V is the wind speed in m/s. It is not possible to extract all of the power in
the wind using a wind turbine as this would make the kinetic energy in the
wind equal to zero. The result of this would be accumulation of air masses
behind the turbine, and thereby a wall of air meeting the incoming wind.
Because of this there is a theoretical optimum of how much which is possible
to extract. This limit was discovered by Betz in 1926, and introduces a
constant (the Betz limit) in Eq. 2.1. The Betz limit is equal to 16/27 or 0.59
[5]. This means that maximum 59% of the energy in the wind can be utilized
by the wind turbine. In addition are the different losses in the mechanical
and electrical system before the kinetic energy in the wind is converted to
electrical energy.
There are two different types of turbine generator systems. These can be
divided into:
• Fixed speed systems.
• Variable speed systems.
In a fixed speed system the generator is of the squirrel cage induction ge-
nerator type. This is connected directly to the grid, and runs at nearly
constant speed. The small speed variations are in the range of 1-2 % around
rated speed. The simplicity of this system is its advantage. Among disad-
vantages is the more robust, and thereby expensive, construction than its
variable speed counterpart. Because the invariability of the turbine speed,
wind speed variations are lead into the drive train as torque variations.
In a variable speed system the mechanical speed of the wind turbine has
to be decoupled from the frequency of the electrical grid. To implement this,
a power electronic converter has to interface the generator and the grid. By
doing this, the generator can operate on different velocities than the one
which correspond to the frequency of the grid [6]. The emphasis in this
thesis is on the fixed speed system and the variable speed systems will not
be further dealt with.
It is evident from Eq. 2.1 that the power output from a wind turbine is
heavily dependent on the wind speed. This is at least true between the cut-in
speed, which is the wind speed where the turbine starts to generate power,
 5

and the rated power wind speed. The cut-in speed is normally in the range
3 − 5 m/s, and the rated wind speed is in the range 12 − 16 m/s. When the
wind speed exceeds the rated power speed the blades are normally pitched to
keep the output at the rated power level. Because of this some of the power
in the wind remains unexploited, but the output power from the turbine is
more predictable. The turbine is usually stopped when the wind speed goes
beyond 25 m/s. The pitch control can be used for compensating for fast wind
speed variations. However in many systems the dynamics of the actuators
controlling the blades are too slow to do this and are only used for limiting
the power output [7].
Fig. 2.1 shows the annual wind distribution and the annual energy
production by a turbine with a blade diameter of 60 m. It has a MAWS
(Mean Annual Wind Speed) of 7 m/s. The graph is of course not equal for
every site, but it has more or less the same shape. The essential information
to withdraw from the two graphs are that the extreme wind speeds occur very
infrequently and contribute to a small amount of the energy production. It is
also clear that the low wind speeds which occur quite often neither contribute
to a large share of the energy production. The largest contribution comes
from the wind speeds around the rated wind speed. Fig. 2.1 does not show
how effective the plant is compared to its theoretical potential, meaning how
productive it is. The capacity factor cf gives this information. It is defined
as:
actual yearly energy production
cf = (2.2)
maximum plant rating × 8760
The energy production is given in watt-hours, the plant rating in watt and
8760 equals the amount of hours in a year. The capacity factor is dependent
on the wind speed. A plant having a MAWS of 7 m/s has a capacity factor
about 30 %, but if the MAWS is 5 m/s the capacity factor drops to 12 %. cf
will of course never reach 100 %. The consequence of this is that the installed
capacity of wind power plants must be larger than for thermal power plants
which have higher capacity factors [4].
The power output from wind power plants will fluctuate. According to a
study made in 1957 the wind speed will vary with three distinct frequencies
[8]. One frequency which is around one minute, another which is around
12 hours and the third is about four days. These variations are caused by
turbulence, diurnal variations and long term weather systems. The two last
variations can be predicted with fairly good certainty. Weather forecast is
very important when it comes to good operation of power systems having
rather high penetration of wind power. The short term fluctuations are not
possible to predict in the same manner. The short term variations are stron-
gly coupled with the size of the wind park, more specifically the number of
6 2. Wind energy

Figure 2.1: The annual production by a turbine of 60 m diameter [8].

wind turbines. The aggregation of many turbines will damp the fluctuations
heavily [9]. This comes from the fact that wind gusts will very seldom hit a
large part of the turbines at the same time. Some turbines can experience
an increase in power output, while others a decrease. According to [9] the
maximum change in one second at a 100 MW wind park was ±2.5 MW du-
ring three months of measuring. The average change in one second was zero.
This result certainly shows the positive aspects of several turbines in one
 7

area. For distributed generation the case is totally different and the power
output fluctuations can be quite severe for a single turbine. This introduces
the requirement for energy storage. An energy storage unit which covers the
fast power fluctuations from a single turbine will improve the power quality
from the power plant. A SMES unit will be a suitable technology for this
purpose as it can deliver high power. The necessity for storing large amounts
of energy is not present for covering the fast fluctuations. For covering the
fluctuations having longer duration a SMES system will probably not be
suitable.
Chapter 3

Superconducting Magnetic
Energy Storage

Parts of this chapter is taken from the project thesis.

3.1 The principles of superconducting magne-

tic energy storage
Energy stored in a normal inductor will fade out rather quickly due to the
ohmic resistance in the coil when the power supply is disconnected. Obviously
this will not be an acceptable energy storage for use in a power system. The
ohmic resistance has to be removed before an inductor can work for this
purpose. This is possible by lowering the temperature of the conductors, and
by this making the conductors superconducting. A superconducting wire
is in a state where the resistance in the material is zero. In this state the
current in a coil can flow for infinite time. This can also be seen from the time
L
constant of a coil τ = where R goes to zero and τ then goes to infinity.
R
There are constraints for a superconducting wire to stay superconducting.
The conductor has to be operated below a critical temperature Tc , below a
critical current Ic and below a critical magnetic field Hc . There should also be
some safety margin between the critical values and the operating conditions
[10].
There are several types of superconducting material. They can be divided
into two groups. High Temperature Superconductors (HTS) and Low Tem-
perature Superconductors (LTS). The HTS types are cooled to 77 K using
liquid nitrogen. A LTS is generally cooled using liquid helium to 4.2 K [10].
There are advantages and drawbacks attended with both the technologies.

9
10 3. Superconducting Magnetic Energy Storage

Liquid nitrogen used for cooling HTS is much cheaper than the liquid helium
used for LTS. It is also an advantage having the higher operating tempera-
ture. One disadvantage regarding HTS is the fact that the materials being
used are brittle and hard to shape, at least into the form of a coil. The price is
also much higher than for the LTS. Because of this the LTS is generally used
for SMES. In a superconducting coil the normal conductor material is the
LTS alloy Niobium-Titanium (Nb/Ti) [11]. This becomes superconducting
when the temperature drops below 9.8 K [10].
The need for cooling to 4.2 K is a drawback comparing SMES with other
forms of energy storage. It lowers the efficiency and introduces refrigerators
and compressors to the system which lower the reliability. The supercon-
ducting coil itself has very good lifetime. In a test of a SMES system in
Bonneville, US in the 1980s a total of 106 charge and discharge cycles were
completed of a total planned 107 . The reason for not reaching the planned
number was not because of the inductor but problems with the refrigerator
[12]. The coil itself showed no sign of wear. Despite losses in the non su-
perconducting parts of the system, and energy needed for cooling, the total
efficiency of a SMES system can exceed 90%, a lot higher than for pumped
hydro (70%) and batteries (70-90%). The reason for the high efficiency is
that there is no need to convert the energy between different forms, chemical
to electrical or mechanical to electrical. SMES is together with capacitors
the only energy storage form which is a pure electrical [13].
SMES has a rather poor energy storage capacity compared to pumped
hydro or Compressed Air Energy Storage (CAES). There are conceptual
design studies for large scale SMES systems which can operate in diurnal
power compensation having the ability to store large amounts of energy [14].
This makes use of several thousand coils to store the energy. However this
is only a study case and a real system is not likely to be constructed in the
near future. While the energy density of a SMES system is low, one of the
main properties of SMES is the ability to deliver large amounts of energy in
a very short time, or said in another way, deliver high power. Combined with
very short response time, this makes SMES one of the most suitable energy
storage solutions to compensate for fast power fluctuations.
Summed up the features of SMES are the following:

• Capability of absorbing and delivering large amounts of power.

• High efficiency.
• Long lifetime.
• Short response time.
• Completely static construction, low maintenance.
• All electric energy storage
3.2. Experimental testing of a superconducting coil 11

3.2 Experimental testing of a superconducting

coil
During the project thesis in the autumn an educational trip to Tokyo Ins-
titute of Technology, Japan was performed. The purpose of the trip was
to participate in a laboratory experiment on a superconducting coil which
took place in November 2009. Due to problems and delays the experiment
did not finish before the stay was over and the results were not ready until
the delivery date of the project. Because of this the experiment results are
included in this thesis.
In a superconducting coil strong electromagnetic forces caused by the high
currents and strong magnetic fields cause big challenges when it comes to the
construction of the coil [14]. A large-scale SMES would require rigid supports
and a very strong construction. Traditional topologies such as solenoids
and toroidal field coils experience these issues strongly. The only current
component in the solenoid is in the toroidal direction, see Fig. 3.1a. Axial
forces exert compressive stress on the solenoid. Forces in the radial direction,
also called hoop forces, cause tensile stress in the toroidal direction which
strain to widen the solenoid. In a toroidal field coil (TFC), which is made up
of several solenoids the only current component is in the poloidal direction,
see Fig. 3.1b. The hoop forces which work on each of the solenoids exert
tensile stress in the poloidal direction. Compressive stress is generated in
the toroidal direction due to the compressive forces which work towards the
centre. All these forces require strong constructions in order to not tear the
coil apart.
The concept of balancing the electromagnetic forces is used in the forced
balanced topology (FBC). Using this typology it is also possible to control
the distribution of the stress. The method of doing this is by choosing an
optimum winding configuration. The number of poloidal turns which is the
optimum is calculated using the following formula [14]:

 √
 α ln 8α( α2 − 1 − α)

N= (3.1)
α2 +1
The variable in the formula is α which is the aspect ratio. This is equal
to the radius of the torus divided by the radius of the torus cross section
(α = torus radius/torus cross section radius). Fig. 3.2 shows the concept of
FBC. It indicates that the coil is composed by several windings, each having
the same number of poloidal turns. By having both poloidal and toroidal
current components, the tensile stress in a pure solenoid will be equalized
by the compressive stress in a TFC. The forces in the radial direction will
12 3. Superconducting Magnetic Energy Storage

(a) Solenoid (b) Toroidal field coils

Figure 3.1: Electromagnetic forces and mechanical stress in traditional to-

pologies [14].

Figure 3.2: The winding configuration and mechanical stress in a FBC.

also work against each other. As a result of the smaller resultant stress it is
possible to reduce the total mass of the coil structure for the same magnetic
energy stored. Theoretically it is feasible to bring down the total mass to a
quarter compared to a TFC and more than halve it compared to a solenoid.
A property like this would be of great importance for large scale SMES.
The superconducting coil in the experiment was constructed as a FBC.
The coil with its dimensions is shown in Fig. 3.3a. It is composed of three
helical windings. Each helical winding does six poloidal turns, and is compo-
sed by an inner and outer winding block. The schematic of the coil is shown
in Fig. 3.3b. Each of the inner winding blocks consists of 329 turns and each
of the outer winding blocks has 259 turns. Together this gives 10584 turns,
and an inductance of 1.8 H. Other parameters can be found in Appendix A.
3.2. Experimental testing of a superconducting coil 13

0.13m 0.27m
0.53m
(a) The bare force balanced coil. (b) The composition of the FBC.

Figure 3.3

During the experiment the coil was kept inside an isolated container. The
laboratory setup is displayed in Fig. 3.4. The two large tanks on the left
contained liquid helium, used to cool the FBC. The total amount of helium
used in the experiment was 5000 l. The smaller container to the right contai-
ned the FBC. The current cables were connected to the top of the container.
It is possible to spot frost on top of the container which was formed from the
condensed humidity in the air.
The critical limits for superconductivity of the FBC were three parame-
ters. Current, magnetic field and temperature. The current limit was 552 A.
The magnetic field limit was 7.1 T. The coil became superconducting below
9.8 K, but to obtain high values of current and magnetic field the tempe-
rature limit was 4.2 K,the temperature of liquid helium which was used as
cooling agent. The test runs starting in November 2009 was the fourth series
of test runs. The first was in February 2007, second in September 2007 and
third in March 2008. Even though the critical current of the FBC is 552 A
the coil was never exited to this value in the first three test runs. A transi-
tion from superconducting state to normal conducting state occurred before
this value was reached. This phenomena is called ”quench”. The first quench
in February 2007 happened at 293 A and 3.8 T, only 53 % of the critical
limits. However the quench currents increased from test run to test run. In
the third test run, the highest quench current was 476 A, 86 % of the cur-
14 3. Superconducting Magnetic Energy Storage

Figure 3.4: The experimental setup in the laboratory. The two left tanks
contain liquid helium. The right one contains the superconducting coil.

rent limit. The phenomena of increasing quench currents is called ”training”.

The physics behind the quench and training is not fully understood. It is
believed to have something to do with the position of each of the supercon-
ducting strands which compose the windings. When the current increases
the electromagnetic forces on the strands increase. At a certain current, the
force will become too big for the strand to keep in position, and it will move
slightly. This small movement will result in a temperature increase, and
move the FBC out of its limits, and a quench occurs. The complex winding
configuration and the fact that it is handmade will make some imperfections
in the windings. To improve the quench performance an optimized winding
technique is necessary.
The objective of the test run in November 2009 was to investigate the
effect of using super cooled liquid helium at a temperature of 2.17 K. The
method of doing this was to reduce the pressure inside the FBC Dewar below
one atmosphere using a vacuum pump. By doing this a larger margin from
operating conditions to the quench limits would be present. The circuit in
the experiment is shown in Fig. 3.5. The DC-power supply is connected to
3.2. Experimental testing of a superconducting coil 15

Figure 3.5: The circuit in the FBC experiment.

Figure 3.6: The training effect of the FBC.

the network through a thyristor rectifier. The branch containing the dump
resistance is connected using a thyristor when a quench is detected. At the
same time the circuit breaker (CB) is opened.
The result of the experiment and the training effect can be seen in Fig. 3.6
and Table 3.1. The current ramp rate was 1 A/s. The current values were
read from an ammeter on the DC-source. The magnetic field was calculated
from the current, and corresponded to the toroidal field. The quench current
increased for each test run, but the first quench current when a new test
run started was lower than the last at the previous test run. However the
quench current was not reduced to the value for the first quench. This showed
that the training effect not disappeared after the coil was warmed to room
temperature. The record quench current occurred on the 105th quench.
16 3. Superconducting Magnetic Energy Storage

Table 3.1: The quench currents and magnetic fields in the fourth test run.
Quench no. Quench current (A) Magnetic field (T) Temperature (K)
82 445 5.7 4.2
83 445 5.7 4.2
84 467 6.0 4.2
85 458 5.9 2.6
86 470 6.0 4.2
87 473 6.1 4.2
88 464 6.0 4.2
89 480 6.2 4.2
90 457 5.9 4.2
91 451 5.8 2.6
92 464 6.0 2.3
93 473 6.1 2.2
94 472 6.1 4.2
95 474 6.1 2.3
96 477 6.1 2.3
97 461 5.9 4.2
98 466 6.0 4.2
99 473 6.1 2.2
100 489 6.3 2.1
101 479 6.1 2.1
102 489 6.3 2.0
103 486 6.2 2.0
104 471 6.0 4.2
105 492 6.3 4.2
106 490 6.3 4.2
107 487 6.2 4.2

492 A corresponded to 89 % of the maximum coil current. This was an

increase compared to the lowest quench current in the fourth run of 10.6 %,
and to the all time lowest of 67.9 %. The record quench current did however
not come when the FBC was cooled using the super cooled helium, but
at the normal temperature of liquid helium. The difference between the
quench current at super cooled conditions and normal conditions were not
big, which indicated that the temperature difference of a few Kelvin were not
crucial. The reason for this could be that the margin from around 2 K to the
superconductivity limit was too small anyway when a strand moved and there
was a temperature rise. The procedure of cooling the liquid helium further
from 4.2 K was done by reducing the helium gas pressure inside the Dewar.
This was a slow process as the pressure decreased, and limited the amounts of
tests per day. Problems with the vacuum pumps did also cause problems in
the experiment which decreased the amounts of test runs using super cooled
liquid helium. Taking these problems into consideration, reaching 89 % of
the theoretical limit was satisfactory. Especially when regarding the complex
winding configuration of the FBC.
Chapter 4

The System Topology

The system which is studied in this thesis is shown in Fig. 4.1. The compo-
nents in the system are a wind turbine, a gear box, an induction generator,
two transformers and a connection to a main grid and the SMES and conver-
ter system. The power in the system is generated by the wind turbine. This is
connected through an ideal gearbox to an induction generator. The gearbox
has a transmission ratio of 1/100. The wind turbine is on the slow rotating
side. The wind turbine is based on the one used in [15]. This has a rated wind
speed of 12 m/s and rated power of 2 MW. The wind turbine has no pitch
control which makes the power output completely dependent on the wind
speed. This may seem unrealistic, but the scope of this thesis is studying
the compensation of power fluctuations due to wind speed variations using
SMES. A slow pitch control would not contribute to this smoothing. The
induction generator is of the normal squirrel cage type. It has a rated po-
wer of 2 MW. The generator is connected to the grid through a transformer
(T1) having a ratio of 690/1100 V. The main grid is connected to magnetize
the induction generator, and supply the constant frequency and stiff voltage.
The main grid is also the part of the system which it is desirable to control
the power flow to. The converter is connected in shunt with the grid through
a transformer (T2). The ratio of this is 1100/1100 V. The purpose of it is
mainly to act as a filter. The converter is of current source converter type as
distinct from the converters in the project thesis which were voltage source
converter and a DC-DC chopper. All the lines in the system are modeled as
ideal, having no impedance. The capacitances at the input of the converter
acts as a filter together with T2. The system frequency is 50 Hz. Detailed
overview of the wind turbine, the generator and the transformer parameters
can be found in Appendix B.

17
18 4. The System Topology


$







#$% 




!"!

Figure 4.1: Single line diagram of the system being studied.

4.1 Per unit representation of the system

Per unit (pu) system analysis generally makes the calculations a lot easier.
Especially when calculations are made in a system where there are several
voltage levels. If the reference values are selected properly, transformation
to a pu-system removes the transformers from the calculations. All of the
different values of voltage, current, impedance and power will normally be
in the interval 0 - 1.0. This makes the comparison between different voltage
levels easier than if the physical values were used. It is easy to distinct bet-
ween normal state and fault state. There are different strategies for deciding
the base values for voltage. In power system analysis the RMS values of
the phase to ground voltage or line to line voltage is normally set as voltage
reference [16]. However in this project the base for the voltage is chosen as
the peak value of the phase to ground voltage. The following value is chosen
4.1. Per unit representation of the system 19

as the power reference in the system, this corresponds to the rated power of
the generator:
Sr = 2.0 MVA (4.1)
The two voltage references in the AC-part of the system are chosen as the
peak value of the phase to ground voltages of the primary and secondary side
of the transformers. Vr2 is also equal to the rated voltage of the induction
generator. 
2
Vr1 = × 1100 V = 898.15 V (4.2)
3

2
Vr2 = × 690 V = 563.38 V (4.3)
3
The base current and impedance on each side of the transformers are:

2 Sr
Ir1 = = 1484.53 A (4.4)
3 Vr1

2 Sr
Ir2 = = 2366.67 A (4.5)
3 Vr2

Vr1
Zr1 = = 0.605 Ω (4.6)
Ir1

Vr2
Zr2 = = 0.238 Ω (4.7)
Ir2
The reference frequency in rad/s is necessary to calculate the pu-values of
inductances and capacitances:

ωr = 2πfr = 100π rad/s (4.8)

The DC-side of the inverter should also be transformed to pu-values. The
reference value of the DC-link voltage is chosen to be two times the AC-side
reference voltage: 
2
Vdcr = 2 × Vr1 = 1466.67 V (4.9)
3
The reference current is calculated by using the three phase power. Consi-
dering the AC-power equal to the DC-power gives:

3
S3ph = 3Vph Iph = Vr1 Ir1 = Vdcr Idcr (4.10)
2
20 4. The System Topology

Eliminating the voltages gives:

3
Idcr = Ir1 = 1113.40 A (4.11)
4
And the reference impedance for the DC-side:
Vdcr
Zdcr = = 1.317 Ω (4.12)
Idcr

4.2 The converter

The ideal converter topology used to connect the SMES system to the rest
of the network will have no harmonic distortion, no usage of reactive po-
wer and no losses. These requirements are of course not possible to fulfil,
and the choice of topology will be a compromise between the different as-
pects. A line commutated converter using thyristors has low on-state losses
and the thyristor devices can cope with large amounts of power. The disad-
vantages are lagging power factor and high low order harmonics pollution.
Neither does a thyristor converter provide the same degree of control as a
self-commutated converter. The requirement of a present strong grid is also
a drawback considering line-commutated converters. Because of these disad-
vantages a self-commutated converter is chosen.
When choosing the converter type among self-commutating types there
are mainly two different types to choose from. They are the voltage source
converter type (VSC), see Fig. 4.2a and the current source converter type
(CSC), see Fig. 4.2b. In a VSC the DC-voltage will always have the same po-
larity. Bidirectional power flow of the converter is achieved by reversing the
DC-current polarity. In a CSC the DC-current will always flow in the same
direction and the bidirectional power flow is achieved by reversing the DC-
voltage polarity. The VSC has a big capacitor on the DC-side. This is sized
to be large enough to sustain the voltage in the DC-link on a constant level
for the expected operating conditions. Because the current flow can be bidi-
rectional, the so called converter valves also have this feature. As Fig. 4.2a
shows this is normally solved having a diode connected in antiparallel with
the switching device which very often is an IGBT. Because the DC-voltage
never switches polarity, there is no need for reverse blocking capability. A
CSC will need a blocking diode in series connection with the switching device
in absence of reverse blocking capabilities in a normal IGBT. [17].
The CSC and VSC topology have quite different properties, and because
of that they have different advantages and disadvantages. The VSC has the
advantage of having lower harmonic pollution. The CSC needs capacitors on
4.2. The converter 21

(a) Voltage source converter. (b) Current source converter.

Figure 4.2

the AC-side to filter out harmonics. The reason for this is the rapid changing
current pulses which are created from the switching of the continuous DC-
current. The capacitors provide a stiff bus interfacing the converter. These
filter capacitors are expensive and bulky and are clearly a disadvantage com-
pared to the VSC. A drawback with the VSC is that the switches are more
vulnerable to high short circuit currents. If two switches on the same leg are
on at the same time the DC-voltage will be short circuited and a large current
will flow. This can cause failure of the converter. In a CSC the short circuit
current is limited by the coil. In fact a short circuit of one leg normally hap-
pens several times during one cycle. Another drawback regarding the VSC
is the capacitor on the DC-side. This has limited lifetime compared to the
inductor in a CSC [18]. Despite these drawbacks the VSC configuration is by
far the preferred over the CSC. The reasons have normally been economics
and performance.
In this master project the purpose is to study a CSC and its feasibility
in SMES applications. The converter which has been used is the CSC shown
in Fig. 4.3a. This converter is different from the converter used in the pro-
ject thesis which is shown in Fig. 4.3b. The fact that the superconducting
coil is a current source requires another converter in addition to the VSC, a
DC-chopper. This is placed on the DC-side to directly manage the current
in the coil. The process of feed in and extraction of energy is quite different
in the two topologies. In a CSC the energy management is directly to and
from the coil as there is only one converter interfacing the power system. In
the VSC/DC-chopper topology the energy management in the SMES is a
two step procedure. If extracting energy from the SMES, the energy is as
a first step withdrawn from the DC-link capacitor. The capacitor is then
charged again by the superconducting coil. If feeding energy into the SMES,
the capacitor is first charged and then it charges the superconducting coil.
Compared to a CSC solution the VSC will as a consequence of this have
faster response because of the DC-link capacitor which can deliver large cur-
 22 4. The System Topology





 


 





(a) The CSC linking the system and the superconducting coil.











 




 

(b) The VSC and DC-DC chopper linking the system and the superconducting coil in the project thesis.

Figure 4.3: The two different converter topologies to connect the supercon-
ducting coil to the grid.

rents quicker than an inductor as this does not allow instantaneous change
of current. The larger inductance, the slower response time is present. A di-
sadvantage regarding the capacitor is as mentioned its limited lifetime. This
ad to the running expenses of the converter. In a CSC topology the DC-DC
chopper is superfluous. Because of this the control system can be made sim-
pler than for the VSC solution. A drawback of the conventional CSC is the
requirement of blocking diodes. This increases the amount of semiconduc-
tors the current has to flow through, and thereby increase the conduction
4.2. The converter 23

losses. A reverse blocking IGBT would help decreasing these losses as this
does not require a blocking diode. On the other hand the VSC/DC-chopper
topology has fewer semiconductor devices conducting in the VSC part, but it
also has semiconductors in the DC-chopper which contribute to losses. The
number of semiconductors the current flows through is actually higher for
the VSC/DC-chopper solution than for the CSC. The reason for this is that
three half-legs is on compared to two in the CSC. Together with the devices
in the chopper this equals five devices to pass compared to four in the CSC.
However two of the half-legs in the VSC will share the current making the
load on each unit smaller.

4.2.1 Sizing of the converter

The two constraining factors on the switching units are applied voltage and
the current in the SMES coil. The voltage applied gives the rate of change
of the current according to the following equation:

di
VL = L (4.13)
dt
From this follows that the SMES will be able to charge and discharge faster
the higher the applied voltage is. This also implies that the power in the
SMES will be higher. The rated turbine power in the system is as mentioned
2 MW. To cope with power oscillations from the turbine the SMES system
should be able to cover fluctuations in the megawatt range. The purpose of
the system in this thesis is to cover for fast power fluctuations, not for diurnal
variations. On basis of this the rated power of the converter is selected to
be 1.5 MW. As is well known this is equal to 1.5 MJ/s. To store 1.5 MJ the
current flowing, can be calculated using:

1
WL = LI 2 (4.14)
2
The inductance of the superconducting coil is 1.8 H, a value found to be
suitable in the project thesis. The rated current is then 1290 A. This will be
the maximum operating current and the switches must be rated to cope with
this current. Reverse blocking diodes will as mentioned decrease the losses.
However these IGBTs are not rated for high power usage. A prototype device
described in [19] is rated at 600 V and 200 A. The manufacturer IXYS can
currently deliver RB-IGBTs with a rating of 1200 V and 55 A [20]. These
rather poor ratings strongly limit their feasibility in energy storage capability
like SMES.
24 4. The System Topology

ABB has in a document given guidelines to choose the voltage ratings of

high power semiconductors [21]. These are taken into account when devices
have been chosen here. It is important that there are safety factors compared
to the operating voltage. On the other hand the semiconducting devices
should not be selected having to high ratings as this would increase the
losses. The CSC is subjected to AC-voltage, and it is the peak of this which
is interesting. The peak of the AC-voltage is calculated and a certain safety
margin, x is added:
√ x
VACpeak = 2 × VN OM RM S × (1 + ) (4.15)
100
The safety factor varies according to the application, but 15 % is not unusual.
The maximum repetitive peak forward voltage is calculated according to:
y
VDR = VACpeak × (1 + ) (4.16)
100
In the system studied here the results from Eq. 4.15 and Eq. 4.16 are given
below. The semiconductors used in the simulations in this thesis are models

Table 4.1: The dimensioning voltages.

VN OM RM S VACpeak VDR
1100 V 1700 V 2900 V

of real ones from ABB. The IGBTs have voltage rating of 3300 V and current
rating of 1500 A. The voltage rating of the diodes is 4500 V and the current
rating is 1650 A. This also gives safety factor for the current of 15 %. The
devices will not break down at 1500 A. According to the data sheets they
can handle much larger currents for shorter durations. There is also a built
in over current protector having the large inductance coil. Data sheets for
the devices can be found in Appendix F.

4.2.2 Losses
These loss considerations are based on the theory given in [22]. In an ideal
semiconductor the current conduction capabilities are infinite at the same
time as the voltage drop is zero. The transition from OFF-state to ON-state
happens instantaneously. However these characteristics are not present in
the real world, which means there are power losses involved when operating
semiconductor devices. The losses in a semiconductor can be divided into
two categories. Conduction losses and switching losses. The former arise
4.2. The converter 25

vT , iT

Vd Vd
Io

Von

t
tri tf v ton trv tf i

tc(on) tc(of f )
PT

Eon ≈ 12 Vd Io tc(on)

Eof f ≈ 12 Vd Io tc(of f )
Econd

Figure 4.4: The power dissipation in a semiconductor.

because the device has a small on-state voltage. The latter arise because the
current does not rise in an instant when the device is turned on, nor does
the voltage fall immediately. When switching off, the current does not fall
instantaneously and the voltage does not rise in an instant either. Fig. 4.4
shows the principal of losses in a semiconductor which is first switched on,
then conducts for a time and then is switched off. Vd is the voltage which
the device blocks, Io is the current conducted when fully on, Von is the on-
state voltage. There are some exaggerations in the figure to make the points
clearer. The magnitude of Von , which in reality is in the order of a few volts
is exaggerated compared with Vd . The length of ton is on the other hand
shortened compared to the length of tc(on) and tc(of f ) . The coloured areas
correspond to the energies lost during one cycle. The conduction losses are
given by Eq. 4.17.
26 4. The System Topology

Pon = Von Io (4.17)

This equation integrated over ton , the green area, Econd in Fig. 4.4 is given.
The energies lost in the switching actions correspond to the two yellow areas,
Eon and Eof f . These are obtained by integration of PT during ton and tof f .
These energies are lost fs times per second, and the averaged switching losses
can be expressed by:

1
PS = Vd Io fs (tc(on) + tc(of f ) ) (4.18)
2
Based on these equations it is desirable to have a low on-state voltage and
short switching times. These losses are important when choosing the ope-
rating conditions for the converter. The switching frequency which greatly
influences the losses will be looked upon in the next section.

4.2.3 Switching frequency selection

The converter will operate on a fixed switching frequency. The value of
this will influence on the switching losses in the converter and the harmonic
pollution. The higher switching frequency, the higher switching losses, but
lower harmonic content. The selection is therefore a compromise between
these two factors. The maximum switching frequency is dependent on the
minimum pulse width and the temperature on the junction. Guidelines given
in [23] are used when considering this.
The maximum theoretical switching frequency is constrained by the turn-
on and turn-off switching times and the rise and fall times of the current.
These times added up is an estimate of the total switching time, and should
not exceed 5 % of the switching period. The intention of this is just that there
has to be time for current conduction during a switching period in addition
to “lost” time in the switching moments. Each of these times are given in
the data sheet in Appendix F. The voltage rise and fall times are ignored in
this consideration as they are very short compared to the equivalent times
of the current. This identity is given by Eq. 4.19:

0.05
fmax1 = (4.19)
td(on) + td(of f ) + tr + tf

Inserting the values from the the data sheet gives a theoretical maximum
switching frequency of 16.8 kHz. This frequency does not take the heat
which is created on the junction and the heat which is dissipated from the
4.2. The converter 27

junction into consideration. Eq. 4.20 does this, and will in the most cases
give a lower switching frequency than Eq. 4.19.
TJ − TC − P
cond
RθJC
fmax2 = (4.20)
Eon + Eof f

The terms in the equation corresponds to the following:

• T J is the temperature on the junction.

• T C is the temperature on the case.

• RθJ C is the thermal resistance of between the junction and the case.
The unit of this quantity is K/W This unit can be compared with
the ohmic resistance in an electrical circuit which has the unit Ω or
V/A In a thermal circuit the temperature difference corresponds to
the voltage in its electrical counterpart, and the power represents the
current. It describes how much power which is dissipated through the
material depending on the temperature difference.

• P cond is the conduction losses in the IGBT. It is composed by the

current through the device, and the voltage drop. Pcond = VCE × I.

• E on is the energy loss when switching ON the IGBT.

• E off is the energy loss when switching OFF the IGBT.

The last three items in the list was further explained in Sec. 4.2.2. Inser-
ting the data from the data sheet gives the maximum switching frequency
1552 Hz. This frequency is used as a maximum limit, and the best switching
frequency below this is chosen based on simulations.
Chapter 5

Control and modulation

5.1 Control system

The objective of the control system is to smooth out the power flow from
the generator to the power system. Ideally the combined power output from
generator and SMES will be constant. The control unit to maintain this is
rather simple and contains only one PI-regulator, see Fig. 5.1. The input to
this regulator is the difference between the reference power and the measu-
red power flow to the grid. The reference power is a constant value, set in
advance. The PI-regulator is tuned using trial and error, and the output is
forwarded into a block which uses so called abc theory to calculate current
references [24]. This theory and its employment in the system will be further
explained in the next section. The parameters of the PI-regulator are found
to be as following:

Kpi,pu = 0.1
Ti = 0.01

Pref,pu 1 + Ti s Pref,abc
Kpi,pu
− Ti s
Pgrid,pu

Figure 5.1: The PI power controller.

29
30 5. Control and modulation

5.2 ABC theory

The abc theory in this system, also called instantaneous abc theory is used
to construct the reference currents for the modulation. The abc theory is
generally used to compensate for reactive power. However in this system it
is used for compensating active power. The concept of the theory will first
be explained as it is stated in [24]. Then its implementation in this system
will be explained.
The idea of the abc theory is to determine the active part of a total
load current which has both active and reactive components, or active and
nonactive current as the terminology is in [24]. The goal is to deliver the
same amount of energy from the source to the load, without having to trans-
port reactive power on the lines. The method of doing this is to establish a
minimized, instantaneous active current component which fulfil the energy
constrain mentioned. The difference between this new minimized active cur-
rent and the uncompensated load current is the nonactive current which now
comes from a type of compensator, as seen in Fig. 5.2. The determination
of the instantaneous active current is carried out using e.g. the Lagrange
Multiplier Method. The objective of this method is to find the extreme va-
lues of a function whose domain is constrained to lie within some particular
boundaries [25]. The active, ipf , and nonactive, iqf , currents of any given
load current, ik are given by the following equation:
if = ipf + iqf , f = (a, b, c) (5.1)

 






(a) Uncompensated currents.





 

(b) Compensated currents.

Figure 5.2: The concept of active and nonactive currents.

5.2. ABC theory 31

The goal is to minimize the load current, and the constraint is that the
nonactive current components iqa , iqb and iqc do not generate any three-
phase instantaneous active power. The minimization of the load current
will of course be −∞, but as this is a invalid and uninteresting solution the
function which is interesting to find the minimum of is given by:

L(iqa , iqb , iqc ) = (ia − iqa )2 + (ib − iqb )2 + (ic − iqc )2 (5.2)

This will minimize the absolute value of the active load currents ipf . The
constraint is given by:

h(iqa , iqb , iqc ) = va iqa + vb iqb + vc iqc = 0 (5.3)

Eq. 5.4 gives the method of the Lagrange multiplier, where λ is the Lagrange
multiplier:
∇h = λ∇g (5.4)
Eq. 5.4 is applied to Eq. 5.2 and Eq. 5.3 which gives a set of three equations:

−2(ia − iqa ) = λva

−2(ib − iqb ) = λvb (5.5)
−2(ic − iqc ) = λvc

Rewritten this gives the following system of equations:

2ia − λva = 2iqa

2ib − λvb = 2iqb
(5.6)
2ic − λvc = 2iqc
va ia + vb ib + vc ic = 0

Eq. 5.6 is solved for λ:

2(va iqa + vb iqb + vc iqc ) 2P3φ

λ= 2
= 2 (5.7)
2
va + vb + vc 2 va + vb2 + vc2

Eq. 5.7 is inserted into Eq. 5.6 and the following is achieved on vector form:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
iqa ia va
⎢ ⎥ ⎢ ⎥ P3φ ⎢ ⎥
⎣ iqb ⎦ = ⎣ ib ⎦ − 2 2 2
⎣ vb ⎦ (5.8)
va + vb + vc
iqc ic vc

The last term in Eq. 5.8 corresponds to the active part of the total load
current in Eq. 5.1. Because of the constraint given on the minimization the
32 5. Control and modulation

   







 

 

(a) Positive power fluctuation. (b) Negative power fluctuation.

Figure 5.3: The different flow directions of current into the SMES.

active currents in Eq. 5.8 will deliver the same amount of instantaneous active
power to the load as the original load currents. This is expressed as:

P3φ = va ia + vb ib + vc ic = va ipa + vb ipb + vc ipc (5.9)

So how can this theory be applied to the SMES power conditioner? The
idea is that the current from the induction generator also consists of an
active and nonactive part. These do not correspond to active and reactive
power, but reference power and power fluctuations. The reference power is
the power generation from the generator set according to predictions of wind
speed, and the power fluctuations arise according to wind speed fluctuations
around the reference wind speed. The active part of the current gives the
reference power, and the nonactive gives the power fluctuations. This can be
expressed by the following principal equations:
Pref erence
iactive = V = iref erence
P (5.10)
inonactive = f luctuations
V = if luctuations

The meaning of this is that the SMES will supply the nonactive current. If
the power fluctuation is positive the sum of the active and nonactive current
components will be larger than the reference current, and the opposite if
the power fluctuation is negative. The concept is shown in Fig. 5.3. A
5.3. Space vector modulation 33

vital detail is that the nonactive current can flow in both directions. What
distinguish this from the conventional abc theory is of course that it is used
for compensating active power. However reactive power fluctuations will also
be included in this compensation as it is part of both the reference current
and the fluctuating currents. This is a drawback of this control system as
it is not possible to compensate for active and reactive power separately.
Another control strategy could be to implement the pq-theory, which enables
this feature.
The output from the abc block in the control system is three phase current
references which correspond to the left side of Eq. 5.8. These currents are
transformed to a current space vector having an angle θ. A ratio between the
space vector and the DC-current is then calculated in per unit. This is called
the modulation factor m and is forwarded into the modulation block which
generates signals for the self commutating semiconductors in the converter.
The modulation technique is explained in the next section.

5.3 Space vector modulation

The number of research projects involving voltage source converters (VSC)

is as mentioned much higher than for current source converters (CSC) [26].
Because of this the different modulation and control techniques for CSC is
not as well known as for its counterpart. The space vector modulation (SVM)
used in this thesis is based on the modulation strategy for a VSC.
In a converter there are only a limited number of possible switch com-
binations. For a VSC there are eight, and in a CSC there are nine. The
different switch combinations are shown in Fig. 5.4. Three of them (SC7,
SC8 and SC9) give a short circuit of the DC-current. Each of the switch
combinations forms space vectors in the complex dq-plane. The three short
circuit combinations give zero vectors. The other six give stationary vectors.
This is illustrated in Fig. 5.5. Each of the stationary space vectors encloses
a sector together with the nearest space vector. It is readable from Fig. 5.5
→ → →
that the space vector is in sector I can be made by adding iac , ibc and a zero
→
vector. This is the concept of SVM. The space vector is is a transformation
of the currents in the three phase input to the converter. It rotates anti-
clockwise at a rotational speed equal to the frequency in the network. The
space vector is derived from the current equations for a three phase balanced
34 5. Control and modulation























Figure 5.4: The different switch combinations.

system having a frequency of ω and a phase shift of φ [27]:

ia = Im cos(ωt + φ)
ib = Im cos(ωt + φ − 2π/3) (5.11)
ic = Im cos(ωt + φ + 2π/3)
→
The current space vector is is then given by:
→ 2
is = (ia + ib ej2π/3 + ic ej4π/3 ) (5.12)
3
By using trigonometric identities and Euler’s formula [28], and inserting
Eq. 5.11 into Eq. 5.12 the expression for the current space vector becomes:
→
is = Im ej(ωt+φ) (5.13)

Eq. 5.13 shows that the space vector rotates at the frequency of the network,
and at a phase delay equal to φ. Compared to the current on the DC-side the
magnitude of the space vector can be calculated by simple considerations.
→
For instance the space vector iac is formed by the switch combination SC1 in
Fig. 5.4. This state implies that the current in phase a, ia (t), defined positive
5.3. Space vector modulation 35

Im
→
ibc

→
is
→
iba
II I →
iac

III 0 Re

π
3

→ →
ica iab
IV V

→
icb

Figure 5.5: Space vector states for current source converter.

into the converter equals the DC-current idc , and the current in phase c, ic (t)
equals −idc . None of the switches in leg B is connected, hence the current in
→
phase b is zero. This is put into Eq. 5.12, and iac equals:
→ 2 2 2
iac = (idc + 0ej2π/3 − idc ej4π/3 ) = idc (1 − ej4π/3 ) = √ idc ejπ/6 (5.14)
3 3 3
The wanted space vector is formed by applying the enclosing fixed space
→
vectors for fractions of a switching period, Ts . Depending on the size of is a
zero vector must also be applied. The desired space vector must be within
the limits of the straight lines drawn between the tips of the adjacent fixed
vectors in Fig. 5.5. However it is advantageous that the space vector has
constant length, or at least not a maximum value depending on where it is.
Hence the circle spanned within the limits of these lines defines the maximum
space vector magnitude.
36 5. Control and modulation

Im
→
ibc

→
is
T2 i→ →
iac
Ts bc

θ T1 i→
Ts ac
Re

Figure 5.6: The calculation of switch times.

→ → → →
is in Fig. 5.5 is composed of iac , ibc and icc . To calculate how long time
→
each fixed vector must be applied to create is a method from [29] is used.
→
Fig. 5.7a illustrates the principle of time intervals. It is readable that iac
→
has to be applied for TT1 and ibc for TT2 to create the space vector. A zero
s s
vector must also be applied unless the desired space vector is equal to the
maximum space vector. The time intervals are calculated considering SVM
for VSC which sectors are shifted π6 in the clockwise direction compared to
a CSC. Therefore it is a term of π6 in the equations to rotate back again:

Im sin( π3 − (θ + π6 )) − S π3
T1 = Ts (5.15)
idc

Im sin((θ + π6 )) − S π3
T2 = Ts (5.16)
idc
The mentioned zero vector time interval is then given as the time rest in the
switching interval:

Tz = Ts − T1 − T2 (5.17)
The above equations are valid for all sectors. The last term in the numerator
of Eq. 5.15 and Eq. 5.16 is rotating the angle back to sector 0. S corresponds
to the sector the space vector is in. The reason for this term is that the
space vector angle, θ, is measured from the real axis. Having the rotating
term allows the same equations being used for every sector. The ratio Ii m in
dc
5.3. Space vector modulation 37

T1 Tz T2 T1 T2 Tz T2 T1
2 2 2 2

→ → → → → → → →
iac icc ibc iac ibc icc ibc iac

Ts Ts
(a) The distribution of switching states. (b) The distribution of switching states in a
symmetrical modulation pattern.

→
Figure 5.7: The switching pattern to create is in one switching period for
two different patterns.

the equations is the modulation factor, m, which is taken as input from the
control system. This can attain values between 0 and 1. Above 1 there is
overmodulation. This should be avoided as it will not give the desired space
vector as the entire switching period is employed before the space vector
→
is formed. The distribution of the switch states to create is in Fig. 5.5 is
illustrated in Fig. 5.7a. This type of distribution is not the only one possible
to create the wanted space vector. The sequence of the applied vectors is not
affecting the created space vector. There are numerous considerations:

• Which zero vector is the best to apply?

• When should the zero vector be applied?

• In which sequence should the fixed space vectors be applied?

• How many times each switching period should each vector be applied?

The first item is fairly easy to decide. In the example here, the two active
→
vectors creating is both involved the lower half-leg of leg C being ON. Consi-
dering this the zero vector also employing this half-leg being ON would be
preferable. This would lead to a minimum amount of switching, as only two
switches will change its state between the transition from active vector to
zero-vector. This applies of course to the other sectors as well.
When it comes to the second item there is no set answer to this. One
very small advantage of placing the zero vector in the middle of the switching
pattern as in Fig. 5.7a is present. This pattern entails that the last active
38 5. Control and modulation

vector before a sector change will be the first active vector in the new sector.
As this is the case, there is no need for switching during sector changes. This
will decrease the switching losses by a fraction, but considering the sectors
only change six times in a cycle compared to the switching frequency in the
kHz −area the advantage of this switching pattern is almost neglectable.
The third item has the same argument as the previous item. The only
time it matters is at the sector changes, and these are not very often compared
to the switching frequency.
The last item refers to the fact that each of the vectors can be applied
several times during one switching period. Check Fig. 5.7b for illustration of
this switching pattern, which applies the two active vectors two times. The
switching states in the two different cases will produce the same vector as the
applied vectors are ON at equal amounts of time in total. The symmetrical
switching pattern is said to have less harmonic pollution than its non sym-
metrical counterpart [30]. Although this did not give any significant result
in the simulations here. Besides, this switching pattern has higher switching
losses due to the fact that it actually switches twice as often as the simpler
pattern. Because of this, the symmetrical pattern has been abandoned in
this thesis, and the pattern shown in Fig. 5.7a is used.

5.4 Principle of operation

The SMES coil in this system works as an active power compensator. When
the output power from the generator is higher than a preset reference value,
the SMES coil will absorb the extra power and store it in the magnetic field.
This can be called Mode 1. The reference in this system is fixed, but in a real
system the reference power can be set by predicting the future wind condi-
tions on site, and thereby predicting the output power from the generator.
The time span of this prediction cannot be too long as the wind speed is
hard to foresee. Mode 2 applies when the power from the generator equals
the reference power. In this mode the energy in the SMES will be stored in
almost lossless conditions. However there are losses in the semiconductors
in the converter. For longer time intervals in this mode, a superconducting
bypass path would be preferable. A path like this would be slower to ope-
rate than the IGBTs in the converter and the response time of the converter
would be increased. On the other hand the usage of this path would be rare
as the output from a wind turbine is seldom constant for long. Mode 3 is
when the power output from the generator is less than the power reference.
In this mode, the SMES will deliver power to the network. Here it will be
explained how these modes of operation are achieved.
5.4. Principle of operation 39

The error signal for the PI power regulator is the difference between the
power reference and the actual power delivered to the network. The output
of the regulator is then forwarded into the ABC-block. Here it is used to
calculate three phase reference currents. These are transformed to a rotating
space vector, which is scaled with the DC-current and the modulation ratio
is obtained. This ratio together with the space vector angle and switching
frequency is passed into the modulation. This block gives the gating signals
to the IGBTs.