Carte Tutelea Muntean Deaconu

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/320709508
Dual Stator Winding Induction Generator for Wind or Hydro Applications
Book · October 2017
CITATIONS READS
2 3,092
3 authors:
Sorin Ioan Deaconu Lucian Tutelea

Polytechnic University of Timisoara Polytechnic University of Timisoara
107 PUBLICATIONS 352 CITATIONS 57 PUBLICATIONS 266 CITATIONS
SEE PROFILE SEE PROFILE
N. Muntean
Polytechnic University of Timisoara
80 PUBLICATIONS 680 CITATIONS
SEE PROFILE
All content following this page was uploaded by Sorin Ioan Deaconu on 30 October 2017.
The user has requested enhancement of the downloaded file.

1 – Scurtă prezentare a istoriei roboticii 5
CONTENTS
Foreword…………………………………………………………………………..7
Chapter 1. Introduction........................................................................................ 9
Chapter 2. Dual stator windings induction generators (DSWIG):

literature review and new proposed topologies.................. 15
Chapter 3. Analytical model of DSWIG............................................................... 18
Chapter 4. Equivalent magnetic circuit considering the saturation................. 23
Chapter 5. Preliminary and optimal design……………………………………. 28
Chapter 6. Parameter optimal identificationn.................................................... 38
Chapter 7. 2D-FEM analysis……………………………………………………..45
Chapter 8. Control strategies, the DSWIG mathematical model and

dynamic simulations………………………………………………… 57
8.1 Mathematical model of the control…………………………………… 57
8.2 The principle of DSWIG control……………………………………... 62
8.3 Control strategy for self excitation case……………………………… 67
8.4 Control strategy for grid connection case…………………………….. 83
8.5 Simulations of start and stop of DSWIG in motor and generator
regimes………………………………………………………………. 90
Chapter 9. Experimental work………………………………………………….. 93

9.1 Parameters identification results……………………………………… 93
9.2 Parameters optimal identification…………………………………….. 104
9.3 Experimental investigation of proposed topologies………………….. 109
9.4 Load operation of DSWIG…………………………………………….114
Chapter 10. Design and simulation software...…………………………………122
Chapter 11. Conclusions………………………………………………………….125
Appendix………………………………………………………………………….. 128
References................................................................................................... .............179
1 – Scurtă prezentare a istoriei roboticii 7
FOREWORD
This book describes a theoretical and experimental study which was intended
to introduce a low cost (PM less), wide-speed-range operation with reduced capacity of
the static power controller, low power electric wind or hydro generators. It reports the
key design equations, preliminary optimal design aspects, FEM analysis, control
design model and the experimental tests for verifying them in the dual stator winding
induction generator (DSWIG). The proposed induction generator consists of a standard
squirrel-cage rotor and a stator with two separate windings wound for a similar number
of poles. At the first stage the single winding, three phase induction machine is
optimally designed for low speed and low frequency considering the influence of the
pole pairs on the main dimensions, weight, active material cost and efficiency behavior
in a saturation regime (non-sinusoidal). The rated power and cost reduction of the
inverter is then analyzed considering dual stator winding induction generators in two
topologies: with the inverter placed on the excitation windings and with the inverter
placed on the main winding. The analytical model used for optimal design and the
model of the double stator winding is validated through finite element analysis with
specialized software, in order to prove the effectiveness of the design equations and
find the main characteristics of the machine. A comparative study is made between
optimal design and FEM design, with emphasis on the influence of saturation. After
this, a system control strategy using the stator flux orientation is outlined. Finally the
proposed solutions are compared with the single winding induction generators. The
tests were carried out using a three-phase induction generator with a rated power of 6
kVA.
In an effort to reduce the initial cost of a cage-rotor (robust) variable speed
induction generator with full (100%) power PWM inverter, a novel dual stator winding
configuration with a 50% rating PWM inverter and a capacitor and 50% diode rectifier
(for additional active power delivery at high speeds) is proposed for a dc link local
power bus. Because of the variable speed, two power electronic converters (one active
rectifier and one diode rectifier) have to be used to deliver the energy produced to the
dc grid. By using such ac converters, the number of poles can be reduced and the rated
generator frequency can be decreased to an arbitrary nominal frequency. Both number
of poles and the rated frequency will be determined within reasonable limits as part of
an optimization problem. The objective-function of this optimization is to maximize
the product between efficiency and power factor with geometrical and economical
restrictions. Maximum values of this product provide high generator efficiency and
lowest generator current and, in consequence, the lowest cost of the converter.
Machine circuit modeling with performance assessment for steady state and a circuit
model for system feed forward controlled dynamics are proposed and backed up by
8 Foreword
numerical power and efficiency versus speed results. Preliminary experiments on a
purposely built prototype already validate the essentials.
The parameter identification of a dual three phase stator winding induction
machine is approached by a genetic optimal algorithm. The estimated parameters are:
the voltage ratio between main and auxiliary winding, the main winding resistance and
leakage reactance, the cage rotor resistance and reactance, the coupling leakage
reactance between main and auxiliary winding, the magnetization of the non-saturated
reactance, equivalent iron loss resistance and mechanical losses including their
variation with speed. The parameters are calculated from standard loadless and short
circuit tests performed on both stator windings by minimizing the sum of squared
errors between measured and computed currents, active power and reactive power at
several points. The proposed method reduces measurement error influence on the
estimated parameters and these cold also be considered the best constant values
approximation for leakage inductances that have some dependence on the current. A
design method for the capacitor bank is proposed, which take into account the self
excitation phenomenon.
December 2016
Authors
Chapter 1
INTRODUCTION
In the context of rapid technical progress, electrical drives have to ensure

operation with high demands on change and adjustment of speed, starting, braking and
reversing, a movement correlation of working mechanisms of the same production
unit, respectively. All these technical requirements have created prospects for the
development of complex drive systems, using power electronic converters based on
semiconductor components that provide automatic management of production
processes, with low energy, turning to computer and microprocessor [1]-[10].
Despite its simple and robust construction, the motion control for induction
generators should take into account the complexity of the dynamic model which is
nonlinear and variable in time and that the physical parameters of the machine are not
always known with great precision. Under these conditions the motion control means
control of the speed and/or the position, respectively torque control, thus obtaining a
faster torque response as the motion control is more efficient. There are two main
control strategies: scalar control and vectorial control [11]-[18]. Usually, electric
machines are designed to be supplied in a sinusoidal regime. If the generator is
connected to the grid through a static frequency converter, because of the higher
harmonics (non-sinusoidal regime), both its parameters and functional characteristics
values are likely to be different from the case of sinusoidal power supply. Connecting a
dc-ac voltage-source inverter in parallel or series with the loads to provide variable
excitation reactive power has facilitated the control of induction generator systems [6]-
[8], but these schemes have simultaneously injected harmonics into the load currents and
induced ripples into the output voltage. The presence of these higher harmonics will
have the appearance of a deforming and saturation regime in the machine as result,
with adverse effects in general in its operation [19]. The appearance of the deforming
regime in the machine is inevitable because any static frequency converter based on
semiconductor technique produces voltages or currents, which contain, in addition to
fundamentals and harmonic, higher odd time harmonics.
The introduction of distributed generation through renewable sources of
energy has unlocked a challenging area for power engineers. As these sources are
intermittent in nature, variable speed electric generators are employed for harnessing
electrical energy from these sources. However, power electronic control is required to
connect these sources to the existing grid [20]. A wind turbine can be designed for a
constant speed or variable speed operation. Variable speed wind turbines can produce
8% to 15% more energy output compared to their constant speed counterparts.
However, in order to provide a fixed frequency and fixed voltage power to their loads,
they require power electronic converters. The major components of a typical wind
10 Introduction – 1
energy conversion system include a wind turbine, generator, interconnection apparatus
and control systems. Most turbine manufacturers have opted for reduction gears
between the low speed turbine rotor and the high speed three-phase generators. Direct
drive configuration, where a generator is directly coupled to the rotor of a wind turbine,
offers high reliability, low maintenance, and possibly low cost for certain turbines [21]-
[23].
For small to medium power wind or hydro turbines, permanent magnet
generators and squirrel cage induction generators are often used because of their
reliability and cost advantages. Interconnection apparatus are devices that are designed
to achieve power control, soft start and interconnection functions. Very often, power
electronic converters are used as such devices. Most modern turbine inverters are force
- commutated PWM inverters that provide a fixed voltage and fixed frequency output
with a high power quality. For certain high power wind or hydro turbines, effective
power control can be achieved with double PWM (pulse width modulation) converters
which provide a bidirectional power flow between the turbine generator and the utility
grid [24]-[30].
Cage rotor induction generators have been employed to operate as wind
turbine generators and small hydroelectric generators in isolated power systems [1],
[2], due to their practical advantages related to low maintenance cost, better transient
performance, ability to operate without dc power supply for field excitation, and
brushless construction. With the development of the packaged high speed gas turbine
and high speed diesel engine, it becomes increasingly common for the prime mover
and the generator to be connected by a gear reducer. On the contrary, a direct
connection between the prime mover and the generator has many advantages, such as
low noise, high efficiency, and high power density. In a split-wound machine, the
stator winding consists of two similar but separate three-phase windings wound for the
same number of poles. Both stators are fed by the same frequency and the rotor is a
standard squirrel cage. The two stator windings are mutually coupled and small
unbalances in the supplied voltages generate circulating currents. Furthermore, because
of the low impedance of the harmonic currents there is a high level of circulating
currents when a non sinusoidal voltage source supply is used, adding losses and
demanding larger semiconductor device ratings [3]-[7].
The induction machine is very suitable for high speed electric power
generation, although there are many problems to be solved, such as the terminal
voltage of a self-excited induction generator being highly dependent on rotor speed,
terminal capacitance, and load [1], [2], [8]-[16]. These disadvantages of the
performance of self-excited induction generators limit their widespread application.
With the development of power electronics, many approaches have been proposed to
solve these problems. A static reactive power generator (implemented with fixed
capacitors and thyristor-controlled inductors) provides variable compensating
capacitance or variable reactive power to maintain output voltage, but it simultaneously
injects harmonics into the load current inducing voltage ripple [1].
One of the main goals is the reduction of the necessary reactive excitation
power by the arrangement of the main windings versus the excitation windings. Some
Dual stator winding induction generator for wind and hydro applications 11
arrangements of the windings can establish an internal positive load current reaction
and thus can determine the desired effect. Such arrangements can be obtained with the
excitation winding and main winding displaced from one another by a certain space
angle α, as in figure 1, where the rotor winding and the consumer impedance are also
shown.
The attractive advantages of this kind of machines are: better controllability in
the very low speed range and the flexibility of the output generator characteristics. Since
the air-gap flux linkages created by the two stator windings and the induced rotor
currents share the same magnetic stator and rotor cores, the main air gap flux saturation
phenomenon is more complicated than that of the normal single-stator-winding squirrel-
cage induction machine [10]. Because of this complexity, a reconsideration of the main
flux linkage saturation effect is required in the design of the machine and in the
development and practical implementation of speed/torque control algorithms. To avoid
deep magnetic saturation in the stator and rotor cores and rotor and stator teeth, magnetic
design methodologies have been suggested both for the dual stator-winding and
brushless doubly fed induction machines [9], [10]. An approach that includes the
influence of magnetic saturation and iron loss using finite-element analysis in the
performance prediction of the brushless doubly fed induction machine was set. The
proposed finite-element model provides very good steady-state predictions and can be
used for the sizing and design optimization of the machine.
Ie
R e , Le
Ue Im
R m , L m
Um
 ZC
r
Ir
L r , R r
Fig. 1 One phase representation of dual three-phase windings full cage rotor induction generator [45].
The dual-stator-winding squirrel-cage induction machine is the most recent

innovation in the family of induction machinery. Broadly, there are two designs: the first
type has two stator windings wound for the same pole numbers with similar or dissimilar
phase numbers, and the second design has two stator windings with dissimilar pole
numbers with the same or unequal phase numbers [2], [17]. There is also the brushless
doubly fed induction machine that has two stator windings wound for dissimilar pole
numbers and a specially designed nested loop rotor structure that couples the two airgap
flux linkages derived from the two stator windings. The synchronous mode of operation
of this the machine appears to be the most profitable, particularly for converter-based
drive control [10], [13], [18].
The research in this domain is apparently reorienting from permanent magnet
synchronous generators (PMSG) towards the three-phase/multi-phase/single-phase
multi-winding stator, squirrel cage rotor induction generators, chiefly dual-stator
generators in order to overcome the drawbacks of named synchronous generators,
namely [4]-[6], [19]:
- difficulties in obtaining magnetic sinusoidal induction distribution along the
machine air gap and, consequently, the appearance of voltage and current distortions,
that determine increased losses and efficiency decrease;
- construction difficulties and low safety level insertion technology of PMSG;
- cogging torque, that complicates the start up of generator sets (e.g. in the case
of unregulated blades wind generators);
- demagnetization risk due to thermal phenomenon determined by Foucault
currents in the poles material;
- high cost of a large number of poles of low rated rotation speed generators;
- limited short-circuit safety-braking torque;
- difficulties in the excitation control, and thus lack of optimization possibility
at variable load currents, as opposed to the optimization opportunities of the induction
machines.
The dual stator windings induction machine (DSWIM) is frequently present in
the literature as an alternative in regenerative energies conversion [2], [9], [31-33].
Recently, dual-winding machines of various types are being considered for various
motor and generating applications since two stator winding sets offer the possibility of
more flexible energy conversion. For example, energy can be transferred not only
between stator and rotor as in the single winding machines, but also indirectly between
stator winding sets [2], [9], [31-32]. The most significant advantages of this machine
lie in its ability to achieve a constant output frequency at variable rotor speeds. The
ability to send the rotor slip energy back to the system improves the efficiency of the
whole system. Therefore, motor parameters should be known to implement controllers
[32]. To achieve optimal dynamic performance during speed and torque tracking and
ensure energy conversion efficiency, rotor resistance and magnetizing inductance
should be precisely known. Implementation of high-gain current controls in field-
oriented algorithms allows one to reduce the effect of other parameter variations.
Sensorless vector control algorithms are more sensitive to parameters accuracy and one
needs all electrical parameters including stator resistance [37]. An important issue met
in practice is the experimental computation of the circuit parameters. The comparison
between the steady state characteristic curves generated with the parameters and those
obtained by direct experimentation reveals significant differences over the entire range
of slip varying from 0 to 1. This leads to the conclusion that to describe the
performance of the induction machine more precisely, one must modify the parameters
obtained from the classical tests to reduce the differences between the estimated and
real performances. To achieve this goal, the use of parameter identification techniques
appears to be a very promising approach [34].
Standard and advanced field-oriented vector controls of induction motors (IM)
require accurate values of the motor parameters which are used in control algorithm.
Methods of induction machine parameter identification from load tests [34-36]
generally require a power converter, speed sensor and mechanical coupling of two
electrical machines. A Genetic algorithm is proposed in [40] to identify the induction
machine parameter from loadless start transients. The start current waveform
comparison with the simulated current and also speed comparison are used to fit the
machine parameters. The method is required to record the speed, currents and voltage
waveform during transients. Standstill, rotating machine and time domain parameter
identification methods are presented in comparison [38] for multiphase induction
machines parameters identification with reasonably good agreement.
In this book a new optimal method to identify the parameters of the two stator
winding induction machines[39], using only industrial measurement equipment
(Voltmeter, Ampere-meter, Power –meter) from standard tests, (no load and two phase
short circuit) is proposed. This method does not need mechanical coupling of the
induction machine with a load machine and does not need any speed measurement. The
method is also able to deal with measurement error.
The induction machine configuration is quite complex. With windings in slots,
the mmf consists of (in three phase or two-phase symmetric windings) a dominant
wave and harmonics. The presence of slot openings on both sides of the air-gap is
bound to amplify (or influence, at least) the mmf step harmonics. Many of them will be
attenuated by rotor-cage-induced currents. To further complicate the picture, the
magnetic saturation of the stator (rotor) teeth and back irons (cores or yokes) also
influences the air-gap flux distribution producing new harmonics [28]. Main field path
saturation, leakage field path saturation, frequency effects and harmonics influence
make the modeling of IMs particularly difficult. Conventional approaches based on
nonlinear magnetic circuit equation are still dominant for magnetization characteristic
(no load current) and core loss assessment though notable errors are still visible in a
careful analysis [29]. Recently, several authors [3-10] have proposed models that
account for saturation in the generalized equations of alternating current machines.
These models are either based on the generalized equations developed for the
unsaturated machines [28] or they are based on the concepts of reluctance and mmf
[59]. It is the purpose of this paper to develop a model that accounts for saturation in an
induction machine. A comparison between this model and another models suggested in
the literature is established. The model developed is used to study the transient
behaviors of a double stator winding induction generator (DSWIG). The theoretical
results obtained are compared with test results obtained from FEM analysis [60-67].
Also, for offshore wind power, fighter aircraft and ship propulsion applications, a
double stator windings induction generator (DSWIG) based high voltage dc generating
system with one or more diode rectifiers connected to the power winding is studied
[41]-[53]. This system can work with the prime mover in a wide speed range under the
change of load, and has a good performance with a small size static excitation
converter [51]. Additionally, in [52], a variable frequency ac generating system using
the DSWIG is explored to satisfy the fast development of airborne and vehicle power
systems, and the corresponding topology and control strategy are proposed as well.
The induction generator with split stator winding where only the control
winding is connected to a power converter could not cover a large speed range without
oversizing the induction generators. Usually, the low speed induction machines have a
large magnetization current, and if the power converter is placed on the load side and
runs as an active rectifier for the capacitor battery on the control side then the rated
power of the active rectifier are not much larger than the power of the control converter
and allow a wide speed range [45], but the generator efficiency is smaller in
comparison with a single stator winding induction generator with the same power,
speed and size. The main advantages between the DSWIG and classical induction
generators is the possibility to reduce the rated power converter and consequently its
cost but the price paid, according to [45], is the generator efficiency reduction of
around 5% or increasing the size and consequently the cost of the induction generator.
The results are disputable and depend on the power converter price versus copper
price. The efficiency reduction origin is a worse utilization of the copper when the
stator windings are split in two windings: one for active power and the other one only
for reactive power.
In this book we propose a new scheme and control strategies able to convey
active and reactive power across both windings but with a low cost power converter.
Also, for the cage rotor induction generator (DSWIG), the main winding (2/3 slot area)
is connected from a dc bus through a roughly 50% rating PWM inverter. The auxiliary
winding (1/3 slot area) is connected to a 3 phase capacitor and through a diode rectifier
(of 50% power rating) to the dc bus. At low speeds only the main winding works as
controlled by the inverter, while at high speeds the auxiliary winding starts delivering
active power into the dc bus. This arrangement allows for optimal power (energy)
extraction at low and high speeds at reasonably low initial cost and satisfactory energy
conversion efficiency [68-73], [74-79], [87-93].
Chapter 2
DUAL STATOR WINDINGS INDUCTION GENERATORS (DSWIG):
LITERATURE REVIEW AND NEW PROPOSED TOPOLOGIES
There is a large variety of excitation schemes for conventional and DSWIG

generators. According to the primary sources (hydro, wind, thermal energies) and to
consumer needs, a capacitive reactive current systems [15] may be used for excitation,
connected at generator terminals, such as: AC capacitances, synchronous compensator,
diode bridge rectifier or inverter.
Diode
AG bridge VDC
AG Inverter VDC rectifier Cf
Cf
Lf Lf Lf
Inverter C
a) b)
Main
winding mw
(mw)
Diode AG Inverter VDC
AG bridge VDC Cf
Excitation rectifier Cf
winding
(ew)
C
ew
C
Inverter C
C
c) d)
Fig. 2 Induction machine power generation schemes

a) with one winding and an inverter; b) with a diode bridge rectifier and an inverter; c) dual winding
induction generator with a diode bridge and an inverter; d) dual winding induction generator with an
inverter (proposed scheme) [45].
The possible induction machine power generation schemes are identified and
shown in fig. 2 [3]. The scheme shown in fig. 2(a) consists of an induction machine
directly connected to a voltage-source PWM inverter that supplies the dc bus. In this
scheme reactive power is supplied from the dc bus to the induction machine.
Therefore, the inverter has to be designed for the rated power, and for decrease due to
high ripple currents a filter capacitor (C f ) is necessary. To reduce the system cost, the
scheme shown in fig. 2(b) is proposed. In this scheme, the induction generator is
directly connected to the full-bridge diode rectifier to supply the dc bus. The excitation
16 DSWIG existing in the literature and new proposed topologies – 2
to the machine is supplied by means of a PWM-VSI connected across the machine
terminals via filter inductors (L f ). A dc-bus capacitor acts as a voltage source to the
inverter. In this case, the inverter supplies the required reactive power to the machine
and also charging current to the dc-bus capacitor so that it can be maintained at a
predetermined voltage (greater than the line–line peak voltage of the machine). The
rating of the inverter in this scheme is considerably less than the inverter rating of the
scheme shown in fig. 2(a).
To further reduce the size of the inverter, a dual-winding-set induction
generator scheme is also proposed as shown in fig. 2(c). In this scheme, the main
winding set is designed to carry rated active power and directly connected to the diode
bridge rectifier. The excitation winding is designed for higher voltage than the main
winding to reduce the inverter current rating which depends upon the turns ratio of the
two winding sets. Figure 2(d) presents the proposed power generation system which
has an inverter in main winding and direct connected capacitances in excitation
winding. Through the direct connection between the excitation winding and the
capacitors, the generator receives the necessary reactive power: in this way the
apparent power of the main winding inverter should decrease by 40%. In an
autonomous regime, the direct connection at the capacitors unleashes the emf voltage
induction without external supplies. For the generator in fig. 2 (c) the inverter in the
excitation winding injects superior harmonics into the main winding causing additional
losses in the iron and windings, which leads to an efficiency decrease. Because the
inverter connected to the main winding supplies the dc bus where several other
generators are connected, the inverter in the excitation winding is not necessary (fig.
2(d)).
The magnetizing effect of the main current and that the best arrangement of
excitation and main windings are at α = - π/2, that is equivalent to the situation when the
machine rotates against the space displacement of the main winding versus excitation one
(with for slots) and 8/9 short step at the main winding in our case. In this case the mutual
magnetic decoupling of the three-phase windings, simplification of generator mathematical
model, and consequently, an easy control of energy conversion system equipped with such
generator are all realized [11], [28].
The double stator windings machine introduced by T.F. Barton in 1927 and
developed by Ph. L. Alger [54], consists of two similar but separate three-phase stator
wound windings with the same number of poles. Both stators are fed with the same
frequency and the rotor is a standard squirrel cage. A dual stator winding induction
generator with a converter connected to the control winding is used to supplies ac loads
with reduced harmonic distortion [43], [50], [53], [56] or dc loads by adding a diode
rectifier on the load side [1], [11], [57], [67]. The induction generator with dual stator
winding where only the control winding is connected to a power converter could not
cover a large speed range without over sizing the induction generators. Usually the low
speed induction machines have a large magnetization current and if the power
converter is placed in the load side and run as an active rectifier with a capacitor
battery on the control side then the rated power of the active rectifier is not much larger
than the power of the control converter and allows a wide speed range [39], but the
generator efficiency is smaller in comparison with a single stator winding induction
generator with the same power, speed and size.
In order to provide better utilization of the copper for DSWIG, a new scheme
is proposed here (figure 3) which contains an active rectifier (inverter) and a diode
rectifier, both designed at half of the rated power. In this configuration both windings
convey active and reactive power through the machine. At low speeds only the main
winding delivers active power to the dc link through the PWM converter, while at high
speeds the auxiliary winding adds more power to the dc link. In general the rating of
the PWM converter (active rectifier) is around 50% while the auxiliary winding may
add an additional 50% in active power. ili This arrangement allows for optimal power
(energy) extraction at low and high speeds at reasonably low initial cost and
satisfactory energy conversion efficiency.
Fig. 3 The proposed solution for variable speed wind or hydro energy conversion system [82].
All the induction generator energy is available in the dc circuit while in [9],
[43] a part of energy is available only for some ac unpretentious loads that are
connected directly to the auxiliary winding (figure 4), and as expected the diode
rectifier may be eliminated. In the wind turbine applications the unpretentious load will
be connected only when available power is greater than 50% of the rated power, which
means a turbine speed that is greater than 80% of the rated speed.
Fig. 4 Proposed solutions for unpretentious loads [82].

Chapter 3
ANALYTICAL MODEL OF DSWIG
The dynamic model of the DSWIG expressed in the complex variable form is set
forth as follows [11] when the parameters on the rotor and the excitation windings are
all converted to the main winding. The voltage and fluxes equations in the stationary
reference frame are given as:
U m  R m  I m  1   m , (1)
U e  R e  I e  1   e , (2)
U r  R r  I r  1  r    r , (3)
 m  Lm  I m  Lme  Ie  Lmr  I r , (4)
 e  Le  Ie  Lme  I m  Lmr  I r , (5)
 r  L r  I r  L mr  I m  I e  , (6)
where the subscripts m, e, and r denote the variables associated with the main winding,
excitation winding, and the equivalent rotor winding, respectively, ω 1 is main electrical
frequency, ω r is the angular rotor electrical frequency, L mr and L me are the mutual
inductance and the amplitude of the mutual leakage inductance between main and
excitation windings respectively:
L m  L m  L me  L mr , (7)
L e  L e  L me  L mr , (8)
Lr  Lr  Lmr , (9)
I0  Ir  Im  Ie , (10)
where the subscript σ denotes the leakage inductance. From the mathematical model of
the DSWIG, the equivalent circuit can be shown in fig. 5.
Ie Re Le
jrr
Lme Lr Ir Rr
Im Rm Lm + 
I0
Ue
U0 Lmr U r= 0
Um
Fig. 5 Complex vector model of the DSWIG under steady state [45].
The equivalent circuit of the DSWIG shown in figure 6 is using the ‘1’ indices for
main winding, ‘2’ for auxiliary winding and ‘r’ for rotor winding. Auxiliary and rotor
windings were reported to the main winding [39], [45], [80].
I1 R1 L1
L12
I’2 R’2 L’2
R’r
V1
I’2a
V’C0 Lm
C’ L’r
Fig. 6 Equivalent circuit of the proposed scheme (figure 3) in no load conditions, with capacitor in the
auxiliary winding [82].
Equations defining the reported parameters are listed below:

I2 V N  kw 1
V2'  k  V2 ; I '2  ; k  1e  1 , (11)
k V2e N 2  kw 2
1
L'2  k 2 L1 ; R '2  k 2 R 2 ; L'f  k 2 L f ; C '  C; '
VDC  kVDC . (12)
k2
where V 2 is phase auxiliary voltage, V’ 2 is phase auxiliary reported voltage, I’ 2 is

phase auxiliary reported current, k is ratio of transformation, N 1 is number of main
winding turns per phase, N 2 number of auxiliary winding turns per phase, V 1e is back
e.m.f. main voltage, V 2e is back e.m.f. auxiliary voltage, k w1 and k w2 are main and
auxiliary winding factors, L’ 2σ and R’ 2 are reported auxiliary inductance and resistance,
L’ f is reported inductive filter inductance, C’ is reported excitation capacitance placed
in the auxiliary windings and V’ DC is the reported d.c. link voltage. With main winding
connected to the active rectifier and auxiliary winding connected to the capacitor bank,
and considering firstly the diode rectifier blocked (case a) the induction generator
equations are:
 
  I 
 V1   R1  j1 L1  L12  L m  j1 L12  L m  j1L m  1 
 
 0 

j1 L12  L m  
R 2  j1 L'2  L12  L m
'
 j1L m  I '2 
 ' 
 0
  


j1L m j1L m
R 'r
s

 j1 L'r  L m 
 I r 


(13)
V  Z  I ; I a  Z 1  V , (14)
1
V 'C0    I'2a . (15)
j1 C'
If the capacitor voltage is larger than equivalent dc voltage (V’ CO > V’ DC ) then
the diodes rectifier is in conduction (case b) and the inductive filter is in series with
auxiliary winding as in the equivalent scheme shown in figure 7. The circuit
20 Analytical model of DSWIG – 3
complexity could be reduced to a simple circuit (figure 8), by applying the
superposition principle and Thévenin's theorem for a.c. circuits and considering the
linearization values of the magnetization inductances.
I1 R1 L1
L12 V4 I’r
L’f R’f If I’2 R’2 L’2 V3 I’s
V1
R 'r
I’2a s
V’2
V’C Lm
C’ L’r
Fig 7 Equivalent circuit of the proposed scheme (figure 3) for case b [82].
R’f L’f Z2i
V’2 V’C
V’C0
Fig. 8 Reduced equivalent circuit of the figure 7 [82].
The corresponding equations are:

2 '
V '2   V DC , (16)

V '2  R '2f I'f  jX '2f I'f  V 'C 0 , (17)
1
Z 2ic  , (18)
' 1
j1C 
Z 2i
1
Z2i  R '2  j1L'2  , (19)
1 1

R1  j1L1 Z3C
1
Z3c  j1L12  , (20)
1 1

j1L m R 'r  j1L r
Z'2f  
 R 'f  j1L'f  Z 2ic ; R '2f  real Z'2f ; X '2f  imag Z'2f .   (21)
The solution of the equation (17) is [83]:

I'2f 
 V2' R '2f  V  R   V
'
2
' 2
2f
'2
C0 
 V2'2 Z'22f
, (22)
Z'22f
 
j  '  d 2   I'  X ' 
I'2f  I'2f d2  a tan  ' 2f ' 2f ' ,
 Vco ;
e (23)
 V  R I 
 2 2 f 2f 
and the equivalent currents and voltages are:
 
j  '  d 2 
V '2  V2'  e  ; V 'C  V '2  I'f  Z'f ; I'2 b   I'2f  V 'C  j1C' ,
Vco  (24)
V1  V3' b
V3' b  V 'C  Z'2  I'2b ; I1b  '
; Isb  I1b  I'2 b , (25)
R1  j1L1
V'4b
V '4b  V 3' b  j1L12  Isb
'
; I'rb   , (26)
R 'r
 j1L'r
s
where current I’ f is calculated from the passivation method, V’ CO is the no load voltage
(in case a) and Z 2i is the equivalent impedance for the circuit from figure 7, obtained
by passivation of auxiliary winding.
Through the experimental tryouts we would need, as a first step, the
computation of the DSWA parameters and characteristics for the stationary regime and
their comparison with the values obtained through finite element analysis. In figure 5
the equivalent phase scheme of the machine is presented. Resistances R m and R e are
measured in DC current and the inductances will be computed through no load (real
and ideal) and short circuit methods [85].
A. No load probe
The equivalent phase scheme for the no load methods is presented in figure 9.
The equations which characterize the functioning in these conditions are given as
follows:
Rm Lm
Vm
Re Le
Ve L’2r
R’r
Fig. 9 Equivalent phase scheme of DSWA in no load probe [85].
V mo R m  jX σm   I1mo  V 'eo  k e  V eo , (27)

 
V 'e  R 'e  jX 'σe  I1' e  V 2 m , (28)
22 Analytical model of DSWIG – 3
I1e
k e  V e  R e  jX σe   k e 2   V 2m , (29)
ke
k e V e  R e  jX σe   I1e   V 2m . (30)
B. Short circuit probe
The equivalent phase scheme in short circuit (figure 10) and the equations are given
as follows [85]:
Rm Lm
Vmsc
L’e L’r
R’e R’r
I’esc A
Fig. 10 Equivalent phase scheme of the DSWA in short circuit probe [85].
V msc R m  jXσm   I1msc  V'esc  k e  Vesc . (31)

V mo R m  jX σm   I1mo V
 eo , (32)
V msc R m  jX σm   I1msc V esc
V eo
V mo R m  jX σm   I1mo   V msc  R m  jX σm   I1msc  , (33)
V esc
V eo
V mo   V msc  R m  jX σm I1mo  I1msc  (34)
V esc
V eo
V mo   V msc
V esc
R m  jX σm  , (35)
I1mo  I1msc
A ratio is made between relations (31) and (35) and the result is:
X 'e  L'e , (36)
 
V msc R m  jX m   I1msc  R 'e  jX 'e  I'esc . (37)
with the help of the previous measurements and equations the stationary regime
parameters are computed for the DSWIG.
Chapter 4
EQUIVALENT MAGNETIC CIRCUIT CONSIDERING THE SATURATION
One of the main goals is the reduction of the reactive excitation power needed
by the arrangement of the load windings versus the excitation windings. Some
arrangements of the windings can establish an internal positive load current reaction
and can thus determine the desired effect. Such arrangements can be obtained with the
excitation winding and load winding displaced from one another by a certain space
angle α, where the rotor winding and the consumer impedance are also shown. Because
of this complexity, a reconsideration of the main flux linkage saturation effect is called
for in the design of the machine and in the development and practical implementation
of speed/torque control algorithms. To avoid deep magnetic saturation in the stator and
rotor cores and rotor and stator teeth, magnetic design methodologies have been
suggested both for the dual stator-winding and brushless doubly fed induction
machines [25], [72].
In the literature there are known various mathematical models associated to
induction machines fed by static frequency and voltage converters. The majority of
these models are based on the association between an induction machine and an
equivalent scheme corresponding to the fundamental and a large number of schemes
corresponding to the various ν frequencies, corresponding to the Fourier series
decomposition of the generator output voltage.
The polar coverage factor is denoted with  i and has the expression [85]:
b
i  i , (38)

where b i is the length of the equivalent polar shoe (=bi*B δ ) and  the pole pitch. The
value of  i depends on the type of the winding and on the magnetic core saturation.
When the saturation is low, it may be considered that the flux density in the air-gap has
a sinusoidal repartition along the pole pitch and:
2
i  . (39)

In the case of magnetic saturation, the mmf of the iron U Hfe is no longer
negligible with respect to the mmf of the air-gap U H , a deformation of the magnetic
curvature takes place, and the polar coverage factor depends on the saturation factor k s
[28]:
U  U Hts  U Htr
k st  Hδ , (40)
U Hδ
for the teeth saturation and a new proposal:
24 Equivalent magnetic circuit considering the saturation – 4
U Hδ  U Hts  U Htr  U Hys  U Hyr
k sα p  , (41)
U Hδ
for the teeth and yoke saturation, where U Hts and U Htr represent the mmf in the stator
and rotor teeth, and U Hys and U Hyr represent the mmf in the stator and rotor yoke.
The form factor k f represents the proportion between the effective value of the
flux density and its mean value:
B
kf  . (42)
B med
At a sinusoidal repartition of the flux density along the pole pitch it is
obtained:

kf  . (43)
2 2
In the references [28], [29] it was shown that if there is yokes saturation (both
stator and rotor), this influences the distribution of the flux density in the air-gap in the
opposite way. The teeth saturation leads to a bending of the flux density curvature
(B 1flat ), and the yoke saturation leads to a sharper peak (B 1 peak ) (figure 11). If the level
of teeth and yoke saturation is identical, then the distribution of the flux density stays
sinusoidal (B 1 ).
The reduction of the yoke equivalent field line is given in the literature in the
shape of tables which depend on the yoke flux density, or depend on the pole pair
number. The tables are determined for a certain geometry (figure 12 for instance) and
for a certain magnetization curvature of the ferromagnetic material. When the yokes
(stator and rotor) are saturated, the analytically calculated mmf is very different with
respect to the one computed using the finite element method [85].
B()
B1 peak (yokes saturation)
B1 (sinusoidal)
B1 flat (teeth saturation)

0 P()
Fig. 11 Sinusoidal, flat and peak air-gap flux density [85].
Considering the air-gap distribution of the flux density sinusoidal, sometimes

with the 3rd harmonic included, the yoke flux density will be computed in 4-5 points,
and the corresponding mmf is calculated through digital integration using the Simpson
formula:
D π 1
 H y dl  2xp  c x , (44)
with
 
cx  f By . (45)
Stator
yoke
Stator
teeth
Magnetic
Field Line
Rotor
teeth
Rx
Rotor
yoke
Shaft
Fig. 12 One pole geometry of the DSWIG.
Usually, the magnetization curvatures H = f (B) are given by the producer up

to a certain value of the flux density B (for instance 1, 8 - 2T). Very often when
designing modern electrical machines there are areas in which the flux density is
higher than the maximum value obtained in the magnetization curvature, which require
high torque densities with respect to the mass and the volume of the machine. A
method of extrapolation of the magnetization curvature is proposed, and this method is
based on the property of relative magnetic permeability of the magnetic material  r to
tend to 1 and is also based on the property of differential permeability  d to tend to 0
when the flux density becomes infinity:
B
μr  , (46)
μ0  H
dB
μd  . (47)
μ 0  dH
The continuity of both the permeability  r and  d is imposed, in the last points
of the given magnetization curvature. The intensity of the magnetic field H x in a
random point may be written as:
   aB
1
Hx  Href end  Bx  Bref end  B0 ea1Bx  e 1 ref end  ,
μ0
(48)
where H refend and B refend are the intensity of the magnetic field, and the flux density
respectively in the last point from the characteristic, and a 1 and B 0 are variables
26 Equivalent magnetic circuit considering the saturation – 4
calculated so that the function is continuous and convergent. They have the
expressions:
1
a1  ln1  μ d1 1  μ d 2   , (49)
x1  x 2
Bn  2  Bn 1
x1  , (50)
2
B  Bn
x 2  n 1 , (51)
2
dH
μ d  μ 0  ref , (52)
dBref
H ref n 1  H ref n  2
μ d1  μ 0  , (53)
Bref n 1  Bref n  2
1
B0  μ d 2  1  e  a 1x1 
. (54)
a1
Figure 13 shows the dependence between the flux density and intensity of the
magnetic field (extended magnetization curvature).
Magnetization curves
3
C1
C2
C0
2.5
2 Cref
C0
B (T)
C1
C2
1.5
0.5
0 50 100 150 200 250 300
H (kA/m)
Fig. 13 Magnetization curves: Cref –references points, C0 –interpolation and extrapolation on original
curve (w2b=0, kFe=1), C1 – interpolation and extrapolation on modified curve with w2b=1, kFe=0.97, C2
– interpolation and extrapolation on modified curve considering w2b=2, kFe=0.97 [85].
Designing classical electrical machines implies that the yokes (rotor and stator)
are not saturated, because it is not an issue of weight, of cost or of size. Only in the last
few years has designing implied the reduction of these sizes as much as possible
through an optimal production, leading to saturated areas in the yoke, overlapping the
teeth saturation phenomena.
In the case of yokes saturation at the nominal frequency, losses are very high.
When discussing the low frequency generating electrical generators, where the shaft is
coupled directly to the wind turbine, there is a possibility of yoke saturation without
any significant increase of iron losses. Flux density in the yoke is calculated with the
relationship:
 w 1
Bref 1  Bref  μ 0  H ref   2b  , (55)
 k Fe 
where
h st
w 2b  , (56)
hy
the depth of the slot is denoted by h st , and the length of the yoke where the flux density
is computed is denoted by h y . In the case of the teeth:
w
w 2b  , (57)
b
w being the length of the slot and b the length of the teeth. Taking into consideration
both the main and the nonlinear leakage inductances, dependent on the current, the
issue of computing the current at a given voltage with respect to the slip if the
inductances are not constant is under discussion. An iterative solving of the circuit is
made, starting from the linear inductance around the nominal load. The current is
computed and the curvatures of the inductances can then be iteratively computed.
Chapter 5
PRELIMINARY AND OPTIMAL DESIGN
Many unknown parameters are involved in the design of the Dual Stator
Winding Induction Generator. As a result, it is necessary to assign some description to
these parameters. They will be further explored in the design equations. Table I gives a
list of the parameters used in the design approach [39].
Table I. Design parameters

Symbol Description
p number of poles
D si inner stator diameter
D so outer stator diameter
N ss stator slots number
N rs rotor slots number
li ideal core length
h ag air-gap height
kc Carter factor for air-gap
h ss total height of stator slot
0 permeability of the air-gap
h rs total height of rotor slot
mwp main winding position
layersm number of layers of main winding
layerse number of layers of excitation winding
ncme number of elementary conductors on main winding
ncee number of elementary conductors on excitation winding
sbm number of turns per coil on main winding
sbe number of turns per coil on excitation winding
dcme diameter of elementary conductors on main winding
dcee slot height
rho_copper stator winding resistivity at 200C
alfa_copper temperature winding coefficient at 200C
Tw winding temperature
S copper copper section
Js current density
Is stator current
Uf phase voltage
f frequency
Pn rated power
n speed
U DC DC voltage
The preliminary parameters of the prototype design are given in Table II and
some of them are presented in figures 14, 15 and 16 [39], [45].
Table II Preliminary parameters and machine design dimensions

Symbol Description Value
p number of poles 8
D si inner stator diameter 180 mm
D so outer stator diameter 267.8 mm
N ss stator slots number 72
N rs rotor slots number 58
li ideal core length 230 mm
h ag air-gap height 0.35 mm
kc Carter factor for air-gap 1.22
h ss total height of stator slot 30.5 mm
0 permeability of the air-gap 12.56 10-7 H/m
h rs total height of rotor slot 31.5 mm
mwp main winding position up
layersm number of layers of main winding 1
layerse number of layers of excitation winding 2
ncme number of elementary conductors on main winding 24
ncee number of elementary conductors on excitation winding 24
Rm main winding phase resistance 2.4 Ω
Re excitation winding phase resistance 3.78 Ω
Rr rotor phase resistance 2.36 Ω
L σm main winding leakage phase inductance 14.5 mH
L σe excitation winding leakage phase inductance 0.6 mH
L me mutual inductance between main and excitation winding 14 mH
L σr rotor winding leakage phase inductance 15.3mH
S copper copper section S 1net = 129 mm2
phase connection Y
Is stator current I 1 = 10,38/6 A
Uf phase voltage U 1 = 139/240 V
f frequency 15 Hz
Pn rated power 6 kVA
n speed 415 rpm
U DC DC voltage 460 V
ke main winding to excitation winding voltage ratio 24/31
It presents data entered into the program optimization, optimization variables

and main outcomes for DSWIG:
%Primary Dimension, Technical request and Technological limitation

%Technical request
Pn=3; % KW -rated power

fn=15; % Hz - base speed
Vfn=220; % V - phase voltage
regim='g'; % functioning regime
poles=8; % number of poles
nphase=3; % number of phase
ParallelPaths=1; % parallel current path
%Optimization variable limitations
30 Preliminary and optimal design – 5
elsp_min=8; % kA/m Minimum Specification electric load
Bagsp_min=0.3; % T Minimum Specification magnetic induction in air
gap
lcpertau_min=0.4; % mm Minimum value of Core stack length
Js_min=2.5; % A/mm^2 Minimum Stator current density
Jr_min=2; % A/mm^2 Minimum Rotor current density
sBt_min=1.2; % T Minimum Specification magnetic induction in
stator Tooth
sBy_min=1; % Minimum magnetic induction in rotor yoke
rBt_min=1.2; % T Minimum magnetic induction in stator yoke
rBy_min=1; % T Minimum magnetic induction in rotor yoke
q1_min=2; % number of stator slots per pole per phase
sMs_min=1; % mm Minimum value of Stator open width
rMs_min=0.3; % mm - Minimum Mouth of rotor slot
sh4_min=0.3; % mm - Minimum stator coil height
rh1_min=0.3; % mm Minimum height of mouth of rotor slot
cSpan_min=0.66; % Coil open
% Electrical rated parameters
Pn=3.000000; % W - rated power
fn=15.000000; % Hz - rated frequency
Vfn=220.000000; % V - dc voltage
I1n=5.611817; % A - Rated current
etan=0.848254; % Rated efficiency
cosphin=-0.809967;% Power factor
Pcu=427.803823; % W - Rated copper loss
Pfe=72.871800; % W - Rated iron loss
Pmec=36.000000; % W - Mechanical loss (given in input file)
sR=2.682759; % Ohm Stator (dc) resistance
rR=2.431359; % Ohm Rotor (dc) resistance
sR=2122.155390; % Ohm Iron loss equivalent resistance
Lm0_sat=0.886296; % H Magnetization inductance (at rated field)
Lsl=0.018615; % H Stator leakage inductance
Lrl=0.022952; % H Rotor leakage resistance
rJ=0.513745; % kgm^2 Rotor inertia
elsp=16.224222; % kA/m -Specification electric load
Bagsp=0.850000; % T –Air gap flux density
lcpertau=1.210553;% Ratio of core stack length per pole pitch
Js=2.500000; % A/mm^2 Stator current density
Jr=2.000000; % A/m^2 Rotor current density
sBt=1.916762; % T - Flux density in stator Tooth
sBy=1.347754; % T - Flux density in stator Yoke
rBt=1.870644; % T - Flux density in rotor Tooth
rBy=1.183570; % T - Flux density in rotor Yoke
sh4=0.300000; % mm - Top of stator teeth height
sMs=1.000000; % mm - Stator slot open width
rMs=0.671976; % mm - Rotor slot open width
q1=3.000000; % Number of stator slots per pole per phase
cSpan=0.777778; % pu - Coil open
rSlots=60.000000 % Number of Rotor slots
For ease of practical realization of the experimental model, the laminations

stack of the stator and rotor that are common in factories manufacturing profiles where
chosen. The data are presented in Table II.
Fig. 14 Main dimensions of stator and rotor from optimal design.
Stator slot dimension

30
25
20
(mm)
15
hs1
hs3
hs4
shOA
10
0
0 20 40 60
steps
Fig. 15 Stator slot dimensions from optimal design [84].

Rotor slot dimension
25
hr1
hr2
hr3
hr4
20
15
(mm )
10
0
0 20 40 60
steps
Fig. 16 Rotor slot dimensions from optimal design [84].
The general optimization problem can be divided into three quasi-independent

sub-problems: choosing the objective function, choosing the optimization variables and
the machine model, and then solving the problem. The optimal design of the DSWIG
generator is subject to multi-objective criteria and constructive constrains such as:
reducing the initial cost, reducing the generator size and weight, improving the
efficiency, and limiting the components’ temperature to a feasible level. The multi-
objective criteria should be aggregated into a single objective function if the design
objective is to obtain a unique solution. This could be a total cost function including a
penalty for unmet constraints. The objective function, C t becomes [16], [45]:
Ct  Ci  C E  C a  C p , (58)
where C i is the initial cost, C E the cost of the lost energy, C a an additional cost to
consider the impact of the machine size and finally C p is the penalty cost. Sixteen
optimization variables are used to control the machine main size dimensions as well as
the slots details. The probability to reach the global optimum using the Hooke Jeeves
(HJ) algorithm could be increased by starting the algorithm several times from
different points of the optimization variable space. A few sample results for the HJ
optimization evolution to the best design are shown in figures 17, 18 and 19.
Using the optimization program, the variation of the initial cost and total
expenses for the machine was observed (for 10 years of functioning) with different
numbers of poles, 2p: 8, 12, 16, 18, 20, and 24. These results along with other
characteristics are shown in Table III [45].
Specific flux density
1 2.2
Air-gap
Stator tooth
0.8 Stator yoke
etan 2 Rotor tooth
Power factor Rotor yoke
0.6
1.8
0.4
1.6
0.2
Flux density (T)

Efficiency
1.4
0
-0.2 1.2
-0.4
1
-0.6
0.8
-0.8
0.6
0 10 20 30 40 50 60
-1
0 10 20 30 40 50 60 steps
steps
a) b)
Fig. 17 a) Efficiency and power factor; b) specific flux density [84].
25 Magnetization inductance
Stator Coils
Cage
Cage Bare
Cage Ring
1
20
0.8
15
Weight (kg)
Lm (H)
0.6
10
0.4
5
0.2
0 0
0 10 20 30 40 50 60 10 20 30 40 50 60 70 80
steps I1 (A)
a) b)
Fig. 18 a) Weight; b) magnetization inductance [84].
Power losses 300
3500
copper
Pcu iron
Pfe cage
Pmec
3000 250
2500
200
Material cost (USD)
2000
Power (W)
150
1500
100
1000
50
500
0
0 10 20 30 40 50 60
0 steps
0 10 20 30 40 50 60
steps
a) b)
Fig. 19 a) power losses; b) material cost [84].
Table III Comparative study of DSWIG depending on poles number
Poles number 2p 8 12 16 18 20 24
Base frequency fn [Hz] 15 24 32.5 36.5 40.5 49
Fabrication cost i_cost [USD] 512.8 425.1 399.1 398.8 384.7 403.3
Objective function (total cost) t_cost [USD] 1478.8 1385.3 1355.8 1367.7 1399.7 1431.2
Efficiency etan 84.83 84.9 84.95 84.79 84.18 84.01
Power factor cosphin 0.81 0.75 0.73 0.72 0.7 0.69
Weight active materials weightM [kg] 88.8 68 58.4 55.4 51.8 48.8
Optimization time sim_time [s] 10 23.8 12.8 8.9 6.8 10.2
Air-gap hag [mm] 0.35 0.35 0.35 0.35 0.35 0.35
Rated current I 1n [A] 5.61 6.05 6.19 6.31 6.54 6.56
Outer stator diameter D so [mm] 386 409 484 519 533 594
Inner stator diameter D si [mm] 286.3 319.6 393.9 433.4 450.1 519.7
Length of stator laminations pack lc [mm] 139.3 107.1 72.0 65.1 60.2 55.3
Air-gap flux density B agsp [T] 0.85 0.85 0.83 0.81 0.81 0.73
Specific solenation elsp [kA/m] 16.22 15.31 15.94 15.0 14.45 13.07
Analyzing the results which were obtained, it can be stated that the cheapest
machine is the one with 20 poles, the one with the lowest global expenses is the one
with 16 poles (which has the highest efficiency), the lightest is the machine with 24
poles; the machine with 8 poles has the best power factor, etc. Depending on the data
presented, a decision has to be made about which option to use for the experimental
model. At 6 KVA electric power for 2p=16 poles (table III) the slip is (-0.05) in figure
20, efficiency 0.84 and power factor 0.73 in figure 21 and at rated current the torque is
higher than 170 Nm in figure 22 [45].
10
- 10
Electric
- 20
- 30
Mechanical
- 40
- 50
-60
-1 -0.8 -0.6 -0.4 -0.2 0
Slip
Fig. 20 Electric and mechanical factor versus slip [45]. Fig. 21 Efficiency and power versus slip [45].
60
50
40
Current (A)
30
20
10
0
-18 -14 -10 -6 -2 0
Output power (kW)
Fig. 22 Current versus torque [45].

The power electronics cost reduction is the main reason to replace the standard
induction generator with full power inverter presented in fig. 2.a, with the DSWIG
generator with reduced power converter presented in fig. 2.c and fig. 2.d. The power
converter KVA reduction by DSWIG (fig. 2 c, d) is presented in table IV: dPinv_c1
respectively dPinv_d1 assume that only the magnetization inductance, reactive energy
is supplied through excitation winding while dPinv_c2 respectively dPinv_d2 assume
all reactive energy is supplied through excitation winding. The first case is favorable
when the power converter is placed on the excitation side (fig. 2 c) but it is possible
only if a capacitive compensator is placed between main winding and diode rectifier
[45].
Table IV DSWIG impact on power converter KVA reduction
Poles 8 12 16 18 20 22
dPinv_c1 [%] 53 45.87 44.82 42.94 42.17 42
dPinv_d1 [%] 11.7 15.92 16.6 17.88 18.42 18.53
dPinv_c2 [%] 41.35 34.02 32.15 30.61 28.15 27.93
dPinv_d2 [%] 19 24.85 26.54 27.99 30.45 30.68
The converter KVA reduction seems to be larger for the scheme presented in
fig. 2.c but this scheme also needs an almost full power diode rectifier and dc-dc boost
converter. The KVA reduction of the excitation inverter (dPinv_c) is decreasing with
the number of poles while the main power converter (dPinv_d) KVA reduction
increases with the number of poles because for large numbers of poles the
magnetization current becomes larger.
The windings of the double stator winding induction generator could be
derived from single a winding induction generator in two main ways: either a) the total
cross section of the both windings will be preserved, or b) the stator copper loses will
be preserved. In both cases the number of turns per coil and the coil length will be the
same as for the single stator winding induction generator.
a) Given total stator winding cross section.
The surface assigned to each winding is a part of the total area in the slot and it
is in direct ratio with the current in the windings [45]:
swe  ke scu1
swm  km scu1 . (59)
ke  km  1
The magnetization current and torque (power) current are orthogonal. The
reactive stator current component is also orthogonal with regard to the active
component. The following equation could be written in this case:
I12  I e2  I m2 , (60)
 Ie 
  arg sin  . (61)
 I1 
A set of k e , k m coefficients according to condition (59), including the winding
cross section surface in direct ratio with the currents is computed in (62):
sin  
ke 
sin    cos  
. (62)
cos  
km 
sin    cos  
The total stator copper losses per phase are:
R R
Pcu1_ DSWA  I e2 Re  I m2 Rm  I12 sin 2   1  I12 cos 2   1 . (63)
ke k
m
Considering the expression of the k e and k m from (62) the DSWIG stator copper
losses become:
Pcu1_ DSWA  Pcu1 1  sin  2   . (64)
The k e , k m coefficients are always positive and then it is clear that copper losses
for DSWIG are larger than for the single winding induction generator. It could be
doubled when the active and reactive current components are equal, a situation close to
the rated DSWIG regime.
b) Given copper stator losses
Increasing the copper cross area is a method to reduce the copper losses. In this
case we impose the same losses as for single stator winding asynchronous generator. In
this condition from (62) we obtain:
sin 2   cos 2  
  1. (65)
ke km
A single solution of (65) is obtained if a minimum of stator copper cross area is
imposed and k e and k m are positive:
ke  sin 2    sin   cos  
. (66)
km  cos 2    sin   cos  
The total stator winding increase area coefficient is:
ke  km  1  sin  2  . (67)
When the active power is equal with the reactive power the DSWIG stator
winding increased area coefficient becomes 2. The generators presented in the table III
are closed to this situation. Considering the required stator slots area the active part of
the stator slots will be increased; that means an increase of iron weight and of
generator outer diameter. Increasing the stator iron weight will also increase the iron
losses with the result that, despite the generator weight increase, the efficiency will
decrease by 0.5% -up to around 1% [45].
Table V Efficiency decreasing when the generator dimensions and weight are preserved and only air-gap
magnetization power is supplied through the excitation winding.
Poles 8 12 16 18 20 24
dPcus [W] 210 222.9 217.2 223.8 249.5 245
Generator efficiency [%] 80 79.8 79.81 79.5 78.7 78.6
Efficiency reduction [%] 4.76 5 5.14 5.25 5.51 5.39
Table VI Efficiency decreasing when the generator dimensions and weight is preserved, but all reactive
power is supplied through the excitation winding.
Poles 8 12 16 18 20 24
dPcus [W] 240.8 242.76 246.11 249.27 264.3 259.14
Generator efficiency [%] 79.41 79.44 79.41 79.21 78.37 78.32
Efficiency reduction [%] 5.41 5.46 5.53 5.58 5.81 5.68
The impact of double stator winding on the generator parameters is presented in
table V and table VI where the machine dimensions are not changed but where the stator
copper losses are increasing. Table VII and table VIII show the geometry and costs of
DSWIG when the copper losses are preserved but the generator stator radial dimensions are
increased.
Table VII Dimensions, weight and initial cost variation for the same stator copper losses, only air-gap
magnetization power is supplied through the excitation winding.
Poles 8 12 16 18 20 24
dWcu [kg] 12.86 11.7 11.76 11.71 11.45 11.27
dWfe [kg] 9.87 9 7.98 7.61 6.95 6.21
dWeigh [kg] 22.72 20.7 19.74 19.33 18.4 17.47
dDso [mm] 33.8 37.3 41.3 40.7 39.7 33.4
dPfe [W] 18.45 29.89 38.24 42.3 43.18 47.7
Generator efficiency [%] 84.39 84.19 84 83.79 83.17 82.9
Efficiency reduction [%] 0.44 0.7 0.91 1 1 1.11
Initial cost increases [USD] 142.38 129.57 128.75 127.8 124.2 121.36
Table VIII Dimensions, weight and initial cost variation for the same stator copper losses, but all reactive
power is supplied through the excitation winding.
Poles 8 12 16 18 20 24
dWcu [kg] 14.73 12.74 12.74 12.49 12.13 11.92
dWfe [kg] 11.2 9.75 8.59 8.08 7.34 6.55
dWeigh [kg] 25.93 22.49 21.33 20.57 19.46 18.47
dDso [mm] 38.3 40.3 44.5 43.3 41.9 38.3
dPfe [W] 20.94 32.35 41.19 44.91 45.56 50.29
Generator efficiency [%] 84.33 84.13 83.97 83.72 83.11 84.84
Efficiency reduction [%] 0.5 0.77 0.98 1.06 1.06 1.17
Initial cost increases [USD] 163.05 141.06 139.38 136.22 131.53 128.32
The dual stator winding induction generator could reduce the inverter (active
rectifier) KVA with 43% for 8 poles, when the inverter is placed on the excitation
winding. The full power rectifier and the dc-dc boost converter required in this case
drastically reduce the advantages of this scheme. The converter KVA reductions are
smaller (19%) when the inverter is placed on the main windings. In this case the
converter power reduction is higher for a large number of poles where the converter for
standard induction machine is also larger due to smaller power factor. The method does
not require other power electronics but it requires a capacitor battery on the excitation
windings. If the DSWIG is made in the same size as the single winding induction
machine, the efficiency will be decreased by around 5%, and that creates cooling
problems. The efficiency could be preserved (only 0.5%-1% efficiency reduction) at
the cost of increasing the DSWIG size to 10% in diameter which would lead to 29% in
the generator weight and about 32%-35% in the active material costs. Economic
benefits will be obtained when the full power inverter price is larger than the generator
price by more than 50%, which is the case for small power machines.
Chapter 6
PARAMETER OPTIMAL IDENTIFICATION
The equivalent phase circuit for DSWIG is presented in figure 23 [83].
R1 L1
L12
R’2 L’2
L’r
Lm Rm R 'r
V1 V '2
s
Fig. 23 Equivalent phase circuit of the DSWIG [83].
The no load and short circuit tests are performed on the two windings
induction machine in order to identify the circuit parameters as in [83]:
a) no load test with the main winding (noted with 1) supplied, the auxiliary
winding (noted with 2) is open;
b) no load test with the auxiliary winding supplied, main winding is open;
c) two-phase short circuit test supplying the main winding with the auxiliary
winding open;
d) two-phase short circuit test supplying the main winding with the auxiliary
winding in short circuit;
e) two-phase short circuit test supplying the auxiliary with the main winding
open;
f) two-phase short circuit test supplying the auxiliary winding with the main
winding in short circuit;
For each of these tests the balance of the active and reactive powers may be
considered, which results in two equations for each test. Additionally, at the open
circuit tests the voltages on the unsupplied winding are measured as well, and two
more equations appear. Resistances R 1 and R 2 may be measured in dc. The iron losses
'
and the mechanical losses (in relation with R m and R r ) which are calculated through a
s
succession of open circuit tests and further equations, are also available. Finally, the
equation system is over determined, but all the measurements are affected by errors.
The relations between the machine parameters and the measured variables are not
linear due to the products and then the equations system is practically impossible to
solve classically even using non linear methods.
A. The Circuit Model
An acceptable solution for the over determined system of equations, knowing

that the measurements are affected by errors, is the minimization of the sum of the
error squares for the active and reactive power, which equals the computation of the
circuit parameters with an optimal problem. In conclusion, for the identification of the
circuit parameters optimization algorithms may be used, having the circuit parameters
as optimization variables, and having the sum of the error squares and the computed
powers (from the equivalent scheme and the candidate parameters) as the objective
function. The domain of search for the parameters may be established considering a
simplified circuit model (neglecting the influence of the rotor at the open circuit tests
and neglecting L m and R m at the short circuit tests) and the maximal influence of the
measurement errors. For tests a, b, c and e the system of equations which is solved is
[83]:
 R  jX sσ  X12σ   Z m Zm  I
 Vs   s  s 
    Rr '   , (68)
 0   Zm  jX r  Z m  I r 
'
 s 
where S is 1 for a and c and 2 for b and e. The slip is 1 for the short circuit test and
should be determined by the torque equation with no load test.
In the parameter optimal identification approaches, the parameters are assumed
to be known (given initial values or randomly generated in the finding interval) and the
stator and rotor currents are computed. Using the optimization algorithm the parameter
is changed in order to minimize the sum of the square error of currents, active and
reactive power. The mechanical loss is also a parameter that should be determined and
consequently it is assumed to be known. Moreover the mechanical losses variation
with the speed could be determined if the no load test is also performed at small
voltage close to the stability limit. The no load tests start with the larger voltage that is
around 60% (50%) of the rated voltage in order to consider the constant non saturate
link inductance in the equation during parameter identification. The link inductance
variation due to the saturation is determined separately considering the maximum
supplied voltage larger than rated voltage.
The mechanical losses are considered through friction torque which has three
components: the constant component or Columbian friction component T 0 , the viscous
friction component, k 1 Ω, and the fan component, k 2 Ω2.
2
ω ω 
Tmec  T0  k 1  k 2   T0  k 1 1 1  s   k 2  1  1  s 2 ,
2
(69)
p  p 
where ω 1 is the angular frequency, p – number of pair’s poles and s -the rotor slip. The
mechanical losses torque is equal to the electromagnetic torque:
40 Parameter optimal identification – 6
R2
Vs2
3p s
Tmec  2
, (70)
ω1 R2
R s  C1  jX sσ  X12σ  C1X 2σ 
s
where C 1 is considered to be magnetization current component and it is a complex
number, usually close to unity with a small imaginary part that is not usually
considered. In order to secure the algorithm accuracy for a large variety of induction
machines, considering that the renewable energy application at very low speed implies
large number of poles (ratio between leakage inductances and link inductance could be
larger that for usual machines) we decide to consider the complex value of C 1 .
 1 j 
C1  1  R s  jX sσ  X12σ      . (71)
 R m Xm 
The equation (70), considering the torque from equation (69) is a polynomial
equation of degree 4 which is solved numerically. The following rows will demonstrate
how to generate simply the polynomial coefficients. At first the equation that should be
solved is:
cts
Ts    Ts Pd s   c t s  Ps   c t s . (72)
Pd s 
In the computer code, the polynomial coefficients are stored in a vector and in
order to find the equations roots, the ‘P’ vector should be computed. The ‘T’ vector
coefficient in the descending degree of ‘s’ is coming directly from (69):

T  k 2 12 ,  k 11  2k 2 12 , T0  k 11  2k 2 12 ,  (73)
where
ω1
1  , (74)
p
In order to compute the denominator polynomial coefficient P d , we introduce
the P 1 , respectively P 2 polynomials whose coefficients are linked to the real,
respectively imaginary part of the denominator from (70):
P1  R s  imagC1  X rσ , realC1  R r  , (75)
P2  X sσ  X12σ  realC1  X rσ , imagC1  R r  , (76)
Pd  P1  P1  P2  P2 , (77)
P  T  Pd , (78)
where * means vector convolution product. The right side coefficient c t from (72) is:
3 2
ct  Vs R r , (79)
1
The four degree polynomial equation (72) has four solutions. The best solution
is only the smallest positive pure real solution. No real solution of equation (72) means
that an induction machine with the candidate parameters does not have an equilibrium
point at the given voltage. New candidate parameters should by tried. For the tests d
and f we have the following equations [83]:
 
  I
 V 1   R 1  jX 1σ  X 12σ   Z m jX 12σ  Z m Zm  1 
 ' 
V2    jX 12σ  Z m ' '

R 2  j X 2σ  X12σ  Z m  Zm  I '2  ,
 ' 
 0   R 'r I
   Zm Zm  jX '2σ  Z m  r 

 s 
(80)
V  for test d
V1   s , (81)
  for test f
0
0  for test d
V '2  V  for test f , (82)
 s
The candidate parameter vector is:
X = [k e R 1 R 2 R r R m X σ X 2σ X 12σ X rσ X m T 0 k 1 k 2 ] . (83)

The estimated stator current I s is computed at the test voltage using the
candidate parameters and equivalent scheme according to performed test (a, b, c, d, e,
f). Active and reactive power are computed for the no load test considering three phase
current and for short circuit 2 phase current:
 

P( a,b)  3V( a,b) real I ( a,b) ,  (84)
 
Q( a,b)  3V( a ,b ) imag I ( a,b) ,   (85)
 

P(c,d ,e, f )  2V(c,d ,e, f ) real I (c,d ,e, f ) ,  (86)
 

Q(c,d ,e, f )  2V(c,d ,e, f ) imag I (c,d ,e, f ) .  (87)
B. The Measured Errors Reduction
The algorithm itself filterers the measured error, but some errors are very large
and then it is better to eliminate those points. Measuring the voltage, current, active
and reactive power gives the possibility of estimating some error level in the
measurements. The relative error for no load test and short circuit test are [83]:
Ps2  Q s2
ε m ( a ,b )  1, (88)
3Vs I s
Ps2  Q s2
ε m ( c ,d ,e ,f )  1 . (89)
2Vs I s
The average error ε av and standard deviation of the error σ ε could be computed
for each test series. The inconsistent measurement points are eliminated from the
parameter computation and also from graphic presentation. A measurement is
considered consistent if the relative error in the measured power satisfies the following
inequality:
ε m  ε av  σ ε , (90)
A thrust weighing factor, w, could be computed for each point based on the
average error, standard deviation of the series test and individual error:
ε2
 2
ε av  ε 2m  σ ε2
we . (91)
C. The Boundaries of the Candidate Parameters Vector
The boundaries of the candidate parameter vector are computed from

experimental data using classical approximation of the induction motor at no load test,
respectively short-circuit and considering the maximum influence of the neglected part
and also the measured accuracy. The auxiliary winding is considered in the stator
frame of the main winding [83]:
V '2  k e  V 2 , (92)
where the ratio factor k e may be considered as the one used in designing, or may be
calculated experimentally, as being part of the unknown parameters of the machine,
and:
V N kw
ke  1  1 1 , (93)
V2 N 2 kw 2
is the value of design. If k e is part of the unknown series, the computation of the search
domain is needed:
 V 
ke max  max  1  1  2 v  , (94)
 V20 
where V 1 is the supply voltage of the main winding, V 20 is the voltage at the terminals of
the auxiliary winding measured:
V 
ke min   10  1  2 v  , (95)
 V2 
and V 10 is the voltage at the terminals of the main winding when the auxiliary winding
is supplied and  v is the maximum value added in both voltmeters (a couple of
percentage points). If the algorithm of identification requires an initial value it may be
considered that:
1  V1 V10 
k e0    . (96)
2  V 20 V 2 

The maximum and minimum values of the stator winding resistances R 1 , and
R 2 are computed considering their dc measured values and dc resistance measurement
including possible resistance variation due to the windings temperature variation:
Rs max  Rsdc 1   Rs  ; Rs min  Rsdc 1   Rs  . (97)
The rotor resistance boundaries are computing considering the short circuit
resistances and the boundaries of the stator resistances:
R 'r max  max R 1sc1   R 1 min ; R 'r min  min R 1sc1   R 1 max . (98)
The minim rotor resistance should by positive, so the computed values will set
to zero if the large measurement error produce a negative value. The short circuit
resistance values are computed from the short circuit tests series c):
Pmsc1
R1sc1  . (99)
2
2 I msc1
The iron losses resistance boundaries:
1  2 v
Rm max  , (100)
P 
min  m0  R1min I m2 0 
1  3 

max  R10  max Vm0 
2
Rm min  max  R10 1  2 v   R1max , (101)

where R 10 is the no load resistance computed from ‘test a)’ series:
3Vm20
R10  . (102)
Pm0
The stator winding leakage reactance’s boundaries are computed using the
short circuit reactances:
X1max
X1 max  min  X1sc 2  ; X1 min  , (103)
10
X
X 2 max  min  X 2 sc 2  ; X 2 min  2 max , (104)
10
X12 max  max  X1sc1   min  X1sc 2  ; X12 min  0 . (105)
The rotor leakage boundaries are the same as the stator coupling leakage
reactance X 12σ. The magnetization reactance boundaries are computed from the no load
tests on the main winding:
X m max  max  X10 1 2v  ; X m min  max  X10   X max . (106)
The boundaries of the mechanical losses components are:
T0 max 

min Pm0  3R1min I m2 0  ; T0 min  0 , (107)
1
T
k1max  0 max ; k1min  0 , (108)
1
k0 max
k2 max  ; k2 min  0 . (109)
1
D. The Objective Function
The objective function is the sum of the squared current, active and reactive
power error for each series tests plus the squared error on the opening winding for the
no load tests. For each of the six tests the active and reactive power errors are
calculated [83]:
 t   a2   b2   c2   d2   e2   2f   va
2 2 ,
vb (110)
 x2   xIav
2  2 wxxI2  wxxP2  wxxQ2 , (111)
xPav   xQav 
2
wx wx wx
where indices x means test case a, b, c, d, e, and f. The current, active and reactive
power errors ε xI , ε xP , ε xQ are computed for each trust point:
  
Ix Px Qx
 xI  1  ;  xP  1 ;  xQ  1  , (112)
Ix Px Qx
The voltage ratio error:
 va
2   wa  vai
2
;  vb
2  
wb vbi
2
, (113)
 
wa


wb
V21i V12i
 vai  1  ;  vbi  1  . (114)
V21i V12i
Chapter 7
2D-FEM ANALYSIS
An approach that includes the influence of magnetic saturation and iron loss
using finite-element analysis in the performance prediction of the dual stator winding
induction machine was set. The proposed finite-element model provides very good
steady-state predictions and can be used for the sizing and design optimization of the
machine.
Induction machine saturation could be increased at low frequency (required in
direct driving) without notable iron losses increases in order to have a larger slot area
required in two stator winding or for aluminum winding which could replace the
copper winding in low cost machines.
The program starts in Matlab and creates the mesh and run in the FEMM 4.2
software platform (figures 24-36). When applied to electrical machines, the problem is
usually reduced to cover only one pole or one pole pair with the help of boundary and
symmetry conditions, in order to reduce the computation time. In addition, it is often
sufficient to reduce the problem to a 2D plane and use the finite element analysis
(FEA). The mesh has 47758 nodes and 95129 elements [39].
Fig. 24 Achieving geometric constructions (4 poles half machine) through the software FEMM 4.2 with
starts from Matlab.
46 2D FEM Analysis – 7
Fig. 25 Magnetic field lines resulting from finite element analysis.
Fig. 26 The location of the two-phase winding conductors of the stator and rotor cage slots and allocation
of material properties.
Fig. 27 The shape of stator and rotor slots in the air gap.
Fig. 28 Assigning properties of ferromagnetic core material for the stator and rotor.
Fig. 29 Portion discredited and field lines in the area from the air gap.
Fig. 30 Portion discredited and field line in the stator tooth and rotor tooth passage from the air gap.
Fig. 31 Magnetic field lines and meshing elements resulting from finite element analysis portion studied
machine.
Fig. 32 Highlighting the field lines in the teeth and the rotor yoke.
Fig. 33 Highlighting the field lines in teeth and the stator yoke.
Fig. 34 Highlighting the three-phase main stator winding in two layers.

Fig. 35 Highlighting of the three-phase excitation stator winding in one layer.
Fig. 36 Highlighting cage rotor winding.
No load (zero rotor current) FEM investigations at standstill, with dc currents

for star connection lead to I Ae = I= -2I Be = -2I Ce . The total distribution of the flux
density for 4 poles is shown in figure 37 for excitation current and in figure 38 for
main winding current [39].
Fig. 37 Total flux dens. at I Ae =4.5A and I Am =0 [39]. Fig. 38 Total flux dens. at I Am =4.5A and I Ae =0
[39].
In figures 39 and 40, detailed distribution of flux density in the rotor and stator
teeth, and in the rotor yoke and in one rotor tooth and two stator teeth is shown.
Fig. 39 Detailed distribution of flux density in the rotor and stator teeth, and in the rotor yoke.
Fig. 40 Detailed distribution of flux density in one rotor tooth and two stator teeth.
The analytical model should be improved for a high saturation magnetic core.
Increasing the stator slot by 40% will increase the saturation of the stator core (figure
41) and will change the line field distribution. The leakage inductances computed
analytically could be validated through a virtual short circuit using ac finite element
analysis where the stator winding currents are given (rated or higher values) at rated
frequency and the rotor bar currents are computed as eddy currents (figure 42). The
end coil leakage inductances are not considered in the 2D FEM approach. No load
(zero rotor current) FEM investigations at standstill, with dc currents for star
connection lead to I Ae = I= -2I Be = -2I Ce [39], [85].
a) b)
Fig. 41 The magnetic field distribution: a) original 6 kVA, 8 poles, 15 Hz rated frequency induction
generator, b) the same generator with 40% larger slot [85].
ωt = 0 ωt = 900
Fig. 42 Virtual short circuit test using AC finite element analyses (f = f N = 15Hz, I mA = 7A peak
current in main winding, I mB =I mC =-3.5A) [85].
Figure 43 shows the air-gap flux density (produced by main and excitation
currents) and figure 44 shows the normal and tangential air-gap magnetic field
strength.
a) b)
c)
Fig. 43 a) Normal air-gap flux density [85]; b) air-gap flux density magnitude; c) air-gap tangential flux
density.
a) b)
c)
Fig. 44 a) Normal air-gap magnetic field strength [85]; b) tangential air-gap magnetic field strength; c)
the magnitude of the magnetic field strength in the air gap.
Figure 45 shows the air-gap flux density (produced by main and excitation
currents) and figure 46 shows the flux density space harmonics. The linkage fluxes (A e
produced by excitation current in excitation winding and A m produced by excitation
current in main winding) are shown in figure 47. Figure 48 are shows inductances
versus current (L e1 excitation phase inductance, L m1 main phase inductance, L ed
excitation d axes inductance, L mq main q axes, L em , L me coupling inductances between
excitation and main winding computed in two ways: linkage flux produced by
excitation current in main winding divided by excitation current respectively linkage
flux produced by main current in excitation winding divided by main current). The
difference between L em and L me are smaller than 0.3% and they should be showing in
FEM accuracy [45], [80].
Air - gap flux density Flux density - space harmonics
0. 6
0. 8
Ie
0. 5
Im
0. 4
Flux density (T)
0. 4
0
0. 3
-0.4 0. 2
Im 0. 1
-0.8 Ie
-1
0
0 200 400 600 800 0 2 4 6 8 10 12 14 16 18 20
Angle (deg) Harmonic order
Fig. 45 The normal component of air-gap flux density [39]. Fig. 46 The flux density space harmonics
at I Am = 4.5A respectively I Ae =4.5A [39].
Linkage flux versus current

2.4 Inductances versus current
Ae 0.41
Led
Le1
Am Lem Lmq
2 0.4
Lme
FLux (Wb)
1.6 0.39
Lm1
0.38
1.2
0.37
0.8
0.36
0.4 0.35
1 2 3 4 5 6 1 2 3 4 5 6
Peak current (A) Peak current (A)
Fig. 47 The linkage fluxes versus current [39]. Fig. 48 The inductances versus current [39].
Chapter 8
CONTROL STRATEGIES, THE DSWIG MATHEMATICAL MODEL AND
DYNAMIC SIMULATIONS
8.1 Mathematical model of the control
The complex variable (space-phasor) model of IM is [28]:

dψ
V s  R s i s  s  jω b  ψ s , (115)
dt
ψs  L s i s  L m  i r , (116)
dψr
Vr  R r ir   jωb  ωr   ψ r , (117)
dt
ψr  L r i r  L m  is , (118)
Te 
3
2
  3
 
p1R e jψs  i*s   p1R e jψ r  i*r ,
2
(119)
J dωr
  Te  Tload , (120)
p1 dt
dθ er
 ωr . (121)
dt
The complex variables may be decomposed in plane along two orthogonal d
and q axes rotating at speed b to obtain the d-q (Park) model:
V s  Vd  jVq , (122)
is  i d  ji q , (123)
ψ s  ψ d  jψ q , (124)
V r  Vdr  jVqr , (125)
i r  i dr  ji qr , (126)
ψ r  ψdr  jψqr . (127)
The voltage equations become:
dψ d
 Vd  R s  i d  ωb ψ q , (128)
dt
dψ q
 Vq  R s  i q  ωb ψ d , (129)
dt
58 Control strategies, DSWIG mathematical model and dynamic simulations – 8
dψdr
 Vdr  R r  i dr  ωb  ωr ψ qr , (130)
dt
dψ qr
 Vqr  R r  i qr  ωb  ωr ψ dr . (131)
dt
From these equations and from phase diagram (figure 49) we have:
2  2   2  
 ed   Ae  cos   Be  cos      Ce  cos    , (132)
3  3   3 
2  2   2  
 eq    Ae  sin    Be  sin       Ce  sin     , (133)
3   3   3 
2  2   2  
 md   Am  sin     Bm  sin       Cm  sin     , (134)
3  3   3 
2  2   2  
 mq   Am  cos   Bm  cos      Cm  cos    . (135)
3   3   3 
Am
Be 
e
q
d
m
Q 
D
1200
1200
m e  Ae
Bm 120 0
1200
Cm
Ce
Fig. 49 Main, excitation, d-q and - winding diagram [39].
Analog results the equations for currents (I ed , I eq , I md , I mq ) and voltages (U ed ,

U eq , U md , U mq ). For  = 0 the ,  model is equal with d, q model [39], [80].
 2 1 1 
ψ eα  3  ψ Ae  2 ψ Be  2 ψ Ce 
  
   , (136)
2 3 3
ψ eβ   ψ  ψ 
3  2
Be Ce 
 2 
 2 3 3 
ψ mα    ψ Bm  ψ Cm 
 3 2 2  .
 (137)
2 
ψ   ψ  ψ  ψ  1 1 
 mβ 3  Am 2 Bm 2 Cm 
Magnetic core saturation is very important for an induction generator with
capacitor excitation because otherwise the system is unstable. Further, deep saturation
occurs when a variable speed induction generator with capacitor reactive power
compensation is used so the induction generator mathematical model should consider
the saturation through an analytical approximation. An analytical continuous and
differentiable function, L M (i 0 ), is a good approximation for magnetization inductions,
for a large current range, only if the following affirmations are true: the function is
positive, the magnetic flux is finite (an additional constant term could consider slowly
flux increase at very large current), the transient induction L Mt is positive, and the
functions L M and L Mt could have a single maximum when the current is increasing
from zero to infinity [39], [80].
  
L M  f i md  i ed  i D 2  i mq  i eq  i Q 2 , (138)
ψmd LM Lmeσ Lmσ LM Lmeσ LM  imd
    
 ψed    LM Lmeσ LM Lmeσ  Leσ LM  ied  , (139)
ψ   LM LM LM Lrσ   iD 
 D 
ψmq LM  Lmeσ  Lmσ LM  Lmeσ LM  imq
    
 ψeq    LM  Lmeσ LM  Lmeσ  Leσ LM  ieq  , (140)
    
 ψQ   LM LM LM  Lrσ   iQ 
 m  L M  is  i r    R cos , (141)
i 0  i s  i r , (142)
 
ψ mβ  L M  isβ  i rβ  ψ R sin θ  , (143)
i 0β  isβ  i rβ , (144)

θ  a tan i 0α , i 0β  , (145)
i 0α
cos θ  , (146)
i 0α  i 02β
2
i 0β
sin θ  , (147)
i 02α  i 02β
dθ
ωb  . (148)
dt
From dependence:
L M  f i 0   0 , (149)
where i 0 is the magnetisation current and LM i0   i0 is bounded i0  0 :
i0  i md  ied  i D 2  i mq  i eq  iQ 2 , (150)
the transitory magnetisation inductance is:
L
L Mt  M  i 0  L M  0 . (151)
i 0
With notation:
 L Mt  L meσ  L mσ L Mt  L meσ L Mt 
 
A   L Mt  L meσ L Mt  L meσ  L eσ L Mt  , (152)
 L Mt L Mt L Mt  L rσ 

 L me σ  L m σ L me σ 0 
 
A  L Mt  I 3   L me σ L me σ  L e σ 0 , (153)
 0 0 L r σ 

A  L Mt  I3  L σ , (154)
 di md 
 
 dt   U md  R m  i md  ω b ψ mq 
di 
A   ed U  R e  i ed  ω b ψ eq   C. (155)
 dt   ed 
 di D   0  R D  i D  ω b  ω r ψ Q 
 
 dt 
 dimq 
 
 dt   Umq  R m  i mq  ωb ψmd 
 dieq   
A    Ueq  R e  i eq  ωbψed   D , (156)
 didt   0  R  i  ω  ω ψ 
r D
 Q   Q Q b
 dt 
 
 i md 
d  
  A C , (157)
1
 i ed
dt  
 iD 
 i mq 
d 
 i eq   A 1  D . (158)
dt  
 iQ 
If L M has the form:
1  ai 0n 1
LM  LM0 . (159)
1  b1  i 0n 1  b 2  i 0n
where L M0 is the initial value of magnetisation inductance and a, b 1 and b 2 are
constants. The coefficients of the function (159) are strictly positive and they could be
chosen to satisfy the condition (149) and also to approximate the experimental (or
design computed) inductance given in several points. The polynomial degree n, is
chosen as small as possible and still has enough function curvature variation.
1
L M0 – is the inductance around zero current, the bending current point,
nb
2
a
the LMo ration represent the saturated flux (maximum value) and finally b 1 controls
b2
the maximum of the inductance curve (for larger values of the b1 there is no
maximum) and guards positive transient inductance.
A good approximation of magnetization inductance is shown in fig.50 where L M
is an analytical approximation, L M FEM are points from design using finite element
computation L Mt analytical approximation of transient inductance and L Mtd is transient
inductance computed by digital derivative from L M FEM [39], [80].
Inductance LM (H)
Fig. 50 Comparison results from analytical and FEM calculation of magnetization inductances [39].
The L M , function (159) coefficients are: n=3, L M0 =0.392H, a=0.03 A-1/2,

b 1 =0.0137 A-1/2, b 2 =0.0036 A-1/3. The fluxes in rotor  Q and  D are:
 i mq 
 
φ Q  L M L M L M  L rσ    i eq  , (160)
 
 iQ 
 i md 
 
φ D  L M L M L M  L rσ    i ed  . (161)
i 
 D 
8.2 The principle of DSWIG control
The principle of DSWIG control is shown in fig. 51 [39], [80], [81].

P
X
IDC VDC
Imq 
Inverter Control
Reg
+
Imd
0
C
1 vm
-K Im
s
ve Ie
DSWIG
Model of T Tem
+ 1 r
wind
vwind  J m  J t   s
turbine
P*
Fig. 51 Principle of DSWIG control [39], [80], [81].
Figure 52 shows the Simulink diagram of the DSWIG with saturated model
and remanence model implemented [81].
1 K*u mu
K*u 1
Vm mu
Gain 3 Im
Gain 5
0 -K -
K*u 2
Constant 1 Gain mu Ie
Gain 6
Matrix 1 2
Multiply |u| Inductance
s
Product Integrator Math
[0 0]
2 K*u m Function
Ve Constant 4 P(u)
Gain 4
O(P) = 5 Inv
Polynomial 1
Product 7 Product 2
Dot Product 1 P(u) 2
Product 3 sqrt |u| Lsig
O(P) = 3
Math Polynomial Math Constant
-K - Function 2 Function 3
0
Gain 1 P(u)
Constant 2 Matrix 1 O(P) = 2
Multiply |u| 2 Product 8
s Polynomial 2
-C-
Product 1 Integrator 1 Math
[0 0] Function 1 Constant 3
Constant 5
3 pp
w
Gain 2
Product 6
Dot Product
Product 4
-K- 3
Tem
Gain 7
Product 5
Fig. 52 Mathematical model of DSWIG-block diagram [81].
A preliminary feed-forward control shown in fig. 53 was implemented to

verify the dynamic of the proposed topology. The auxiliary winding voltage becomes
larger than dc voltage, after 80% of the rated speed and it starts to deliver the power to
the dc link. The power in the main winding is maintained constant by controlling the
main voltage V 1 as it is shown in fig. 53 b [82].
a)
b)
Fig. 53 Feed-forward control a) block diagram, b) nonlinear V 1 and f 2 function [82].
Numerical simulations of the dynamic process are used to prove the proposed
control strategy. The matrix form equations (162-164) in a stator reference frame are
used to model the DSWIG in transient regime. The main inductance’s L m saturation is
considered through an analytical approximation (165-166) where L m0 , a, b 1 , b 2
parameters are given from a curve fitting process considering the magnetization
inductances values from FEM.
 
 I α1   Vα1  R 1I α1 
 '   
 I α 2   L1  Vα' 2  R 2 I 'α 2  , (162)
 '   ' ' 

 I αr 
    R r I αr  ω r βr 
 
 
 Iβ1 
 '   Vβ1  R1Iβ1 
 
 Iβ 2   L1  V '  R I '  , (163)
β2 2 β2
 '   ' 
 ' 
   ω r αr  R r Iβr 
 Iβr 
 
 Lm  L1σ  L12σ Lm  L12σ Lm 
 
L   Lm  L12σ Lm  L12σ  L2σ Lm  , (164)
 Lm Lm 
Lm  Lrσ 

1  aI 02
L m  L m0  , (165)
1  b1I 02  b 2 I 30
I0  I α1  I 'α 2  I α' r   I
2
β1  Iβ' 2  Iβ' r 2
, (166)
 α1   I α1 
   
    L   I 'α 2  ,
'
(167)
 α' 2   ' 
   I αr 
 αr 
   Iβ1 
 β1   
    L   I'  ,
'
(168)
 β2  
β2

 '   Iβ' r 
 βr   
3

Tem  P1 β' r  I 'αr  α' r  Iβ' r .
2
 (169)
The DSWIG model, based on the previous equations is shown in figure 54 [82].
V1abc main V1, V1 I1, I1 main I1abc
abc DSWIG abc
ortogonal
V2abc auxiliary V2, V2 nonlinear I2, I2 auxiliary I2abc
abc model abc
r Stator coordinate Tem
a)
I1, I2, Ir
V1, V2  axis 
I
I 1
r Eq.
s
(17) Saturation
r
model
Lm Lm
L-1 Eq. I0
(21)
r I0
r  Eq. (20)
 axis I 1
V1, V2 Eq.
s
(18) I
I1, I2, Ir
1, 2, r

I1, I2, Ir
Eq. (22)
L matrix
Lm construction Matrix L-1
Eq. (19) Inversation
1, 2, r
I1, I2, Ir Eq. (23)
Ir, Ir, r, r 3 Tem

2

R1  ' r I ' r   ' r I ' r 
b)
Fig. 54 The DSWIG model a) orthogonal transformation; b) orthogonal nonlinear model [82].
A mathematical model of the DSWIG and control is combined with the circuit
model in figure 55 [82].
Matematical model of
the DSWIG and control I1
Filter and rectifier circuit
Feed- V model
1 V2a
forward
control I2a
V2b
 VDC
DSWIG I2 V2
model I2b
V2c
V2 I2c
C
Tem
Fig. 55 System simulation-block diagram [82].
The feed-forward control is set to extract electrical power in direct ratio to the
cube of the shaft speed. Only the active rectifier is able to extract electrical power at
low speed, figure 56.a. The iron losses and mechanical loses are not considered in the
dynamic simulations and consequently the efficiency, figure 56.b, contains only the
copper losses.
Efficiency
2000 0.86
Paux
0 0.84
Pmain
-2000 0.82
Power (W)
-4000 0.8
Pin
-6000 0.78
-8000 0.76
1 2 3 4 5 6 7 1 2 3 4 5 6 7
time (s) time (s)
a) b)
Fig. 56 Power (a) and efficiency (b) variation [82].
The prime mover is assumed to rotate the generator with a variable speed and a
test variation of the speed is considered in the simulation, figure 57. The voltages and
currents variation considering the speed test’s profile are shown in figure 58 [80], [82].
Shaft speed
50
45
Speed (rad/s)
40
35
30
25
20
1 2 3 time (s) 4 5 6 7
Fig. 57 Shaft speed profile [82].

Main phase voltage
400 Main phase current
10
300
200 5
100
Voltage (V)
Current (A)
0 0
-100
-200 -5
-300
-400 -10
1 2 3 4 5 6 7 1 2 3 4 5 6 7
time (s) time (s)
Auxiliary phase voltage Auxiliary phase current

400
10
300
200
5
100
Voltage (V)
0 Current (A) 0
-100
-5
-200
-300
-10
-400
1 2 3 4 5 6 7 1 2 3 4 5 6 7
time (s) time (s)
Fig. 58 Main and auxiliary voltages and currents variation [82].
Only the fundamental harmonics are considered in the voltage applied to the
main windings. When the speed is low and the diode rectifier is closed all the voltages
and the currents are sinusoidal. When the diode rectifier is open the harmonics are
spread in auxiliary current and voltage, but also in the main current as shown in fig. 59
at the rated power in the auxiliary winding.
Main phase current Auxiliary phase voltage
10 a 400
b c
a b c
300
5 200
100
Current (A)
Voltage (V)
0 0
-100
-5 -200
-300
-10 -400
5.5 5.51 5.52 5.53 5.54 5.55 5.56 5.57 5.58 5.59 5.6 5.5 5.51 5.52 5.53 5.54 5.55 5.56 5.57 5.58 5.59 5.6
time (s) time (s)
Auxiliary phase current Rectifier current

a b c
10 7.4
7.2
5
7
Current (A)
Current (A)
6.8
0
6.6
6.4
-5
6.2
6
-10
5.8
5.5 5.51 5.52 5.53 5.54 5.55 5.56 5.57 5.58 5.59 5.6 5.5 5.51 5.52 5.53 5.54 5.55 5.56 5.57 5.58 5.59 5.6
time (s) time (s)
Fig. 59 Currents (main, auxiliary, rectifier) and auxiliary voltage waveform (zoom in) for rated power in
the auxiliary windings [82].
8.3 Control strategy for self excitation case
The control strategy for self excitation case is illustrated in fig. 60 via a
dedicated Matlab Simulink code [39], [81].
Dot Product
Voltage Power
-K - Vm Im
R_load
Current
1
-K - Ve Ie
s
C Integrator
26 w Tem
Speed
Dot Product 2
IG
a)
Rload Cexcitation
im Vm ie Ve
Main Excitation
winding winding
Tshaft
Input
speed

Short circuit
rotor winding
b)
Fig. 60 a) Simulink diagram of the DSWIG drive, self excitation case; b) real block diagram [39], [81].
The nonlinear model of DSWIG [39] is used to study the dual winding
induction generator with an active rectifier on the main windings and a passive bank
capacitor on the excitation side. The self excitation ability is essential in autonomous
applications. Moreover as a backup for power electronics failure, the unpretentious
load (heaters, light bulbs) could be connected directly to the main winding
transforming the scheme from figure 2d in an emergency solution, figure 61 a. A
Matlab Simulink model to study self excitation of DSWIG was developed, figure 61 c,
considering the wind turbine model, figure 61 b and saturated DSWIG model (another
block diagram), in figure 62 [80].
Table IX Turbine parameters

Symbol Description Value
P N TURBINE rated turbine power 3.5 kW
VN wind nominal speed 8 m/s
nn nominal turbine speed 220 rpm
J turbine and generator inertia 80 kgm2
B viscous friction coefficient 0.01 Nm2/rad
C0 aerodynamic p.u. torque at zero speed 0.0068
C2 reversing value of 1/λ pu 0.5
a)
B
V 1 Vpu V2pu
u2
Vn
Tpu − 
1
1  Tn
+ Js
 * pu −
1 pu Cpu
/ Eq. (10) Tem
n
b)
c)
Fig. 61 a) emergency solution for unpretentious load; b) Turbine model;
c) Block diagram of the DSWIG, self excitation case [80].
Fig. 62 Another block diagram of the DSWIG with remanence and saturation [80].
The aerodynamic torque and speed is computed by a wind turbine model based
on [51-53] considering the per-unit tip-speed ratio λ pu [80]:
V ω
Vpu  , ωpu  , (170)
Vn ωn
1 Vpu
 , (171)
λ pu ωpu
C  1  C3
Cpu  C0  1   C 2 e λ , (172)
λ pu  λ pu 
 
Ta  Cpu  u 2  Tan , (173)

J  ω  Ta  Tem  B  ω , (174)
1
C3  , C1  C0  1  C3  e  C3 . (175)
C2  1
The minimum value of the capacitor which is able to start self excitation
depends on the speed and for no load, ideal (no loss) generator; it was computed with
(176) and shown in fig. 63 [80].
1 1
C min  2  . (176)
ω L M 0  L meσ  L eσ  p  L M 0  L meσ  L eσ 
2
220
200
180
160
140
Cmin (uF)
120
100
80
60
40
12 14 16 18 20 22 24 26 28 30
Mechanical speed (rad/s)
Fig. 63 The minimum value of the capacitor versus speed [39].
The capacitor is dimensioned to produce all reactive energy necessary at rate

load, so the active rectifier is designed only for rated active power. From the reactive
energy balance the capacitor values are C n =71.5 μF. The dynamic model which
includes the winding losses proves that self excitation starts slowly with a 58 μF
capacitor, a value that is notably larger than for the ideal generator (given in fig.63).
With the C n capacitor self excitation starts at 23.4 rad/s. The no-load self excitation
process at rated speed and C n capacitor is shown in fig. 64 (voltages, currents and
torque) and in fig. 65 (power, magnetization current and magnetization inductance). At
3.947 s, the generator is sudden charged with the rated load (load resistance is reduced
from 4840  to a 48.4 ) [80], [81].
Fig. 64 Main phase voltages, excitation phase voltages, main current, excitation current and torque, self
excitation case [39].
The dynamic simulation proves that the induction generator is able to start and
run with full load without any control.
Fig. 65 Output power, input power, magnetization current magnitude, magnetization inductance and
transient magnetization inductance, self excitation case [39].
The self-starting of the wind turbine is very slow and according to the digital
simulation it required 6 minutes to accelerate from 0 to the rated speed, even at 8m/s
wind speed (the turbine staring torque is only 0.68% of the rated torque). In figure 66
the block parameters of wind turbine and DSWIG from Simulink are shown. Figures
67 and 68 present turbine torque and speed starting from zero speed. The time
necessary for starting is about 6 minutes. In order to avoid a long time simulation, it
was assumed that initial speed of the wind turbine is 10 rad/s [80], [81], [86].
Fig. 66 The block parameters of wind turbine and DSWIG from Simulink.
Turbine output torque
200
150
Torque (Nm)
100
50
0
0 50 100 150 200 250 300 350 400 450 500
Time (s)
Fig. 67 Turbine torque considering starting from zero speed.

Turbine speed
35
30
25
Speed (rad/s)
20
15
10
0
0 50 100 150 200 250 300 350 400 450 500
Time (s)
Fig. 68 Turbine speed considering starting from zero.
The wind turbine speed, figure 69, is increasing under aerodynamics torque up
to 36rad/s, where the self excitation process brings the DSWIG at maximum power. A
sudden resistive torque, figure 70, is applied to the wind turbine through the generator
reducing the speed to the rated speed. The rated value capacitor, C n =80μF and rated
resistive load R n =62  are connected to the generator terminal during the entire
starting process. Overvoltage and over currents could be noticed on both windings after
the self excitation process, figure 71. Digital simulation, figure 72, proves that self
excitation DSWIG with a capacitor is able to produce the rated power at rated wind
speed [80].
Turbine speed Ratio between turbine speed and wind speed
40 1.6
35 1.4
30 1.2
Speed (rad/s)
(p.u.)
25 1
20 0.8
15 0.6
10 0.4
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time (s) Time (s)
a) b)
Fig. 69 a) Turbine speed during the autoexcitation proces [80]; b) ratio between turbine speed and wind
speed at high wind speed (8 m/s).
Torque (Nm)
Fig. 70 Turbine and generator torque during self excitation proces and nominal load
at high wind speed (8 m/s) [80].
Load voltage Excitation voltage
600 600
400 400
200 200
Voltage (V)
Voltage (V)
0 0
-200 -200
-400 -400
-600 -600
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time (s) Time (s)
a) b)
Load current
Excitation current
10
15
10
5
5
Current (A)
Current (A)
0 0
-5
-5
-10
-10 -15
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time (s) Time (s)
c) d)
Fig. 71 Self excitation and nominal load case: a) main phase voltage; b) excitation phase voltage; c)
main phase current; d) excitation phase current [80].
Fig. 72 Output and input power for self excitation and nominal load case at high wind speed (8 m/s) [80].
The induction generator self excitation occurs also at low wind speed (5.5m/s)
and it stay excited after that even at low wind speed (4.5m/s) with a small load (load
resistance 124). The aerodynamics and electromagnetic torques, figure 73, as well
the power variation in figure 74 show that the system is close to the stability limit. The
wind turbine speed, figure 75 is not adapted to the wind speed and consequently the
efficiency power conversion is small [80].
60
50
Turbine
40
30
Generator
20
10
0
0 20 40 60 80 100
Time (s)
Fig. 73 Turbine and generator torque during autoexcitation proces at low wind speed (wind speed 5.5 m/s
from 0 to 50 s; it the decreases to 4.5 m/s at 60 s when it remains constant) [80].
Output and input power

0
-200
Output
-400
-600
Input
-800
-1000
-1200
0 20 40 60 80 100
Time (s)
Fig. 74 Output and input power for self excitation case at low wind speed (5.5m/s  4.5 m/s) [80].
Ratio between turbine speed and wind speed
Turbine speed
1.8
28
1.6
26
1.4
Speed (rad/s)
24
(p.u.)
1.2
22
1
20
0.8
18
0 20 40 60 80 100 0 10 20 30 40 50 60 70 80 90
Time (s) Time (s)
a) b)
Fig. 75 a) Turbine speed during autoexcitation proces [80]; b) ratio between turbine speed and wind speed
at low wind speed (5.5m/s  4.5 m/s).
A larger capacitor (120 μF) could improve the system stability at low wind
speed and also increase the output power at low speed, figure 76, but it will increase
the generator losses at high speed [80].
Power (W)
Fig. 76 Output and input power for self excitation case at low wind speed (5.5m/s  4.5 m/s) and 120 μF
capacitor [80].
For a large capacitor (120 μF), with no load self excitation (R load =12400 Ω)
and high wind speed (8 m/s), an increase in voltage across main and excitation
windings nearly of 600V is observed. In figures (77-82) torque, voltages in main and
excitation windings, power, magnetization current and magnetization inductance in this
situation are shown [80].
180
160
140
Turbine
120
100
80
60
Generator
40
20
-20
0 5 10 15 20 25 30
Time (s)
Fig. 77 Turbine and generator torque during self excitation proces with large capacitor (120 μF) and no
load, at high wind speed (8 m/s).
Load voltage
600
400
200
Voltage (V)
-200
-400
-600
0 5 10 15 20 25 30
Time (s)
Fig. 78 Main winding voltages during self excitation proces with large capacitor (120 μF) and no load, at
high wind speed (8 m/s).
Excitation voltage
600
400
200
Voltage (V)
-200
-400
-600
0 5 10 15 20 25 30
Time (s)
Fig. 79 Excitation winding voltages during self excitation proces with large capacitor (120 μF) and no
load, at high wind speed (8 m/s).
Power (W)
Fig. 80 Ouput and imput power during self excitation proces with large capacitor (120 μF) and no load, at
Magnetization current (magnitude value)

25
20
15
Current (A)
10
0
0 5 10 15 20 25 30
Time (s)
Fig. 81 Magnetization current during self excitation proces with large capacitor (120 μF) and no load, at
Magnetization inductance)
0.3
0.25
Inductance (H)
0.2
0.15
0.1
0.05
0 5 10 15 20 25 30
Time (s)
Fig. 82 Magnetization inductance during self excitation proces with large capacitor (120 μF) and no load,
at high wind speed (8 m/s).
The capacitor bank must be designed in order to offer enough reactive power,
but at the same time not too much as to create a self-excited phenomenon. A capacitor
bank is used to compensate the reactive power. Equation (177) shows the capacity
computation considering that all reactive power is produced by the capacitor bank [86].
Q
Q  3  U e2  ω  C  C  (177)
3  U e2  ω
At 1 = 200 rad/s, n  477.7 rot / min, f  31.84 Hz  , the induction machine
required a reactive power Q = 720 VAR. On the another hand, the self-excited
phenomenon should be avoided which means that the next condition is accomplished:
1
XC  XL   ω1  L , (178)
ω1  C
1
C . (179)
L ω12
According to the Table II values, the capacity is computed to avoid the self-
excited phenomenon at ω 1 =200 rad/s.
L L  L 2σ  L12σ  L m  0.38 Hz . (180)
For X C = X L we obtain:
1
C 2  C  65.5 μF . (181)
ω1  L
A smaller value (C = 60 F) was chosen in order to secure the system against
the self-excitation phenomenon. The energy stored in capacitors must be lower than the
magnetization energy. This is shown below:
ω1  200 rad / s  f  31.84 Hz  60 106  65.54 106 , (182)
ω1  210 rad / s  f  33.43 Hz  60 106  59.45 106 . (183)
From previous calculations, it is observed that up to a speed of 200 rad/s the
capacitor can not provide enough reactive energy to excite the machine. The self-
excited danger appears only over this rotation speed.
The self-excited phenomenon was simulated in Matlab / Simulink with
DSWIG as generator, with a large resistance connected to the main winding terminals:
R load =1500 Ω, and the capacitor battery connected to the auxiliary winding terminals
(Fig. 83). Various values of capacitors have been introduced to view the behavior of
the machine at 200 rad/s. The saturated model of the machine (184) was tuned from
magnetization inductance observed from FEM (finite element method) simulation. In
this way it could acquire the high sinusoidal voltage necessary to determine the
inductances:
 2
k lsat1  I ev 1 
Lm     L m0 , (184)
k  I 3
 k  I 2 
 lsat3 ev lsat 2 ev 
 L  L m  L 2σ  L 12σ , (185)
where I ev represent the peak auxiliary current. Figure 83 shows the self-excited
phenomenon obtained from simulation, at ω = 212.8 rad/s and C = 60 μF [86].
a)
b)
c)
Fig. 83 Self-excited process simulate case: a) main line voltage, b) auxiliary line voltage, c) auxiliary
current [86].
8.4 Control strategy for grid connection case
Using a controlled converter the speed range will be extended from 10 rad/s to
26 rad/s. The controlled converter will be used to boost the voltage in order to provide
the extra magnetization current at low speed. The active power is decreasing with the
cube of speed, and the torque with the square of speed which means the active current
through the active rectifier is decreasing and it is possible to introduce a reactive
component. In this case the control strategy is illustrated in fig. 84 and the Simulink
diagram of the DSWIG drive with controlled converter in fig. 85 [80], [81].
a)
Three phase
active rectifier
(Inverter)
iDC
Cexcitation
im
VDC
ie
 Rectifier
control Main Excitation
winding winding
Tshaft
Input
speed

Short circuit
rotor winding
DSWIG electrical model

b)
1
1 8 -K -
w 0.1s+1
Gain 11 Gain 6 K*u 1
Transfer Fcn
Vm
cos Gain 7
1
Product 1
s
Trigonometric
Integrator 3 Function
sin
-7.5
Trigonometric
Constant Function 1
Voltage
c)
Fig. 84 a) The control strategy of the DSWIG drive with controlled converter; b) Block diagram; c)
converter control [39], [80], [81].
w Vm Dot Product Saturation
Voltage
Converter
Vm Im 1/3 sqrt
Dot Product 1 C1 Math

Function 2 Power
Product
1/3 sqrt
1 Current
-K - Ve Ie Dot Product 3 C2 Math
s
Function 1
C Integrator
w Tem
Speed
Dot Product 2 Saturation 1
IG
Fig. 85 Simulink diagram of the DSWIG drive with controlled converter [81].
The dynamic simulations are presented in fig. 86-90 for speed variation [80], [81].
a)
b)
Fig. 86 Main phase voltages, main phase currents, excitation phase voltages, excitation current: a)
waveforms; b) waveforms in zoom [81].
Fig. 87 Magnetisation and transient magnetization inductances [81].
Fig. 88 Rms main phase currents, rms excitation current and torque [81].
Fig. 89 Output power, input power, efficiency and wind speed variation [81].
Fig. 90 Magnetization current magnitude and magnetization inductance.
Figure 91 shows the original proposed scheme for reduced cost low speed wind or
hydro energy conversion system with DSWIG, and figure 92 shows dynamic simulations
for this case [82].
Fig. 91 The proposed solution for low speed wind or hydro energy conversion system [82].
In figure 93 the wind profile and turbine speed are shown, in figure 94 main
and excitation voltages and currents, in figure 95 are shown torque (turbine and
generator) and power (input and output), in figure 96 main load and excitation currents
(rms values), and in figure 97 efficiency and magnetization inductance respectively
[80].
Fig. 92 Input power, output power, efficiency and speed variation for new proposed solution [82].
Wind speed Turbine speed

24
8
22
7.5
20
7
18
Speed (rad/s)
6.5
Speed (m/s)
6 16
5.5 14
5 12
4.5
10
4
8
0 20 40 60 80 100 0 20 40 60 80 100
Time (s) Time (s)
a) b)
Fig. 93 a)Wind speed profile; b) turbine speed [80].
Main current
Main voltage
300 6
4
200
2
100
Current (A)
Voltage (V)
0
0
-2
-100
-4
-200
-6
-300
0 20 40 60 80 100 0 20 40 60 80 100
Time (s) Time (s)
a) b)
Excitation current
Excitation voltage
5
300 4
3
200
2
100 1
Current (A)
Voltage (V)
0
0
-1
-100
-2
-200 -3
-4
-300
-5
0 20 40 60 80 100 0 20 40 60 80 100
Time (s) Time (s)
c) d)
Fig. 94 Grid connection for speed variation with controlled converter: a) main phase voltage; b) main
phase current; c) excitation phase voltage; d) excitation phase current.
Power (W)
a) b)
Fig. 95 a) Torque of turbine and generator; b) input and output power for the generator for speed variation
with controlled converter [80].
Excitation current (rms value)
4
Load current (rms value)
5
3.5
4.5 3
4 2.5
Current (A)
Current (A)
2
3.5
1.5
3
1
2.5 0.5
2 0
0 20 40 60 80 100 0 20 40 60 80 100
Time (s) Time (s)
a) b)
Fig. 96 a) Main winding load current [80]; b) excitation current for speed variation (rms values).
Efficiency
0.85 Magnetization inductance)
0.29
0.8
0.28
0.75
Inductance (H)
0.7 0.27
0.65
0.26
0.6
0.25
0.55
0.5 0.24
0 20 40 60 80 100 0 20 40 60 80 100
Time (s) Time (s)
a) b)
Fig. 97 a) Efficiency; b) magnetization inductance for speed variation with controlled converter.
8.5 Simulations of start and stop of DSWIG in motor and generator

regimes
Scheme simulations for starting and stopping in the motor regime of DSWIG
are shown in figure 98, the results of the simulation in figures 99, 100 and for the
generator regime in figure 101 [86].
Fig. 98 Simulink diagram for simulation start and stop of DSWIG in the motor regime.
Current
Voltage
Speed
Fig. 99 Simulation start in the motor regime for DSWIG.
Current
Voltage
Speed
Fig. 100 Simulation stop in the motor regime for DSWIG.

Current
Voltage
Speed
a)
Current
Voltage
Speed
b)
Fig. 101 Simulation start (a) or stop (b) in the generator regime for DSWIG.
Chapter 9
EXPERIMENTAL WORK
9.1 Parameters identification results
The no load and short circuit tests which were performed on the two windings
induction machine in order to identify the circuit parameters were presented in chapter
6. The tests were performed on the 11 kW, 8 poles, 630 V, and 50 Hz dual stator three
phase induction machine with a cage rotor. The main dimensions are: stator core outer
diameter D so = 267.8 mm, stator core inner diameter D si =180 mm, core length L c = 230
mm, number of stator slots N s = 72, with cross section S s = 128 mm2, and rotor slots N r
= 58, with cross section S r = 111 mm2. The first winding is a two layer winding with 8
teeth coil span and 0.94 mm2 cross section placed near an air-gap while the second
winding has a single layer winding with 1.73 mm2 cross section. Each winding has 96
turns per phase. The prime mover is simulated by a three-phase cage-type induction
machine driven by an inverter of ABB ACS 800-11 (figure 102) [80], [82], [86].
Fig. 102 Generator DSWIG (front right) coupled with prime mover induction motor (front left) [84].
94 Experimental work – 9
The aim of the experimental tryouts as a first step is the computation of the
DSWIG parameters and characteristics for the stationary regime and their comparison
with the values obtained through optimal design and finite element analysis.
No load probes
Through an ideal no load probe we understand the situation in which the
machine is rotated from the exterior by the prime mover induction motor with the
synchronism speed, which for 8 poles and 50 Hz is of 750 rpm. This is done both for
the main winding (m) supplied from the three-phased autotransformer and the
excitation winding (e) in open circuit, and also the other way around (figure 103). The
real no load probe (figure 104) is made by taking turns in supplying by autotransformer
the two windings (m, e), without supplying at the inverter of prime mover induction
machine (the current absorbed by it is null). The results of the measurements are
presented in appendix 1.
400V, 50Hz
AT
R m R
Inver V, A, S
S I.M. I.G.
ter W
T A e T
f=50Hz
V
Fig. 103 Experimental scheme for ideal no load probes of DSWIG .
400V, 50Hz
K AT
m R
V, A, S
I.G.
W
e T
n
Fig. 104 Experimental scheme for real no load probes of DSWIG .
The equivalent phase scheme for the no load methods ( V m ≠ 0 and V e =0,
respectively V e ≠ 0 and V m =0) is shown in fig 9. The equations which characterize the
functioning in these conditions are (27)-(30).
Short circuit probes
The setup of the short circuit is done through a low voltage supply of two
phases of the main winding or of the excitation winding, the other stator winding being
either in open circuit or in short circuit. This way, there is no rotation tendency of the
shaft. The nominal values of the currents should not exceed rated values. The
measurement schemes are shown in figure 105.
AT, 400V, 50Hz
V, A, m
I.G.
W
m
e e e
V, A, W
a)
m
AT V, A, m
I.G.
W
m
e e e
V, A, W
b)
e
AT V, A, e
I.G.
W
e
m m m
V, A, W
c)
e
AT V, A, e I.G.
W
e
m m m
V, A, W
d)
Fig. 105 Experimental schemes for the short circuit regime of DSWIG a) with open circuit excitation; b)
with short circuit excitation; c) with the main winding in open circuit; d) with the main winding in short
circuit.
An equivalent phase scheme in short circuit is shown in figure 10 and the

equations are (31)-(37). The results of the measurements are presented in appendix 1.
Some pictures from the preliminary tests are included in the same appendix. In these
experiments, the value of the auxiliary excitation capacitors is 35 μF, and the value of the
filter inductances is 23 mH.
With the help of the previous measurements and equations, the stationary
regime parameters are computed for the DSWIG. In figures 106-112 some experimental
characteristics and parameters are shown [83], [84].
a) b)
Fig. 106 Ideal no load excitation winding and main winding: a) voltages versus current; b) active and
reactive power versus current [84].
a) b)
Fig. 107 Short-circ. active and reactive power versus current: a) excitation winding; b) main winding [84].
300
250
Vm0
Vem0
200
Ve0
150
100
50
Vme0
Vm0
0
0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
a) b)
Fig. 108 Voltages versus current: a) real no load; b) short-circuit [84].
Fig. 109 Real no load active and reactive power versus current [84].
160
Xm0iq
140
120
Zm0i
Xm0q
100
Xm0iz
80
60
Zm0
40
20
Xm0z
0
0 0.5 1 1.5 2 2.5
a) b)
Fig. 110 Impedances and reactances versus current at ideal and real no load:
a) main winding; b) excitation winding [84].
16 20
18
14 Xesc1z
Zmsc1
16 Zesc1
12
Xmsc1q Xesc1q
14
10 Xmsc1z
12
8
10
Zmsc2
Zesc2
6 8
4 6 Xesc2z
Xmsc2z
Xesc2q
Xmsc2q 4
2 0 1 2 3 4 5 6 7 8 9 10
0 2 4 6 8 10 12
a) b)
Fig. 111 Impedances and reactances versus current at short-circuit:
a) main winding; b) excitation winding [84].
12 8
X12mrq
10 7.5
X12mrz
Rmsc1
7
8
6.5
6
Xe1z Resc2
6
4 Rmsc2
Xe1q
5.5
2
5
Xm1q Xm1z Resc1
0
4.5
-2 4
1 2 3 4 5 6 7 8 0 2 4 6 8 10 12
a) b)
Fig. 112 a) Reactances (main and mutual) versus current; b) resistances versus current [84].
The magnetizing effect of the main current and the best arrangement of
excitation and main windings (at α = - π/2) results from experimental and simulation
determinations. It makes a comparison between experimental results and simulations
[85].
Fig. 113 Voltage and current per phase in the main winding.
Im
Fig. 114 The low voltage and current phase in the main winding.
Fig. 115 Voltage and current in the main winding load operation [85].
Fig. 116 Voltage and current in the main winding load operation (higher load) [85].
Main current
Main voltage
C A B
300 6
C A B
200 4
100 2
Current (A)
Voltage (V)
0
0
-2
-100
-4
-200
-6
-300
59.96 59.965 59.97 59.975 59.98 59.985 59.99 59.995 60 59.95 59.96 59.97 59.98 59.99 60
Time (s) Time (s)
a) b)
Fig. 117 a) voltages in the main winding; b) currents in the main winding from simulation.
Main voltage Main current

80 5
C A B C A B
4
60
3
40
2
20 1
Current (A)
Voltage (V)
0 0
-1
-20
-2
-40
-3
-60
-4
-80 -5
99.9 99.92 99.94 99.96 99.98 100 99.9 99.92 99.94 99.96 99.98 100
Time (s) Time (s)
a) b)
Fig. 118 a) voltages in the main winding; b) currents in the main winding from simulation.
Almost an 180 degree phase shift between current and voltage in the main
winding as against the experimental results (figures 113-116) and the simulation
(figures 117-118) can be observed. The machine operates as a generator near unity
power factor. For the results shown in figure 115 the reference torque for induction
motor driving the generator (turbine emulator) was 16.5%, and the current drawn 12.12
A, speed of 362.5 rpm, cutting RMS current 5.12 A, mains voltage frequency the main
winding 24 Hz and 14.62% power. In figure 116, the reference torque for induction
motor driving the generator (turbine emulator) was 32.8%, and the current drawn 13.3
A, speed of 365.5 rpm, cutting RMS current 5.82 A, mains voltage frequency the main
winding 24 Hz and 40% power.
It should be noted that the excitation winding is a generator for determination
of the experiment (figures 119-121) and the simulations (figures 122-123). In this case
the phase shift is influenced by the presence of the capacitor. Check and thus
conclusions reached in chapter 8, that active power generator can supply both the
primary winding and by the excitation [85].
Ve
Ie
Fig. 119 Phase voltage and current in the excitation winding.
Fig. 120 Voltages in main (yellow) and excitation (blue) windings and main current (withe) with
resistive load in isolated network.
Fig. 121 Winding excitation voltage, when the capacitor is coupled, with the capacity of 35 μF, while
charging the main task of winding [85].
Excitation voltage Excitation current
C 5
A B C A B
300
4
200 3
2
100
1
Voltage (V)
Current (A)
0 0
-1
-100
-2
-200 -3
-4
-300
-5
59.96 59.965 59.97 59.975 59.98 59.985 59.99 59.995 60 59.96 59.965 59.97 59.975 59.98 59.985 59.99 59.995 60
Time (s) Time (s)
a) b)
Fig. 122 a) voltages in the excitation winding; b) currents in the excitation winding from simulation.
Excitation voltage Excitation current
80
0.6
C A B C A B
60
0.4
40
0.2
20
Voltage (V)
Current (A)
0 0
-20
-0.2
-40
-0.4
-60
-80 -0.6
99.9 99.92 99.94 99.96 99.98 100 99.9 99.92 99.94 99.96 99.98 100
Time (s) Time (s)
a) b)
Fig. 123 a) voltages in the excitation winding; b) currents in the excitation winding from simulation.
9.2 Parameters optimal identification
A genetic algorithm with a population size of 200 and a generation number of

100 was used to find the machine parameter based on no load and a short circuit test
series. The algorithm transferred is always the best machine code for the new
generation and it has a mutation rate 0.5%. The parameter boundary and their found
values are given in table X [83].
Table X The parameter boundary and their founded values
Variable Minimum Maximum Found value
ke 0.933 1.073 0.992
R 1 (Ω) 3.315 4.485 4.212
R 2 (Ω) 2.04 2.76 2.491
R r (Ω) 1.637 3.177 2.243
R m (Ω) 877.39 1239.26 1418.2
X 1σ (Ω) 0.29 2.91 0.2989
X 2σ (Ω) 0.49 4.9 3.9399
X 12σ (Ω) 0 9.219 7.352
X rσ (Ω) 0 9.219 2.835
X m (Ω) 107.66 122.19 108.47
T 0 (Nm) 0 0.516 0.3043
k 1 (Nm s) 0 6.567*10-3 1.494*10-3
k 2 (Nm s2) 0 84*10-6 51*10-6
P mec (W) 58.027
The no load measured current, active power and reactive power are shown in
figure 124 in comparison with the computed values, based on parameter
identifications. The short circuit with an open winding measured current, active power
and reactive power are shown in figure 125 in comparison with the computed values.
The short circuit with the one winding in short-circuits measured current, active power
and reactive power is shown in figure 126 in comparison with the computed values
[83].
No load current No load - active power
300
2.6
2.4
2.2 250
1.8 200
Current (A)
Power (W)
1.6
1.4 150
1.2
main model
1 main exp 100 main model
aux model main exp
0.8 aux exp aux model
aux exp
0.6 50
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Voltage (V) Voltage (V)
a) b)
No load - reactive power
2000
1800
1600
1400
1200
Reactive (VAr)
1000
800
600
400 main model

main exp
200 aux model
aux exp
0
0 50 100 150 200 250 300
Voltage (V)
c)
Fig. 124 No load test and computation comparisons: a) current; b) active power; c) reactive power [83].
Short circuit 1 - current

8 Short circuit 1 - active power
800
7
700
6
600
5 500
Current (A)
Power (W)
4 400
3 300
main model
2 main exp 200
aux model main model
aux exp main exp
1 100 aux model
aux exp
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
a) b)
Short circuit 1 - reactive power
1800
1600
1400
1200
Reactive (VAr)
1000
800
600
400
main model
main exp
200 aux model
aux exp
0
0 20 40 60 80 100 120
Voltage (V)
c)
Fig. 125 The short circuit with one open winding comparison: a) current; b) active power; c) reactive
power [83].
Short circuit 2 - current Short circuit 2 - active power

11 1400
10
1200
9
8 1000
7
800
Current (A)
Power (W)
6
600
5
4 400
3 main model main model

main exp 200 main exp
2 aux model aux model
aux exp aux exp
1 0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
a) b)
Short circuit 2 - reactive power
1200
1000
800
Reactive (VAr)
600
400
main model
200 main exp
aux model
aux exp
0
0 10 20 30 40 50 60 70 80
Voltage (V)
c)
Fig. 126 The short circuit type 2 comparisons: a) current; b) active power; c) reactive power [83].
The current, active and reactive power average errors (rms value) are shown in
Table XI for each test [83].
Table XI The current, active and reactive power average errors

Test type I (%) P (%) Q (%)
a 5.4 6.0 9.9
b 3.8 5.8 6.1
c 4.9 10.4 3.0
d 1.9 2.5 5.8
e 3.2 6.0 3.5
f 1.3 1.3 2.4
The overall average error is 5.3%. The larger error in the no load tests could be
explained by measurement error and voltage and machine unbalance. The voltage
unbalanced average on the main winding was 0.7% but at low voltage the unbalanced
level was 1.7%. This produces a 2.2% average unbalanced current with the maximum
at 4%. When the auxiliary winding was supplied the unbalanced current level was
3.2% with some peaks at 4.4%. The variation of the short circuit resistance, fig. 127,
versus test voltage also shows the level of measurement error. The short circuit
reactance is less affected by random error but the coupling leakage reactance of the
stator windings and the rotor leakage reactance are affected by saturation and they are
not constant in a real machine as in the model [83].
6.6 16
R1sc1
6.4 X2sc1
14
6.2
6 12
R2sc2 X1sc1
5.8
R1sc2 10
5.6
8
5.4
5.2
6
X2sc2
5
R2sc1 4
4.8
X1sc2
4.6
0 2 4 6 8 10 12 2
0 2 4 6 8 10 12
Voltage (V)
Voltage (V)
a) b)
Fig. 127 The short circuit: a) resistance’s; b) reactance’s [83].
The parameter variation and the measurement error are comparable with the
average error between model current and power and measured current and power
respectively. The optimization algorithm features are presented through the objective
function evolution, figure 128 and through induction machine parameter evolutions,
figure 129.
0.25
cost min
cost max
cost med
0.2
Cost function
0.15
0.1
0.05
20 40 60 80 100 120 140 160 180
Generation
Fig. 128 Population evolution statistics [83].

Windings resistance
Second windig voltage ratio 4.5
1.06
1.05 4
R1
R2
1.04
Rr
3.5
1.03
(Ohm)
1.02
k e=V 1/V 2
1.01
2.5
1
0.99
2
0.98
0.97 1.5
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
Generation Generation
a) b)
Leakage reactances Magnetizsation reactance

8 118
117
7
X1
X2 116
6 X12
Xr 115
5
114
Xm (Ohm)
(Ohm)
4 113
112
3
111
2
110
1 109
0 108
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
c) d)
Iron losses resistance
1450 Mechanical losses at synchronous speed
62
60
1400
58
56
1350
Rm (Ohm)
Pmec (W)
54
52
1300
50
1250 48
46
1200 44
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
e) f)
Fig. 129 The induction machine parameter evolutions: a) voltages ratio; b) windings resistances;
c) leakage reactance’s; d) magnetization reactance; e) iron losses resistance; f) mechanical losses at
synchronous speed [83].
9.3 Experimental investigation of the proposed topologies
A general purpose existing DSWIG was used to prove the proposed scheme
and control experimentally. The existing (6kVA, 400V/415 rpm) prototype of the
DSWIG has two stator windings with the same number of turns per phase and 2/1 wire
gauge. The main winding (2/3 slot area) is connected to a dc bus through a roughly
50% rating PWM inverter. The auxiliary winding (1/3 slot area) is connected to 3
phase capacitor and through a diode rectifier (of 50% power rating) to a secondary dc
bus (figure 130). Steady state and dynamic simulation had shown that in order to
connect the diode rectifier to the same dc link, a turn ratio of around 1.26 between
auxiliary and main winding was necessary. The actual DSWIG could be used to prove
the principle if the auxiliary winding is connected, through a diode rectifier, to a
secondary dc link voltage with the ratio of the dc link voltage and secondary dc voltage
equal at 1.26. A bidirectional power flow industrial converter is used to implement the
dc link circuit, and equivalent dc load and the active rectifier. The dc link voltage value
for the bidirectional industrial power converter is V DC =580V [82].
The secondary dc voltage, V DC2 , was set at 460V considering the required ratio
between dc link voltage and secondary dc voltage. The secondary dc link voltage was
emulated through a variable resistive load connected in parallel with the diode rectifier
and controlled through a static contactor to hold the secondary dc voltage at a constant
value (figure 130). The scope of the test was to prove that the maximum power
available from a wind turbine could be extracted using the DSWIG proposed scheme.
The current harmonics, usually introduced by the diode rectifier, are not presented
when using this experimental setup but the control of the active and reactive power
flow with a single converter remain. The prime mover is simulated by a three-phase
cage-type induction machine driven by an inverter of ABB ACS 800-11. The
parameters of the prototype are given above. The cross sections of the windings of the
existing machine are not the same as is required for the proposed strategy and will be
de-rated at 4kW.
The experimental results, figure 131.a, prove that DSWIG in the proposed
configuration could double the power at high speed where the power in the main
winding is constant and the power through the auxiliary winding is increasing with the
speed. Using the active rectifier the power extraction covers the whole wide speed
range. The extraction power capability (at a given rated current 6.2A for the main
winding) is larger than the power available from a wind turbine. A small variation of
the rectified auxiliary voltage, figure 131b, proves that it is possible to deliver power in
constant voltage dc bus. It also shows that the main winding currents remain constant.
The active rectifier output power can be seen in figure 132 [82]. The self-excitation
phenomenon is strongly dependent on the magnetization inductance nonlinearity and a
test validation was necessary. The experiments were carried out with a double capacity
value in order to reduce the terminal voltage, for safety reasons. The self-excited
phenomenon occurs at lower speed. The following charts show the selfexcited
phenomenon obtained from measurements, for ω = 155.32 rad/s, C=120 μF, figure 133
[80], [86].
Grid Prime mover emulator
R
Drive
S Power IM
Converter
T
Bidirectional power
Auxiliary Main flow industrial converter Grid
C
C Active Cf DC
DSWIG Rectifier Load
(Inverter) Inverter
C
Diode
Rectifier VDC 2
Static
Contactor
Load
Secondary dc link emulator
Fig. 130 Test setup [82].

4000
main
3500 aixiliary
total
3000 turbine
2500
Power (W)
2000
1500
1000
500
0
8 10 12 14 16 18 20 22 24 26 28
frequency f1(Hz)
a) b)
Fig. 131 Preliminary experimental results [82].
2800
2600
2400
2200
Output power (W)
2000
1800
1600
1400
1200
1000
10 15 20 25 30 35 40
speed (rad/s)
Fig. 132 The active rectifier output power [82].
a)
b)
c)
Fig. 133 Self-excited process real case: a) main line voltage, b) excitation line voltage, c) excitation
current [86].
The auxiliary voltage winding increases with speed, as it is shown in fig. 134
where significant differences between test and simulation voltage (Sim1) could be
observed. This error is produced by the difference between the magnetization
inductance curve obtained from FEM, and the real one. Fortunately, a new
magnetization inductance curve could be computed from the self-excited test (fig.135).
In the following chart, the blue curve represents the actual voltage obtained in the
laboratory. From this, the inductance in relation to the current amplitude (186) was
calculated. In fig. 135 comparisons can be noted between the inductance obtained from
FEM (Sim1 curve) and another inductance obtained from calculations (in blue) [86].
U ef
L m  L 2σ  L12σ  (186)
ωel  I eţ
Fig. 134 Auxiliary winding voltage: Sim 1- obtained from FEM,

Sim 2- obtained by changing the coefficients [86].
Fig. 135 Inductance values: Sim 1 - obtained from FEM, Sim 2 - obtained from self-excited test [86].
With the magnetization inductance from the experiments, the coefficients of

the new magnetization inductance were recalculated. These are presented in Table XII.
With these new coefficients, the simulation was repeated for a capacity of 60 F and it
was noted that the self-excited phenomenon occurs at a speed greater than 200 rad/s
[86].
Table XII The magnetization inductance coefficients (186)
The coefficients Sim 1 (FEM) Sim 2

k lsat1 0.03 0.0195
k lsat2 0.0137 0.0197
k lsat3 0.0036 0.0018
9.4 Load operation of DSWIG
The scalar control method was selected in order to address driving DSWIG.
This method is much simpler than other ways of controlling the induction machine
(direct torque control DTC, and vector control directly). The scalar method, which
maintains constant frequency voltage ratio when the machine determines to maintain
constant stator flux results. With this method it is sufficient to prescribe a single
variable sizes, rather than calculating a corresponding relationship of the electric
machine in steady regime. The details of these control schemes are shown in appendix
2.
The DSWIG experimental test setup has many parts which are presented in
detail in appendix 3. Its block design is shown in figure 136 and in figures 137 and 138
both the control and the force parts of the stand are shown [86].
Fig. 135 Block diagram of the experimental setup [86].
Fig. 136 The experiment setup for DSWIG (the control scheme).
Fig. 137 The experiment setup for DSWIG (the force scheme).
Compared with DSWIG simulations for start and stop from subsections 8.5,
the results obtained from the experimental evidence bench when starting and stopping
a in the motor (figure 138) or the generator (figure 139) regimes [86] are shown here.
Current
Voltage
Speed
a)
Current
Voltage
Speed
b)
Fig. 138 The experimental results for starting (a) and for stop (b) DSWIG in the motor regime.
Current
Voltage
Speed
a)
Current
Voltage
Speed
b)
Fig. 139 The experimental results for starting (a) and for stop (b) DSWIG in the generator regime.
The capacitor battery supplies the DSWIG with almost the entire reactive
power at the maximum speed and no load operation: in this way the total apparent
power of the inverter is reduced. At low speed the capacitor battery offers less reactive
power, but the inverter still magnetizes the generator. In the next two figures the
influence of the capacitor value on reactive power consumption is shown. In this
experiment, the wind turbine was emulated by an 11 KW induction motor driven by a
bidirectional inverter. The inverter is limited in current, speed, and torque, with the
possibility of prescribing the necessary torque. The rated torque of the drive motor is
denoted by M n and is equal to 140 Nm. This fact is used to load the DSWIG at
different torques and speeds. The active power from the inverter does not depend on
the capacitor bank presence (fig.140), as the reactive power is decreasing when the
capacitor bank is connected (fig.141) [86].
a) b)
Fig. 140 Active power generated by DSWIG: a) without capacitor battery, b) with capacitor battery [86].
a) b)
Fig. 141 Reactive power absorbed by DSWIG through inverter: a) without capacitor battery, b) with
capacitor battery [86].
The DSWIG capability to generate different active power on the two windings
was proven in laboratory tests at different speeds. The capacitor battery still remains
connected to the auxiliary winding, in parallel with an unpretentious load. The test
results can be seen in the following graphs. The test results without capacitor battery at
speed of  = 190 [rad/s] drive torque M=30%M n Nm and R load =110 Ω is shown in fig.
142. The powers measured during this experiment are: the power on the resistance =
844 W, the power delivered to the DC bus = 661 W and the reactive power through the
inverter = 1421 VAR. The test results with capacitor battery at speed  = 190 [rad/s]
drive torque M=30%M n Nm and R load =110 Ω are shown in fig. 143. The powers
measured during this experiment are: the power on the resistance = 792 W, the power
delivered to the DC bus = 700 W and the reactive power through the inverter = 811
VAR. The DC voltage source is used to power the DSWIG as a motor through the
inverter. The drive motor spins the DSWIG as a generator. When the DC voltage is
greater than the DC source voltage, the diode D is blocked [86].
Ic Ib Ia
a)
b)
c)
d)
e)
Fig. 142 Experimental results without capacitors: a) the main winding currents, b) the main winding
voltage, c) the resistive load current, d) the induced line voltage, e) the capacitor current [86].
Ic Ib Ia
a)
b)
c)
d)
e)
Fig. 143 Experimental results with capacitors: a) the main winding currents, b) the main winding voltage,
c) the resistive load current, d) the induced line voltage, e) the capacitor current [86].
The generated power at the main winding of DSWIG is consumed by

electronic load. This electronic load has voltage, current and power limitation, and is
operating in constant voltage mode.
Chapter 10
DESIGN AND SIMULATION SOFTWARE
The optimal design software of the induction machine [94] was used to design
the dual stator winding induction generator. After that, new software was developed for
parameter computation using the finite element method, parameter identification from
test results, steady state and dynamic simulation of DSWIG.
The magnetization curve based on finite element analyses is computed by the
“dswa.m” MATLAB script, presented in appendix 4.1: that is an interface to the
FEMM 4.2 (finite element software) [95], [96]. The motor data as analyzed (geometric
dimensions, materials for lamination and rotor cage, no-load rated pick current) is
given in the input file. An example of an input file “g3kW8p.m” is presented in
appendix 4.2. The variable “i0_pu” set the per unit magnetization current for finite
element analyses. It could be a scalar, for example “i0_pu =1;”, when only two FEM
simulations are performed: one for rated magnetization current in the excitation
winding and another one for rated magnetization current in the main winding, or a
vector (i0_pu = 0.2:0.2:1.4;) when FEM analyses are run for all values derived from
the “i0_pu” vector. The software computes the linkage flux produced by the excitation
current in the excitation winding, the excitation current in the main winding, the main
current in the main winding and the main current in the excitation winding. The
excitation winding and main winding self-inductance is computed from these fluxes, in
addition to the coupling inductance between excitation and main winding. The two
“three phase windings” are geometrically shifted by 90 degrees, and consequently the
“d axis” of excitation winding is magnetically coupled with the “q axis” of main
winding, in the same way as the “d axis” of main winding is coupled with the“q axis”
of excitation winding. The tables of linkage fluxes and inductances are stored in two
output files: one for when the excitation winding is supplied by current and another one
for when main winding is supplied by current. The names of the output files are given
in the input file through the variables “fname_psie” when magnetization current is set
in the excitation winding, and “fname_psim” when magnetization current is set in the
main winding.
In addition to main script “dswa.m” and input file, “g3kW8p2.m”, the
magnetization curve computation software also contains the following scripts:
“constant.m”,“save_dat”, “set_mat.m”, “M19.m”, and function “htrapez.m”.
The “constant.m” script provides the universal constants as magnetic
permeability of free space and basic electric conductor’s features (copper and
aluminum resistivity at 20 0C, resistivity thermal coefficient and density).
The “set_mat.m” script sets the rotor bar features (aluminum or copper)
according to the stated material in the input file and basic material features from
“constant.m”. It also set the magnetic core features based on the stated core material
and its features given in a material features file. For example if the magnetic material
used is M19, given in input file by its name, and then a script with the name“M19.m”
should exist in the same folder and provide the magnetic saturation curve, power loss
coefficients, electric resistivity and iron density. The “save_data.m” script is used to
store the computed linkage flux and inductance as tables in a two output files.
The “htrapez.m” function is used to divide the slot area in layers (compute the
height of the layer based of slots dimensions and required area of the layer).
Considering the heavy computation effort required for magnetization curve
computation, a data post processing feature is added through the “post_proc.m” script
presented in appendix 4.3. The output data, produced by “dswa”, is reloaded and the
flux and inductances versus current are presented using the “dswa_fig.m” script as in
fig. 47-48 (chapter 7). The air gap flux density versus position and harmonics contents
are also shown, similar to fig. 45-46. Only the first most important harmonics are
presented in fig. 46.
Additional post processing features are the rational polynomial approximation
of the magnetization inductances, in equation 159 (chapter 8), shown in fig. 50 by
running the “Lm_apr.m” script (see appendix 4) after the “post_proc.m” script or
independently when the “fname_psie” variable has been previously initialized with a
valid file name that contain a magnetization curve computed with “dswa.m”. The
“Lm_apr.m” script represents the approximation function and sample points while the
approximation polynomial coefficients are computed based on the input parameter
L m0 =0.392; L tmin =0; a 1 =8.2; b 1 =3.751; b 2 =1/6.5^3; these are given first by the user
experience (see chapter 8) and could then be slightly adjusted by the user in order to
result in a good correlation between approximation curves and sample points. Optimal
algorithm like the modified Hooke Jeeves algorithm or a genetic algorithm could be
used to adjust the input parameter in order to minimize the cumulative square error in
the sample points. The main script of this software is “inductanta.m”, which is
presented in appendix 5 with an example of an input file. Preparing the data and
minimization algorithm are implemented in the “set_hg.m”, “optim.m” for the
modified Hoke Jeeves algorithm and “set_ga.m” and “gaOptim.m” for the genetic
algorithm. The output graphics are shown by using the “desen.m” script. These scripts
are not included in the book.
An optimal genetic algorithm was proposed for parameter identification of the
equivalent circuit (fig. 23 chapter 6) from the test results. It was implemented in the
main script “miparam.m”, presented in appendix 5 and in the following auxiliary
scripts: “pt_fig.m”- test graphics;“set_ga.m” – parameter of genetic algorithm settings;
“par_gaOptim.m” – kernel of the genetic algorithm minimization; “par_grafic.m”,
“pt_fig2.m” – shown optimal parameter identification graphics; “save_param.m” –
save the computed parameter in the output file; “cost_obj.m” – objective function
(minimum sum of square errors) evaluation; “par_best.m” – storing the best fitting
parameter; “circ_eval.m” – evaluate the circuit model using actual parameter;
“mutation.m” – implementation of the mutation concept; “circ_err.m” – error between
124 Design and Simulation Software – 10
circuit evaluation and sample test results. The test results and parameter identification
results are presented in paragraph 9.1 respectively in paragraph 9.2.
The dynamic model and simulation of different control strategies were
developed in Matlab-Simulink and the block diagrams are presented in chapter 8. The
dynamic saturated model of the double stator winding machine, presented in fig. 52, is
the most important block and it is used in all dynamic simulation scenarios.
Chapter 11
CONCLUSIONS
The optimal design had shown the parameters and performance complex
dependence on the number of poles. However, the 16 poles generator is considered the
optimum because it has the maximum efficiency, while the active material cost is also
near its minimum and the power factor (which influences the power convertor cost) is
not large. The double stator winding induction generator could reduce the inverter
(active rectifier) KVA by between 43% 28% (for 8 to 24 poles), when the inverter is
placed on the excitation winding. The power converter reduction decreases when the
number of poles is increasing because the inverter has to handle more reactive power
with an increasing number of poles. The full power rectifier and the dc-dc boost
converter required in this case drastically reduce the advantages of this scheme. The
converter KVA reductions are smaller (19% to 30.6%) when the inverter is placed on
main windings. In this case the converter power reduction is higher for a large number
of poles where the converter for the standard induction machine is also larger due to
smaller power factor. The method does not require other power electronics but it
requires a capacitor battery on the excitation windings. If the DSWIG is the same size
as the single winding induction machine, the efficiency will be decreased by around
5% and create cooling problems. The efficiency could be reduced (only 0.5%-1%
efficiency reduction) at the cost of increasing the DSWIG size by 10%-6.5% (8 -24
poles) in diameter which means 29% to 37.8% in the generator weight and about 32%-
35% in the active material costs. Economic benefits will be obtained when the full
power inverter price is larger than generator price by more than 50%, which is the case
for small power machines. This book presents a new topological variant of the dual
stator winding induction machine (DSWIG) operating in generator mode (84%
efficiency at 170 Nm and 250 rpm, 3kW for 1.93 Nm/kg at 8 poles and 3.48 Nm/kg at
24 poles). The dedicated optimal DSWIG design could further reduce the DSWIG over
sizing by several percent, but new winding connection topologies and inverter controls
could improve further the DSWIG concept in order to be competitive with wind or
hydro power plants.
A mathematically perfect saturated model should be implemented in flux
rotating frame coordinates in order that the transients and steady state magnetization
inductance could be considered. In our application a long simulation time is required,
without fast transients. Following this, a simplified version of the model using stator
frame coordinate is applied with only the steady state magnetization inductance. The
model and simulation results are still accurate despite the large difference between
steady state inductance and transient inductances, because the magnetization current
magnitude is constant during the fundamental currents and voltages. The dynamic
126 Conclusions – 11
simulation proves that the induction generator is able to start and run with full load
without any control and with controlled converter. The active controller is also used to
adjust the active power flow in order to extract the maximum energy for a given speed
wind. The main advantage of the DSWIG is its improved capability to operate at
variable low speed.
It presents an optimal method for induction machine parameter identifications
which could be allayed to the classical rotor cage machine with a single three phase
winding in the stator and also to a dual stator winding cage-rotor induction machine.
The estimate is performed by using an equivalent scheme to the DSWIG and a simple
standard test without requiring mechanical coupling of two machines or a mechanical
sensor. The model uses constant parameters as they are required in practical
applications. The linkage inductance variation with the magnetization current could
also be added using no load test points at larger voltage. During the parameter
identification the machine speed is also an estimate at each no-load test point. If the
speed variation range is small then the mechanical losses distribution in Columbian
friction, viscous friction and fan friction are poor in precision but the total mechanical
losses (sum of the components) are more precise. For future work, a better thrust
weighing factor could be computed considering the unbalanced current level and the
short circuit resistance deviation from and to the average value.
Digital simulations as well the preliminary experimental results prove that the
reduced power active rectifier, able to cover only the rated active power, could be used
for a large speed range when the prime mover power increases in direct ratio or more
with the speed. The reactive power is produced only by the capacitor battery at rated
speed and the active rectifier is able to assist the reactive power at low speed. The
proposed configuration is able to run self excited at full power at rated speed wind in
emergence mode (without any power electronics) with the same excited capacitor and
at low power with some stability problem at low wind speed. Increasing the capacitor
battery by 50% will improve the system stability and self excited generator power
capability at low wind speed.
In contrast to other dual stator winding cage-rotor induction generator systems,
the one presented here (through steady state, transient modeling and preliminary test
results) uses a (40-50) % power rating PWM converter in the main winding to deliver
both active and some reactive power in a controlled manner acting alone at low speeds.
At high speeds the (33-50) % power rating excitation winding with parallel terminal
capacitors and diode rectifier produces (40-50) % more active power. A wide speed
range brushless generator with a (40-50) % PWM converter was thus obtained, with
self starting capability.
The DSWIG scheme proposed in this paper uses an inverter with apparent
power lower than the corresponding generator power. The expected ratio between the
inverter power and the generator power is 50% in the case of DSWIG. This is similar
to DFIG applications. The advantage of DSWIG is the lack of brushes. DSWIG can be
used in variable speed applications. It is possible to extract low power even at low
speeds, which cannot be obtained when the generator is directly connected to the grid,
or when the generator has an inverter on the excitation winding and a diode bridge on
the main winding. The DSWIG typology is an advantageous solution when it supplies
unpretentious loads. The inverter on the main winding is used to transfer active power
and also reactive power required for generator magnetization at low speeds, when the
capacitor could not provide enough reactive power. A method to determine the
saturated inductance of the DSWIG was also developed. Digital simulations and
experimental results, in good correspondence, prove the validity of the theoretical
considerations.
APPENDIX
Appendix 1 – The results of the experimental measurements for parameter

identification
Test 1 The ideal no load test with main winding feed from auto-transformer
(AT) and excitation winding is open
Test 2 The ideal no load test with excitation winding feed from AT and main
winding is open
Test 3 The real no load test with main winding feed from AT and excitation
winding is open
Test 4 The real no load test with excitation winding feed from AT and main
winding is open
130 Appendix
Test 5 Short circuit test with main winding feed from AT, with reduced voltage
in two phases, and excitation winding is open
Test 6 Short circuit test with main winding feed from AT with reduced voltage
in two phases, and excitation winding in short circuit
Test 7 Short circuit test with excitation winding feed from AT, with reduced
voltage in two phases, and main winding is open
Test 8 Short circuit test with excitation winding feed from AT, with reduced voltage in
two phases, and main winding in short circuit
132 Appendix
Test 9 The real no load test with star connection of main winding (open) and
delta connection of excitation winding (feed with high voltage)
Test 10 The real no load test with delta connection of main winding (feed with high
voltage) and star connection of excitation winding (open)
Fig. A1.1 Three phase Autotransformer and portable analyzer and phase current measuring pliers.
Fig. A1.2 The portable analyzer with voltages measuring probes.

134 Appendix
Fig. A1.3 The terminal box of DSWIG with voltages probes.
Fig. A1.4 FLUKE 434 portable analyzer to measure currents, voltages, power, power factor, and phase
shifts.
Fig. A1.5 Panel for the power frequency converter drive machine
136 Appendix
Appendix 2 – Schemes for data acquisition used to the experimental work
A first attempt to control the induction machine DSWIG was based on a

simple control, open loop. The program was developed in Simulink using both the
default libraries blocks and special blocks used for communicating with dSpace. This
library is called "Library Board RTI1103" and is divided into two main sections
corresponding to the two processors stored on the device, namely "Master PPC (Power
PC)" and "Slave DSP F240". Accessing either of them displays function blocks that
provide associate access input / output processor information (figure A2.1) [86].
Fig. A2.1 Library PPC Master and Slave RTI 1103 F240 DSP [86].
The scheme used is shown in figure A2.2 [86].
Fig. A.2.2 Programs implemented in open loop [86].

The “Discrete – Time Integrator 1” block, from Fig A 2.2, provides the voltage angle
by integrating the radian frequency (electric speed). The symmetrical three-phase
voltage system, with constant V/f ratio, is obtaining by multiplying the frequency with
sinusoidal waveforms from the three “sin” block outputs. The argument of the function
is composed of a reported value (the prescribed angular velocity that is equivalent to
the maximum angular speed of 314 rad/s). Using dSpace equipment for communication
with an inverter motor control requires the use of special blocks from the dSpace
library. The PWM signal generation block is DS1103SL_DSP_PWM3 which carries
out the comparison of the triangular signal from the signal generator with the
comparator signal. The function for generating commands to the inverter represented
by PWM signals could be made explicitly and in a relay using virtual triangular signal
generators that would achieve sinusoidal signals with signals triangular comparison.
Using the DS1103SL_DSP_PWM3 block has a number of advantages, one of which is
access to an interrupt generated by slave to master DSP with high utility readings. The
three signals synthesized block is DS1103SL_DSP_PWM3 fill factor. The duty cycle
sinusoidal waveform appears below and is used for generating the inverter control,
which, depending on its amplitude and frequency to control the inverter, produces the
same shapes and voltages relative to the value of the duty cycle. Since the fill factor is
defined in theory with values in the range [0, 1], 3 sinusoidal signals must be brought
into this range, without changing the frequency. To achieve this, the signals are
multiplied by 0.5 so that the amount there of is in the range [-0.5, 0.5] was added
followed by 0.5, to move signals in the interval [0, 1]. The Blocks "Enable PWM" and
"Enable INV" are used to activate and deactivate the inverter control. They were
assigned to one button in the interface of the Control Desk. "Enable INV" is a block
with which a signal is sent directly to the inverter with the aid of the
DS1103SL_DSP_BIT_OUT_C13 block from the Simulink library. "Enable PWM" is
an indirect block of commands used to control the inverter. The signal of this block
enters DS1103SL_DSP_PWM3 so that once arranged, the block function will not stop
comparing sine wave with the triangle, but will stop sending signals to the inverter. In
the logic port Stop PWM: value 0 - interrupts the connection with the inverter; value 1
- liaising with the inverter.
The block "NOT" is a logical operator, to deny the value of the input. It was
used as buffer between the dSpace interface board and the inverter which when
mounted on dSpace has negated output, thus reversing the inverter controls. It was
required to enable Stop PWM port logic and eliminate possible human error. The block
"double" may be used to convert an input signal to any of the Simulink functions in an
output signal of the type and the user specified scale. Entry can be any real or complex
signal. If the input signal and output signal are real, it will show true for a complex
signal.
The following sizes are required measurement of key process control.
Calculating the degree of importance (torque, electrical power, mechanical power, etc)
requires knowledge of both the forms and values of currents collected 3 phase
induction machine and the voltages generated by the inverter. By knowing the status of
the transistors of the voltage inverter and the DC circuit of the inverter, the 3 line
138 Appendix
voltages can be determined. These parameters are used to measure the current and
voltage sensors. The currents from the inverter stages are measured using three sensors
LEM LA 25-P. They can measure the nominal effective value of an alternating current
of 25A, with a conversion rate of 1: 1000, in order to measure the current need for
mounting a transducer precision resistor in the circuit. To measure the DC voltage
value of the inverter will be used a voltage sensor LEM LV25-P type, with the
possibility of measuring a voltage of 600 VDC maximum, with a conversion rate of
2500: 1000.
The second control program of the induction machine DSWIG is based on the
first version of the program with the control method dot, which was added to block the
acquisition of the currents and voltages in the DC circuit of the inverter (figure A2.3)
[86].
Fig. A2.3 The program implemented with unfiltered signal acquisition [86].
For acquiring the interest signals, the 4 un-multiplexed channels of dSpace are
being used, from Master PPC processor with blocks ADC (analog to digital converter).
In addition to these four channels, all dSpace also has the Master PPC processor and
the other four multiplexed channels. Finally the channels are obtained by multiplexing
number 16. The use of block data acquisition provides the possibility of selecting one
of the 4 channels Master un-multiplexed on PPC, which is making the acquisition. So
the DS1103ADC_C17 block is used to measure the DC voltage and the
DS1103ADC_C18, DS1103ADC_C19, DS1103ADC_C20 currents are used to acquire
all those blocks referring to un-multiplexed channels.
Supported input voltage values for the 4 channels per processor PPC Master
are in the range of [-10V, 10V]. These channels have a +/- 15V overvoltage protection.
Conversion rate analogue channels between the input signal and the signal are shown
in Simulink program is 10: 1. Values obtained by acquisition software to show the true
value of the measured values need a transformation that takes into account the
conversion rates of both sensors and the dSpace community. The waveforms of the 3
phase currents in the acquired at different frequencies, are shown in figures A2.4, A2.5
and A2.6 [86].
Fig. A2.4 Current acquisition at frequency f = 5,093 Hz [86].
Fig. A2.5 Current acquisition at frequency f = 26,101Hz [86].
Fig. A2.6 Current acquisition at frequency f = 40.107 Hz [86].

140 Appendix
It can be seen from the measured values of current waveforms that they are
heavily distorted and affected by disturbances that vary with frequency change. One of
the problems still unresolved in the current acquisition is the synchronizing the
acquisition block with PWM block.
In the next version of the program, this synchronization is achieved using
Trigger Block. By adding Trigger block to a subsystem or to a model (figure A2.7), it is
possible for the execution of the subsystem or model to be triggered by an external
signal. Trigger block can be configured in order to allow a change in the value of the
external signal to trigger the execution of a subsystem at every step of the integration
process. This block can be configured to allow triggering the execution of the subsystem
when calling a specific function. It is therefore possible for an S-Function to activate the
execution of a subsystem or a model several times during a "sample time". A subsystem
may contain just a block Trigger. The acquisition program structure was comprised of
three signals into a subsystem ADC trigger and activation of this block is controlled by
DS1103SLAVE_PWMINT [86].
Fig. A2.7 The program implemented with synchronized signal acquisition [86].
Using the MINT DS1103SLAVE_PW block can generate some interruption on

almost all "sampling period" programs, which is reported by ST1PWM. In practice,
this interrupt is used to synchronize the generation of the slave PWM signal with, for
example, the operation of a digital to analog converter Master. The ADC Trigger block
contains the acquisition of the 3 signals and functioning external triggers (figure A2.8)
[86].
Fig. A2.8 Signal acquisition using the DS1103 SLAVE_PWMINT block [86].
Because different tasks in one program may be interrupted between them, the
form data transfer vector can be interrupted, thus producing incorrect information. This
can be avoided if the data transfer passes through a data buffer. Simulink can choose
the consistency of the data, if the data transfer between various tasks passed through a
buffer, or the user can determine that the data should be transferred as quickly as
possible without a buffer, with the risk of incorrect information. This block used to
implement a transition program allows non-buffered data. It must be placed on the
same level as the reception of the data block (in our case with Trigger ADC) so that it
is connected directly to the input block or to the subsystem that receives and uses the
data. The waveforms of the currents after the synchronization acquisition of the data
are shown in figures A2.9, A2.10 and A2.11.
Fig. A2.9 The waveforms of currents with synchronization acquisition at f = 5.093 Hz [86].
142 Appendix
The enhancement waveform acquisition process is exclusively synchronized.

From the previous figures we can see the absence of trouble disturbances introduced to
any operating frequency. However, a current filter is required to perform precise
control, so capacitors are placed in parallel with coaxial cable measuring resistance to
filter parasitic signals (figure A2.12). The value of capacity was assessed by testing
with regard to the phase angle between the input and the output [86].
Fig. A2.12 The practical realization of the low-pass filter.

Simplifying the analysis of the 3 phase circuit of the machine will introduce a
new scheme of measurement that takes into account the conversion of coordinate
phases (abc) to coordinate system stationary (). This is done using the Clarke
transformation and its inverse. Transform (abc - ) [86]:
 1 1 
1  2  2 
X α    Xa 
X   2   0 3

3  
 Xb . (187)
 β 3  2 2   
 X 0  1 1 1   X c 
2 2
 2 
Transform (-abc):
 
Xa 
 1 0 1 X 
  α
X   2   1 3
1   Xβ  ,
  (188)
 b 3  2 2 
 X c   
 1 3  X0 
  1
 2 2 
where X may be current, voltage or flux. The measurement scheme is shown in figure
A2.13.
Fig. A2.13 Transformation program implemented with transform from phases coordinates (abc) to
stationary coordinate system () [86].
144 Appendix
It should be noted from the scheme that both the currents and the voltages
passing through a switch (abc-). Applying the transformation matrix corresponding
quantities directly on the program produces [86]:
- for currents
 1  2
I α  I a   I b  I c   , (189)
 2  3
I b  Ic   2 
3
 Iβ  I β 
3
I b  Ic  , (190)
3 2 3
- for voltages
2  U 1 U U 
U α    U a  DC   U b  DC  U c  DC  , (191)
3  2 2 2 2 
2  3  U U 
Uβ      U b  DC  U c  DC  . (192)
3  2  2 2 
To view of the waveform of the machine rotating flux, the flux is calculated as
coordinates () in the program. This is done by integrating the difference between the
voltage () and the related voltage drop on the motor windings:
α   U α  I α  R s dt , (193)
 
β   U β  Iβ  R s dt , (194)
where R s = 4.2 [] is the phase resistance of the machine. The control is implemented
in discreet time and then all “continuous-time” blocks should be transformed in
“discreet-time” blocks. The Matlab Simulink tool program “Simulink Model
Discretizer” (figure A2.14) [86] was used to translate the “continuous-time” models
into “discreet time mode”.
Fig. A2.14 Simulink Model Discretizer.

Acquire the 3-phase currents in the main winding are shown in figures A2.15,
A2.16 and A2.17 [86].
Fig. A2.15 The currents waveforms acquired with coordinate transformation f = 5.093 at Hz [86].
Fig. A2.16 The currents waveforms acquired with coordinate transformation f = 26.101 Hz [86].
Fig. A2.17 The currents waveforms acquired with coordinate transformation f = 40.107 Hz [86].
Compared to previous versions in which the three voltages in coordinates (abc)

were synthesized directly, the control diagram in figure A2.18 shows what happens
where voltages are obtained in coordinates (), which generate three sinusoidal
146 Appendix
signals that serve as PWM. Figure A2.19 shows the block diagram of the process of
synthesizing the two sinusoidal signals in coordinates () [86].
Fig. A2.18 Control diagram where voltages are obtained in coordinates (), which generates three
sinusoidal signals that serve as PWM [86].
Fig. A2.19 Voltages synthesizing in coordinates (),achievement U boost at f boost and maintenance of angle
 between [- 2, + 2] [86].
To limit the gradient of increase or decrease in speed prescribed in order to
avoid sudden changes that could cause absorption of a very high current or voltage rise
circuit DC inverter, the block diagram Transfer Fcn1 is introduced into the scheme
with the transfer function:
1
, (195)
2s 1
that occurs in a program discretizing with the method described above. Transformation
by integration of the speed in the angle is realized by an integrator. To limit the angle
value obtained by integrating, the speed that would reach high values while hindering
the Simulink's calculation is performed maintaining the value angle:
θ   ωdt , (196)
between [- 2, + 2]. This is achieved by resetting the integrator at well-defined times.
When the angle  reaches one of the limitations [- 2, + 2], the block "Dead Zone"
provides outputs of 1 for limitation in the positive or -1 for limiting the downside.
Using this signal module that enters in the block "Detect Increase" gives a command
and the integrator is reset to the original value. The difference from other programs is
that the voltage drop of phase resistance of DSWIG is considered.
When the machine is operating as a generator, there is a danger that the energy
produced by the generator might not be dissipated on the electronic resistance. In such
a situation all the surplus energy would be injected into the capacitors in the DC circuit
of the inverter. This phenomenon would increase voltage in the intermediate circuit,
causing the destruction of the inverter. In order to limit and avoid this phenomenon, a
protection program was created that cut continued inverter control (block "Protection")
above a certain DC voltage. This action is felt as the removal of the magnetizing
current of the generator. The protection circuit is shown in figure A2.20 [86].
Fig. A2.20 DC overvoltage protection [86].
In a generator regime, from a technical standpoint, the stand is limited to a DC

voltage circuit 400 V. This limitation is imposed by the electronic load which is
connected directly to the intermediate circuit of the inverter. As a protection,
experiments are carried out at 350 V with an overvoltage limit of 380 V; if it is
exceeded, the inverter output switches off.
The previous programs were written in order to create driving induction
windings in the stator (DSWIG), both as a motor and as a generator which had a
148 Appendix
capacitor in excitation winding auxiliary unpretentious (resistance-phase) consumer
and parallel compensation reactive power.
The following is the calculation scheme of the power circulating through the
system (figure A2.21). In the block "P, Q, M, FLUX", calculations are performed using
the coordinates sizes () voltages which are obtained after applying nonlinearity
[86].
Fig. A2.21 Calculation of active power, reactive power, torque and fluxes [86].
The equations used for calculation are:

- active power
2

P   Iα  U α1  Iβ  Uβ1 ,
3
 (197)
- reactive power
2

Q   I α  U β1  Iβ  U α1 ,
3
 (198)
- torque
2

M  p   Iβ  α1  I α  β1 ,
3
 (199)
A lower apparent power through the inverter is achieved by using a capacitor
battery because reactive power required for the induction machine magnetization is
supplied by the battery of compensation. Checking the value of the capacitor is carried
out to ensure that the magnetization energy of the induction machine is higher than that
stored in capacitors. Capacitors (figure A2.22) are connected to the excitation winding
stand and have the following characteristics: C = 60 [F]; U = 450 [VAC]; f = 50/60
Hz [86].
Fig. A2.22 Capacitors C= 60 F used in the excitation winding.

150 Appendix
Appendix 3 – Experimental test setup for DSWIG
The DSWIG experimental test setup consists of: PC (Personal Computer) with
programs Matlab Simulink and Control Desk, dSpace DS1103 PPC Controller Board,
Danfoss inverter, ABB inverter, DSWIG machine, prime mover (induction machine),
electronic load, three-phase resistive load, DC voltage supply, voltage and current
sensors, communication interface between dSpace and inverter, galvanically separated
with optical fiber, circuit breakers and switches.
In this experimental test setup, the PC interface represents the entire ensemble
motor - generator - inverter with a member of staff that controls the machines. It is the
central unit that controls the whole system that performs real-time the following
functions: activation and deactivation of inverter enabling and disabling the generation
of PWM, changing the speed of the machine, calculating active power, reactive power
and torque, graphical representation of waveforms acquired and synthesized and sizes
processing, etc. The PC runs the Matlab Simulink programs, which were made of
electric machine control schemes DSWIG and the Desk Control program which is the
interface with the dSpace site. Command of the inverter is achieved through dSpace's
computer. Achieving communication between dSpace DS1103 PPC Controller Board
and the inverter is performed using a communication boards (figure A3.1) which is
mounted in the slot Slave I / O (CP31). It also performs the galvanic separation
between the inverter and dSpace due to sending control signals as light through optical
fiber [86].
Fig. A3.1 DS1103 control board.

DSpace control signals sent from Slave to the inverter and the terminal slot I /
O are listed in the table A3.1 [86]:
Table A3.1 Signals mapping

Simbol Semnalele din Simulink Semnal Pin
UP1 DS1103SL_DSP_PWM3 Duty cycle a SPWM1 7
VP1 DS1103SL_DSP_PWM3 Duty cycle b SPWM2 8
WP1 DS1103SL_DSP_PWM3 Duty cycle c SPWM3 9
EN1 DS1103_DSP_BIT_OUT_C13 SBIO 36
RST1 DS1103_DSP_BIT_OUT_C10 STMRCLK 12
The method of operation of the module is based on the operation of the

complementary transistors of the transistors in the same leg. By using the 3 signals
UP1, VP1, WP1 generated by DS1103SL_DSP_PWM3 Duty cycle a / b / c, 3
transistors are controlled from the top of the inverter. Denial signals, UP1, VP1, WP1 ,
represent even transistors command from the bottom of the legs and are automatically
generated. EN1 and activation signals RST1 represent the signals which activate and
reset the inverter. The acquisition system, which in this case is the device dSpace has
analog inputs that allow an input voltage up to 10V input operation that performs
sampling system information. We therefore need equipment to measure the levels of
interest in the system and to convert them into signals accepted entries from dSpace. In
this application, sampling currents on the three phases of DSWIG's main winding and
the DC voltage of the converter were necessary. The sensor boards to acquire and
convert the signals described above were designed and implemented for this reason.
Sensor boards have been designed to be able to use two types of sources,
allowing more sensors on the same board, and different configurations of resistance
measurement. Board sensors contain: power source, current sensors, different
connectors, LED for voltage presence on both the plus and minus, resistance
measurement.
Current sensors used are LEM LA-25 (figure A3.2). They have three pins, two
used for supply and one for measurement [86].
Fig. A3.2 Wiring diagram of the current sensor.
The supply voltage is 12 ÷ 15 V and the current drawn is 10 + I s mA where

I s = 25 mA. Power supply is from Traco Power TMP 15215. The supply voltage is
between 100-240 VAC, and the output voltage is 15 V and output current 500 mA.
Internal diagram of such a sensor is shown in figure A3.3.
Prefabricated voltage sensors are used for different values of voltages. The use
of these sensors is necessary for a board that sends signals from the sensor through a
152 Appendix
BNC to dSpace. The supply voltage is between 12 ÷ 15 V and the current drawn is 10
+ I s mA where I s = 25 mA. The power supply voltage of the sensor board is also used
to power the current sensor (figures A3.4 and A3.5) [86].
Fig. A3.3 Internal scheme of the current sensor.
Fig. A3.4 Internal scheme of the voltage sensor.
Fig A3.5 The voltage sensor and transmitted board.

To command DSWIG, a voltage inverter model FC-Danfoss 302P4K0 was
used. This frequency converter is high performance and can support multiple methods
of control for motors such as U / f = constant, VVC plus (voltage vector control) or
flux vector control. Technical specifications: for input U = 3x380 – 500 V, f = 50/60
Hz, I = 9/7.4 A and for output U = 3x0 – U IN , f = 0 – 590 Hz, I = 10/8.2 A, P N = 4 KW.
The converter is unidirectional between supply and DC circuit and bidirectional
between DC circuit and electrical machine (figure A3.6) [86].
Fig. A3.6 Inverter internal scheme.
This converter can control either speed or torque on the motor shaft. The speed
control can be used for open-loop speed control if the motor does not require feedback
or closed loop, in which case a PID controller is used and the reaction requires input
from the machine shaft speed. This latest version provides greater accuracy than the
original [86].
Fig. A3.7 Danfoss inverter control panel (left) and control board (right).
154 Appendix
The torque control can use either the voltage vector control and torque (torque
VVC plus) open-loop or closed-loop control of flux with speed measurement to the
motor shaft. The latter offers superior performance in all four quadrants of operation
and for the entire speed range of the machine. VVC plus open-loop is an application
based on current measurement and offers limited accuracy for torque calculation. VVC
plus (Voltage Vector Control) in comparison with the scalar control, improves
dynamics and stability at any change of the reference speed and torque load. It is worth
mentioning that this project local control panel was replaced by a plate that allows
direct access to commands transistors (figure A3.7).
The voltage inverter used to control motor drive is model ACS800-11-0025-3
+ ABB + K466 + P904 E200. It is a four-quadrant inverter, which allows the machine
operation both as a motor and as a generator. The inverter controls motors through the
direct torque control method with increased accuracy and speed in torque control
without the use of encoders and tachometers. The current is measured on two of the
motor phases and the voltage of the DC circuit. The third phase current is measured for
protection against ground faults. ACS800 series has a simplified control panel because
it uses an alphanumeric intelligent inverter which can be mounted on or in a place that
allows easy access to the user (figure A3.8). Technical specifications for ABB
ACS800-11-0025-3: U N = 400 V (variation range 380-415 V), I cont. max = 47 A, I max =
68 A, P cont. max = 22 kW, = 97 % to the rated power [86].
Fig. A3.8 Schematic diagram of the ABB inverter.
The main circuit consists of two IGBT converters, a line-side converter and a
motor-side, both integrated into the same chassis. The line-side converter is a
bidirectional converter (capable of transferring energy to both sides) connected
between the power supply and the DC circuit. The motor-side converter is also a
bidirectional converter connected between the DC network and drive motor. It deals
with the operation of the motor. Filter EMC (Electromagnetic Compatibility) is used to
ensure proper functioning of equipment in an environment electromagnetic
interference. An LCL filter is used to eliminate harmonics. Varistors are used to protect
against overvoltage peaks.
The induction generator (figure A3.9) represents the element which helps to
develop the control scheme. This machine, with two windings in the stator, is
connected directly to the Danfoss inverter. Excitation winding is connected to three-
phase resistance and a capacitors battery is connected in parallel. The drive motor will
be connected by a rigid coupling with the DSWIG shaft [86].
Fig. A3.9 DSWIG generator.
The drive motor (figure A3.10) is an induction machine with the following
parameters: P = 11 kW, insulation class F, for delta stator winding connection U = 230
V, I = 33 A, for Y stator winding connection U = 400 V, I = 24,7 A, cosφ= 0,76, n=
725 rot/min.
Fig. A3.10 Electric drive motor.

156 Appendix
The programmable voltage source used to power Danfoss inverter is model
Topcon Quadro TC.P.10.400.400.PV from Regatron. This allows for the control of
voltage, current and power output. Depending on the specifications for output voltage
and current values, it can operate as either a constant voltage source or a constant
current source. Technical specifications: V IN = 3x(360-440) V AC , I IN = 3x20 A rms , f IN
= 48-68 Hz, P OUT = 0 – 10 kW, V OUT = 0 - 400 V DC , I OUT = 0 – 31 A, internal
resistance 0 – 1000 mΩ. Its programming can be undertaken locally from the front
panel (figure A3.11), where the desired parameters and operating limitations can be
set. This panel also displays values and error messages in real time [86].
Fig. A3.11 Front panel DC source.
Another method of programming is available on a PC using the

communication interface RS-232 or RS-422 (figure A3.12). There is software that can
create all the settings that were accessible on the front panel, and various programs can
be written using Visual C ++, to set photovoltaic panels where the source is used for
this purpose. If using such a charger photovoltaic MPPT (Maximum Power Point
Tracking) can follow the position of point of its operation on the characteristic U = f (I)
[86].
Fig. A3.12 DC source software.

An electronic load type EL9400-050 produced by Elektro-Automatik (figure
A3.13) was used for the generator power. Technical specifications: power supply -
U alim = 115/230 V, f = 50/60 Hz – DC input - U IN = 0 – 400 V, P IN = 0 – 2400 W, P 2IN
= 0 – 1500 W in steady state regime, I N = 50 A, U IN.max = 500 V - R = 0 – 10 Ω with
display resolution 10 mΩ, R = 0 – 400 Ω with display resolution 100 mΩ. This type of
source has several different modes and levels of control are available both on the front
panel and through software if it is linked to a PC. Procedures are: CC – constant
current, CV - constant voltage, CP - constant power and CR - constant resistance [86].
Fig. A3.13 Front panel electronic load.
The “electronic load” behavior depends on the selected mode: CC, CV,
CP or CR, until some prescribed limits is reached when its behavior is changed
to keep the parameter under its imposed limit. For example, in CV mode it
maintains a constant voltage until current reaches its imposed limits when the
voltage decreases, in order to keep the current under imposed limits. In CC
mode it maintain the prescribed constant current until the voltage reaches its
limit, when the current is let to be changed in order to keep the voltage limits.
Failure to comply the constructive limits leads to the overload and the self-
protection system will disconnect the “electronic load” that is harmful for the
rest of the system. In this case, the entire generator power flows in the DC link
capacitor, increasing quickly the dc voltage up to dangerous values for the
equipment and operators. A reliable protection system is required to connect a
load resistance in the dc circuit, and stop the prime mover.
To takeover the winding main task was used for seizure of power electronics
and the excitation winding was used a resistive load (figure A 3.14). This is three-
phase and is directly connected to the generator in parallel with the capacitor. Has
placed five-step adjustment from left to right denoted the maximum current that can
158 Appendix
absorb (1A, 2A, 3A, 5A, 12A) to the line voltage 400 V and the various combinations
that can be made. Resistance steps are: step 1 R = 230 Ω, step 2 R = 143 Ω, step 3 R =
90 Ω, step 4 R = 25 Ω, step 5 R = 12 Ω [86].
Fig. A3.14 Resistive load.

Appendix 4 – The magnetization curve computation using finite element analysis
% This script provides an interface to FEMM 4.2 for DSWIG
clearall;
closeall;
fname='g2'; % request for the input file name

eval(fname,'fred_error=1'); % evaluation of the input file
w_name=['Ae';'Be';'Ce';'Am';'Bm';'Cm']; % Name of the motor phase
constant; % universal and material constant
set_mat; % setings the material features
pp=poles/2; % pole pairs
poles1=poles/gcd(Nrs,poles); % number of simulated poles in FEM
q1=Nss/(3*poles); % number of slot per pole per phase
ym1=round(cSpanm*3*q1); % main coil opening
ye1=round(cSpane*3*q1); % excitation
if(layerse==1) ye1=3*q1; end
if(layersm==1) ym1=3*q1; end
Dro=Dsi-2*hag; % outer stator diameter

hsy=(Dso-Dsi)/2-hst; % height of stator yoke
hry=(Dro-Dri)/2-hrt; % height of rotor yoke
if(mwp==1)
nce1=ncme; % number of elementary conductors on base layer
dce1=dcme; % diameter of elementary conductors on base layer
sb1=sbm%number of turns per coil on base layer
layers1=layersm;
nce2=ncee; % number of elementary conductors on upper layer
dce2=dcee; % diameter of elementary conductors on upper layer
sb2=sbe; %number of turns per coil on excitation on upper layer
layers2=layerse;
y1=ym1;
y2=ye1;
win_name='me';
else
nce1=ncee; % number of elementary conductors on base layer
dce1=dcee; % diameter of elementary conductors on base layer
sb1=sbe%number of turns per coil on base layer
layers1=layerse;
160 Appendix
nce2=ncme; % number of elementary conductors on upper layer
dce2=dcme; % diameter of elementary conductors on upper layer
sb2=sbm; % number of turns per coil on excitation on upper layer
layers2=layersm;
y1=ye1;
y2=ym1;
win_name='em';
end
% Computation of Stator slot dimensions
scu1=0.25*pi*sb1*nce1*(dce1)^2*layers1;
scu2=0.25*pi*sb2*nce2*(dce2)^2*layers2;
scu=scu1+scu2; % copper cross sections in a slot
hs1=hst-hs2-hs3-hs4;
ws2=(hs2*ws1+hs1*ws3)/(hs1+hs2);
sau=(ws1+ws2)*hs1/2;
ka1=scu1/scu; % ratio between main coil copper area and total copper area
sa1=ka1*sau;
sa2=sau-sa1;
hc1=htrapez(ws1,ws2,hs1,sa1);
hc2=hs1-hc1;
wsc1u=(hc2*ws1+hc1*ws2)/hs1;
if(layers1==2)
hc1L1=htrapez(ws1,wsc1u,hc1,sa1/2);
hc1L2=hc1-hc1L1;
wsc1L1=(hc1L2*ws1+hc1L1*wsc1u)/hc1;
else
hc1L1=hc1;
hc1L2=0;
wsc1L1=wsc1u;
end
if(layers2==2)
hc2L1=htrapez(wsc1u,ws2,hc2,sa2/2);
hc2L2=hc2-hc2L1;
wsc2L1=(hc2L2*wsc1u+hc2L1*ws2)/hc2;
else
hc2L1=hc2;
hc2L2=0;
wsc2L1=ws2;
end
tryclosefemm; end
openfemm;
newdocument(0);
%Seting the material

mi_probdef(0,'millimeters','planar',1e-8,li,25);
mi_getmaterial('Air');
if(stcore_fem)
%try mi_getmaterial(stcore_fem); catch stcore_fem=0; end
eval('mi_getmaterial(stcore_fem)', 'stcore_fem=0');
end
if(~stcore_fem)
stcore_fem=stcore_mat;
mi_addmaterial(stcore_fem,1,1,0,0,1e-6/rho_ironst,lamThick*1000,0,ssf*lfe/li,0);
mi_addbhpoints(stcore_fem,BHs_ref(1,:),BHs_ref(2,:));
end
if(kmrid<1)
if(strcmp(rtcore_mat,shaft_mat)) shaft_fem=rtcore_fem;
elseif(strcmp(stcore_mat,shaft_mat)) shaft_fem=stcore_fem;
end
if(~(strcmp(shaft_fem,stcore_fem)|strcmp(rtcore_fem,shaft_fem)))
mi_addmaterial(shaft_fem,1,1,0,0,1e-6/rho_ironrt,RlamThick,0,1,0);
mi_addbhpoints(shaft_fem,BHsh_ref(1,:),BHsh_ref(2,:));
end
end
mi_addmaterial('Stcond1',1,1,0,0,1e-6/rho_copper,0,0,0,4,0,0,nce1,dce1);
mi_addmaterial('Stcond2',1,1,0,0,1e-6/rho_copper,0,0,0,4,0,0,nce2,dce2);
mi_addmaterial('Rtcond',1,1,0,0,1e-6/rho_rtb,0,0,0,0);
mi_addcircprop('Am',0,1);
mi_addcircprop('Bm',0,1);
mi_addcircprop('Cm',0,1);
mi_addcircprop('Ae',0,1);
mi_addcircprop('Be',0,1);
mi_addcircprop('Ce',0,1);
%Deffine the boundary

if(rem(poles1,2))
simetry=5; % antiperiodic
else
simetry=4; %periodic
end
mi_addboundprop('Zero',0,0,0,0,0,0,0,0,0);
mi_addboundprop('Periodic_r',0,0,0,0,0,0,0,0,simetry);
mi_addboundprop('Periodic_a1',0,0,0,0,0,0,0,0,simetry);
162 Appendix
mi_addboundprop('Periodic_a2',0,0,0,0,0,0,0,0,simetry);
mi_addboundprop('Periodic_s',0,0,0,0,0,0,0,0,simetry);
%Stator coil distribution matrix

Cname=['ABC'];
islot=1:Nss;
islot_angle=rem(180*poles*islot/Nss,360);
[iss_angleslot_index]=sort(islot_angle);
zone1=slot_index(1:fix(Nss/6));
zone2=slot_index(fix(Nss/6)+1:Nss/3);
zone3=slot_index(Nss/3+1:fix(Nss/2));
zone4=slot_index(fix(Nss/2)+1:2*Nss/3);
zone5=slot_index(2*Nss/3+1:fix(5*Nss/6));
zone6=slot_index(fix(5*Nss/6)+1:Nss);
Cdist(zone1)=1;
Cdist(zone2)=-2;
Cdist(zone3)=3;
Cdist(zone4)=-1;
Cdist(zone5)=2;
Cdist(zone6)=-3;
Cdist1(1,:)=Cdist;
if(layers1==2)
Cdist1(2,fix(3*q1)-y1+1:Nss)=Cdist(1:Nss-fix(3*q1)+y1);
if(y1~=3*q1) Cdist1(2,1:fix(3*q1)-y1)=(-1)^poles*Cdist(Nss-fix(3*q1)+y1+1:Nss);
end
end
dns=fix((3*q1-y2+y1)/2);
Cdist2(1,1:Nss-dns)=Cdist(dns+1:Nss);
if(dns) Cdist2(1,Nss-dns+1:Nss)=Cdist(1:dns); end
if(layers2==2)
Cdist2(2,fix(3*q1)-y2+1:Nss)=Cdist2(1,1:Nss-fix(3*q1)+y2);
if(y2~=3*q1) Cdist2(2,1:fix(3*q1)-y2)=(-1)^poles*Cdist(1,Nss-fix(3*q1)+y2+1:Nss);
end
end
mi_zoom(Dso/2*cos(pi/pp*poles1),0,Dso/2,Dso/2);
% Drawing the geometry

%The stator
if(poles1~=poles)
mi_drawline(Dsi/2,0,Dso/2,0);
mi_drawline(Dsi/2*cos(pi/pp*poles1),Dsi/2*sin(pi/pp*poles1),Dso/2*cos(pi/pp*poles
1),Dso/2*sin(pi/pp*poles1));
mi_addarc(Dso/2,0,Dso/2*cos(pi/pp*poles1),Dso/2*sin(pi/pp*poles1),180/pp*poles1,5
);
else
mi_addarc(Dso/2,0,-Dso/2,0,180,20);
mi_addarc(-Dso/2,0,-Dso/2,0,180,20);
end
hs=[0 hs4 hs3 hs2 hc2L2 hc2L1 hc1L2 hc1L1];

ws=[ws4 ws4 ws3 ws2 wsc2L1 wsc1u wsc1L1 ws1];
ii=find(hs~=0);
hs=[0 hs(ii)];
ws=[ws(1) ws(ii)];
xspred=Dsi/2;
yspred=0;
ys=cumsum(hs);
xs=ws/2;
hs0=sqrt(Dsi^2-ws4^2)/2
for(i_slot=1:Nss/poles*poles1)
cosA=cos((2*i_slot-1)*pi/Nss); sinA=sin((2*i_slot-1)*pi/Nss);
xs1=(hs0+ys)*cosA+xs*sinA;
ys1=(hs0+ys)*sinA-xs*cosA;
mi_drawpolyline([xs1',ys1']);
xs2=(hs0+ys)*cosA-xs*sinA;
ys2=(hs0+ys)*sinA+xs*cosA;
ns=length(hs);
for(j_seg=4: ns)
mi_addsegment(xs1(j_seg),ys1(j_seg),xs2(j_seg),ys2(j_seg));
end
mi_addarc(xspred,yspred,xs1(1),ys1(1),180/Nss-180/pi*asin(ws(1)/Dsi),1);
mi_addarc(xs1(1),ys1(1),xs2(1),ys2(1),180/pi*asin(ws(1)/Dsi),1);
xspred=xs2(1);
yspred=ys2(1);
i_reg=[1,4:ns-1];
ysc=(ys(i_reg)+ys(i_reg+1))/2;
mi_addblocklabel((hs0+ysc(1))*cosA,(hs0+ysc(1))*sinA);
mi_selectlabel((hs0+ysc(1))*cosA,(hs0+ysc(1))*sinA);
mi_setblockprop('Air',1,ws(2)/3,'none',0,0,1)
mi_clearselected;
if(layers2==2)
164 Appendix
mi_setblockprop('Stcond2',1,ws2/3,[Cname(abs(Cdist2(2,i_slot))),win_name(2)],0,1,si
gn(Cdist2(2,i_slot))*sb2);
mi_clearselected;
end
mi_clearselected;
mi_addblocklabel((hs0+ysc(2+layers2))*cosA,(hs0+ysc(2+layers2))*sinA);
mi_selectlabel((hs0+ysc(2+layers2))*cosA,(hs0+ysc(2+layers2))*sinA);
if(layers1==2)
mi_clearselected;
mi_addblocklabel((hs0+ysc(3+layers2))*cosA,(hs0+ysc(3+layers2))*sinA);
mi_selectlabel((hs0+ysc(3+layers2))*cosA,(hs0+ysc(3+layers2))*sinA);
end
mi_clearselected;
end
xs1(1)=Dsi/2*cos(2*pi/Nss*i_slot);
ys1(1)=Dsi/2*sin(2*pi/Nss*i_slot);
mi_addarc(xspred,yspred,xs1(1),ys1(1),180/Nss-180/pi*asin(ws(1)/Dsi),1);
%The rotor
if(poles1~=poles)
mi_drawline(Dro/2,0,Dri/2,0);
mi_drawline(Dri/2*cos(pi/pp*poles1),Dri/2*sin(pi/pp*poles1),Dro/2*cos(pi/pp*poles1
),Dro/2*sin(pi/pp*poles1));
mi_addarc(Dri/2,0,Dri/2*cos(pi/pp*poles1),Dri/2*sin(pi/pp*poles1),180/pp*poles1,5);
else
mi_addarc(Dri/2,0,-Dri/2,0,180,5);
mi_addarc(-Dri/2,0,-Dri/2,0,180,5);
end
hr=[0 hr4 hr3 hr1];

wr=[wr4 wr4 wr3 wr1];
xspred=Dro/2;
yspred=0;
ys=cumsum(hr);
xs=wr/2;
hs0=sqrt(Dro^2-wr4^2)/2
for(i_slot=1:Nrs/poles*poles1)
cosA=cos((2*i_slot-1)*pi/Nrs); sinA=sin((2*i_slot-1)*pi/Nrs);
xs1=(hs0-ys)*cosA+xs*sinA;
ys1=(hs0-ys)*sinA-xs*cosA;
xs2=(hs0-ys)*cosA-xs*sinA;
ys2=(hs0-ys)*sinA+xs*cosA;
mi_addsegment(xs1(4),ys1(4),xs2(4),ys2(4));
mi_addarc(xspred,yspred,xs1(1),ys1(1),180/Nrs-180/pi*asin(wr(1)/Dro),1);
mi_addarc(xs1(1),ys1(1),xs2(1),ys2(1),180/pi*asin(wr(1)/Dro),1);
xspred=xs2(1);
yspred=ys2(1);
ysc=(ys(1)+ys(end))/2;
mi_addblocklabel((hs0-ysc)*cosA,(hs0-ysc)*sinA);
mi_selectlabel((hs0-ysc)*cosA,(hs0-ysc)*sinA);
mi_setblockprop('Rtcond',1,wr(2)/3,'none',0,0,1)
mi_clearselected;
end
xs1(1)=Dro/2*cos(2*pi/Nrs*i_slot);
ys1(1)=Dro/2*sin(2*pi/Nrs*i_slot);
mi_addarc(xspred,yspred,xs1(1),ys1(1),180/Nrs-180/pi*asin(wr(1)/Dro),1);
mi_drawarc(Dro/2+hag/2,0,(Dro/2+hag/2)*cos(pi/pp*poles1),(Dro/2+hag/2)*sin(pi/pp
*poles1),180/pp*poles1,1);
mi_addsegment(Dro/2+hag/2, 0, Dsi, 0);
mi_addsegment([Dro/2+hag/2; Dsi]*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
mi_addsegment(Dro/2, 0, Dro/2+hag/2, 0);
mi_addsegment([Dro/2; Dro/2+hag/2]*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
%Boundary conditions
mi_selectarcsegment(Dso/2*[cos(pi/poles) sin(pi/poles)]);
mi_selectarcsegment(Dri/2*[cos(pi/poles) sin(pi/poles)]);
mi_setarcsegmentprop(5,'Zero',0,0);
mi_clearselected;
166 Appendix
mi_selectsegment((Dri+hry)/2,0);
mi_selectsegment((Dri+hry)/2*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
mi_setsegmentprop('Periodic_r',3,1,0,0);
mi_clearselected;
mi_selectsegment(Dro/2+hag/4,0);
mi_selectsegment((Dro/2+hag/4)*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
mi_setsegmentprop('Periodic_a1',1,1,0,0);
mi_clearselected;
mi_selectsegment(Dro/2+3*hag/4,0);
mi_selectsegment((Dro/2+3*hag/4)*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
mi_setsegmentprop('Periodic_a2',1,1,0,0);
mi_clearselected;
mi_selectsegment((Dsi+Dso)/4,0);
mi_selectsegment((Dsi+Dso)/4*[cos(pi/pp*poles1) sin(pi/pp*poles1)]);
mi_setsegmentprop('Periodic_s',3,1,0,0);
mi_clearselected;
%Material settings
mi_addblocklabel((Dso-hsy)/2*cos(pi/poles),(Dso-hsy)/2*sin(pi/poles));
mi_selectlabel((Dso-hsy)/2*cos(pi/poles),(Dso-hsy)/2*sin(pi/poles));
mi_setblockprop(stcore_fem,1,ws1/5,'none',0,0,1)
mi_clearselected;
mi_addblocklabel((Dri+hry)/2*cos(pi/poles),(Dri+hry)/2*sin(pi/poles));
mi_selectlabel((Dri+hry)/2*cos(pi/poles),(Dri+hry)/2*sin(pi/poles));
mi_setblockprop(stcore_fem,1,wr1/2,'none',0,0,1)
mi_clearselected;
mi_addblocklabel((Dro+hag/2)/2*cos(pi/poles),(Dro+hag/2)/2*sin(pi/poles));
mi_addblocklabel((Dsi-hag/2)/2*cos(pi/poles),(Dsi-hag/2)/2*sin(pi/poles));
mi_selectlabel((Dro+hag/2)/2*cos(pi/poles),(Dro+hag/2)/2*sin(pi/poles));
mi_selectlabel((Dsi-hag/2)/2*cos(pi/poles),(Dsi-hag/2)/2*sin(pi/poles));
mi_setblockprop('Air',1,hag/2,'none',0,0,1)
mi_clearselected;
fem_ffile='g1';
I0_tab=I0n*i0_pu;
npcm=length(i0_pu);
% Running FEM analyses in a loop for magnetization curve

foricm=1:npcm
% Magnetization curve when excitation windings are supplied
I0=I0_tab(icm);
mi_setcurrent('Ae',I0);
mi_setcurrent('Be',-I0/2);
mi_setcurrent('Ce',-I0/2);
mi_setcurrent('Am',0);
mi_setcurrent('Bm',0);
mi_setcurrent('Cm',0);
mi_saveas([fem_ffile,'.fem']);
mi_analyze(0); %flag 0 - visible windows, 0 hide windows
%Data extraction from the first simulation

mi_loadsolution;
fori_w=1:6
rfemm=mo_getcircuitproperties(w_name(i_w,:));
psi_ie(icm,i_w)=rfemm(3)*poles/(poles1);
I_ie(icm,i_w)=rfemm(1);
V_ie(icm,i_w)=rfemm(1);
end
psidee(icm)=psi_ie(icm,1:3)*[2 -1 -1]'/3;
psiqee(icm)=(psi_ie(icm,2)-psi_ie(icm,3))/sqrt(3);
psidem(icm)=psi_ie(icm,4:6)*[2 -1 -1]'/3;
psiqem(icm)=(psi_ie(icm,5)-psi_ie(icm,6))/sqrt(3);
Le1(icm)=psi_ie(icm,1)/I_ie(icm,1);
Led(icm)=psidee(icm)/I_ie(icm,1);
Lem(icm)=psiqem(icm)/I_ie(icm,1);
mo_addcontour((Dsi-hag)/2,0);
mo_addcontour(0.5*(Dsi-hag)*cos(pi/pp*poles1),0.5*(Dsi-hag)*sin(pi/pp*poles1));
mo_bendcontour(180/pp*poles1,1);
s1=sprintf('%2.1f',i0_pu(icm)); ii=find(s1=='.'); s1(ii)='p';
f_Bename=[fname,'_Be',s1,'.txt'];
mo_makeplot(2,512,f_Bename,0);
mo_clearcontour;
%Second simulation - Magnetization curve when main windings are supplied

mi_setcurrent('Am',I0);
mi_setcurrent('Bm',-I0/2);
mi_setcurrent('Cm',-I0/2);
mi_setcurrent('Ae',0);
mi_setcurrent('Be',0);
168 Appendix
mi_setcurrent('Ce',0);
mi_saveas([fem_ffile,'.fem']);
mi_analyze(0); %flag 0 - visible windows, 0 hide windows
%Data extraction from the second simulation

mi_loadsolution;
fori_w=1:6
rfemm=mo_getcircuitproperties(w_name(i_w,:));
psi_im(icm,i_w)=rfemm(3)*poles/(poles1);
I_im(icm,i_w)=rfemm(1);
V_im(icm,i_w)=rfemm(1);
end
psidme(icm)=psi_im(icm,1:3)*[2 -1 -1]'/3;
psiqme(icm)=(psi_im(icm,2)-psi_im(icm,3))/sqrt(3);
psiqmm(icm)=psi_im(icm,4:6)*[2 -1 -1]'/3;
psidmm(icm)=(psi_im(icm,5)-psi_im(icm,6))/sqrt(3);
Lm1(icm)=psi_im(icm,4)/I_im(icm,4);
Lmq(icm)=psiqmm(icm)/I_im(icm,4);
Lme(icm)=psiqme(icm)/I_im(icm,4);
mo_addcontour((Dsi-hag)/2,0);
mo_addcontour(0.5*(Dsi-hag)*cos(pi/pp*poles1),0.5*(Dsi-hag)*sin(pi/pp*poles1));
mo_bendcontour(180/pp*poles1,1);
s1=sprintf('%2.1f',i0_pu(icm)); ii=find(s1=='.'); s1(ii)='p';
f_Bmname=[fname,'_Bm',s1,'.txt'];
mo_makeplot(2,512,f_Bmname,0);
mo_clearcontour;
end
mf_ag=load(f_Bmname);
save_data;
Appendix 4.1 – The input file
poles=8; % number of poles

Dsi=286; % mm iner stator diameter
Dso=386; % mm outer stator diameter
Nss=72; % stator slors number
Nrs=60; % Number of rotor slots
Dri=197; % mm rotor inner diameter
li=140; % mm ideal core length
hag=0.35;
%Stator slot dimensions

hst=31.4; % mm total height of stator slot
hs4=0.3; % mm stator slot tip height
hs3=1.238; % mm stator slot wedge height
hs2=0.65; % mm slot insulation
ws1=10.4; % mm stator slot width
ws3=8.2; % mm stator slot width
ws4=1.8; % mm stator slot opening
%Rotor slot dimension

hrt=18.4; % mm total height of rotor slot
hr4=0.5; % mm rotor slot tip height
hr3=1.8;
hr1=16.1;
wr1=6.32;
wr3=7.48;
wr4=0.67; % mm rotor slot opening
mwp=1; % main winding position: 1 -near yoke, 2-near air-gap

layersm=2; % number of layers for main winding
layerse=1; % number of layers for excitation winding
ncme=4; % number of elementary conductors on main winding
dcme=1.18; % diameter of elementary conductors on main winding
sbm=18; % number of turns per coil on main winding
ncee=2; % number of elementary conductors on excitation winding
dcee=1.18; % diameter of elementary conductors on excitation winding
sbe=36; % number of turns per coil on excitation winding
cSpanm=0.77;
%Rotor data
kmrid=1; % non magnetic shaft
stcore_mat='M19'; % Magnetic material data for stator

rtbar_mat='Al'
I0n=3.64; % A magnetization pick current

i0_pu=1; % 0.2:0.2:1.4;
fname_psie='g3kW8p2_psie.txt';
fname_psim='g3kW8p2_psim.txt';
170 Appendix
Appendix 4.2
%post_proc
closeall
%clear all
fname='g1';
eval(fname,'fred_error=1');
poles1=poles/gcd(Nrs,poles);
me=load(fname_psie);
mm=load(fname_psim);
mBm=load('g1_Bm1p0.txt');
mBe=load('g1_Be1p0.txt');
ns=length(mBe);
tetae=poles1*180*mBe(:,1)/mBe(ns,1);
I0_tab=me(:,1);
psi_ie=me(:,7:12);
psi_im=mm(:,7:12);
psidee=me(:,13);
psiqme=-mm(:,14);
psiqem=me(:,16);
psiqmm=mm(:,16);
Le1=me(:,17);
Led=me(:,18);
Lem=me(:,19);
Lm1=mm(:,17);
Lmq=mm(:,18);
Lme=-mm(:,19);
dswa_fig;
figure
plot(tetae,mBe(:,2),tetae,mBm(:,2))
legend('Ie','Im');
xlabel('Angle (deg)');
ylabel('Flux density (T)');
grid
title('Air-gap flux density');
Bmh=2*fft(mBm(:,2))/ns;
Beh=2*fft(mBe(:,2))/ns;
hord=0:2/poles1:2*(ns-1)/poles1;
figure
bar(hord',[abs(Beh),abs(Bmh)]);
legend('Ie','Im');
xlabel('Harmonic order');
ylabel('Flux density magnitude (T)');
title('Flux density - space harmonics');
grid
Appendix 4.3 – The magnetization inductance approximation
me=load(fname_psie);
I0_tab=me(:,1);
Lem=me(:,19);
psidee=me(:,13);
Lm0=0.392;
Ltmin=0;
a1=8.2;
b1=3.751;
b2=1/6.5^3;
p1=[a1*b2 0 1]*(Lm0-Ltmin);
p2=[b2 b1*b2 0 1];
p3=[p1 0];
pd1=polyder(p3);
pd2=polyder(p2);
p4=conv(pd1,p2)-conv(pd2,p3);
x=0.1:0.1:30;
ixt=I0_tab(1:end-1)+diff(I0_tab)/2;
Lt=diff(Lem)./diff(I0_tab).*ixt+interp1(I0_tab,Lem,ixt);
fx=polyval(p1,x)./polyval(p2,x)+Ltmin;
gx=polyval(p4,x)./(polyval(p2,x).^2)+Ltmin;
figure
plot(I0_tab,Lem,'o',x,fx,x,gx,ixt,Lt,'*')
grid
xlabel('Magnetization current Im (A)');
ylabel('Inductance Lm (H)');
172 Appendix
Appendix 4.4 – The “inductance.m” script
closeall
clearall
ShowCurveReadings=0;
f_input=input('Input data - file name ','s');

eval(f_input);
m1=load(imput_table);
i_tab=m1(:,1);
Lm_tab=m1(:,2);
L_dtab=diff(Lm_tab)./diff(i_tab);
i_dtab=(i_tab(1:end-1)+i_tab(2:end))/2;
L_tranztab=L_dtab.*i_dtab+spline(i_tab,Lm_tab,i_dtab);
f_tab=i_tab.*Lm_tab;
L_tranzf2=diff(f_tab)./diff(i_tab);
ifrun_mode=='h'
set_hg;
optim; % optimization routine
else
set_ga;
gaOptim;
l_opt=Xbest;
end
desen; %show the output graphics
Example of input file script
%Data In
n1=3;
%minimal values
Lm0_min=0.37;
a1_min=0;
b1_min=0;
b2_min=1.7;
Lsc_min=0;
%maxim values
Lm0_max=0.528;
a1_max=2;
b1_max=2;
b2_max=18;
Lsc_max=0.042682;
%initial values
Lm0=0.281;
a1=0.6;
b1=0.6;
b2=9;
Lsc=0.02134
Var_low=[Lm0_min,a1_min,b1_min,b2_min,Lsc_min];
Var_up=[Lm0_max,a1_max,b1_max,b2_max,Lsc_max];
nstp=2;
%Initial values for optimization variable (is used only for HJ)
Var0=[Lm0,a1,b1,b2,Lsc];
Xc=Var0;
%Minimum variation
dXmin=[Lm0_min/1000,0.0001,0.0001,b2_min/1000,Lsc_max/1000];
k_emp=0.5; % minimum weight for quadratic error

k_emm=0.5; % maximum weight error
run_mode='h';
output_file='ind_c.m'; % Name of the output file
output_ev2best='ind1_ev2b1.txt';
imput_table='ind_tab.txt';
174 Appendix
Appendix 5 – Optimal parameter identification
closeall
clearall
output_file='par1.txt';
pp=4; % numbers of poles

fn=50; % rated frequency (the test frequency)
em=0.01; % assumed measurement error
ev=0.01; %voltage measurement error
erdc=0.15; %dc resistance measurement error including temperature variation
R1dc=3.9; %Ohm - dc resistance - first winding

R2dc=2.4; %Ohm - dc resistance - second winding
m1nL=3; % phase number at no load test

m1sc=2; % phase number at short circuit test
m3=load('.\\test\\test3.txt'); %Proba de gol real înfãsurareaprincipalãalimentatã de la

AT siexcitaþiaîngol
m4=load('.\\test\\test4.txt'); %Proba de gol real excitaþiaalimentatã de la AT
siînfãºurareaprincipalãîngol
m5=load('.\\test\\test5.txt'); %Proba de scurtcircuit, înfãsurareaprincipalãalimentatã de
la AT în 2 faze siexcitaþiaîngol
m6=load('.\\test\\test6.txt'); %Proba de scurtcircuit, înfãsurareaprincipalãalimentatã de
la AT în 2 faze siexcitaþiaînscurtcircuit
m7=load('.\\test\\test7.txt'); %Proba de scurtcircuit, excitaþiaalimentatã de la AT
siînfãºurareaprincipalãîngol
m8=load('.\\test\\test8.txt'); %iProba de scurtcircuit,excitaþiaalimentatã de la AT cu
tensiuneredusãsiînfãsurareaprincipalãînscurtcircuit
W0=2*pi*fn/pp;
Vm0=(m3(:,1)+m3(:,2)+m3(:,3))/3;
Vm02=2/3*sqrt(m3(:,1).^2+m3(:,2).^2+m3(:,3).^2-m3(:,1).*m3(:,2)-m3(:,1).*m3(:,3)-
m3(:,2).*m3(:,3));
Im0=(m3(:,4)+m3(:,5)+m3(:,6))/3;
Im02=2/3*sqrt(m3(:,4).^2+m3(:,5).^2+m3(:,6).^2-m3(:,4).*m3(:,5)-m3(:,4).*m3(:,6)-
m3(:,5).*m3(:,6));
Vem0=(m3(:,9)+m3(:,10)+m3(:,11))/3; %voltage in excitation, main is suplied
Vem02=2/3*sqrt(m3(:,9).^2+m3(:,10).^2+m3(:,11).^2-m3(:,9).*m3(:,10)-
m3(:,9).*m3(:,11)-m3(:,10).*m3(:,11));
Pm0=m3(:,7)*1000; %kW->W
Qm0=m3(:,8)*1000; %kW->W
Ve0=(m4(:,1)+m4(:,2)+m4(:,3))/3;
Ie0=(m4(:,4)+m4(:,5)+m4(:,6))/3;
Vme0=(m4(:,9)+m4(:,10)+m4(:,11))/3;
Ve02=2/3*sqrt(m4(:,1).^2+m4(:,2).^2+m4(:,3).^2-m4(:,1).*m4(:,2)-m4(:,1).*m4(:,3)-
m4(:,2).*m4(:,3));
Ie02=2/3*sqrt(m4(:,4).^2+m4(:,5).^2+m4(:,6).^2-m4(:,4).*m4(:,5)-m4(:,4).*m4(:,6)-
m4(:,5).*m4(:,6));
Vme02=2/3*sqrt(m4(:,9).^2+m4(:,10).^2+m4(:,11).^2-m4(:,9).*m4(:,10)-
m4(:,9).*m4(:,11)-m4(:,10).*m4(:,11));
Pe0=m4(:,7);
Qe0=m4(:,8);
Vmsc1=(m5(:,1)+m5(:,3))/2;
Imsc1=(m5(:,4)+m5(:,6))/2;
%Vemsc=(m5(:,9)+m5(:,10)+m5(:,11))/3; %voltage in excitation, main is supplied
Pmsc1=m5(:,7);
Qmsc1=m5(:,8);
Vmsc2=(m6(:,1)+m6(:,3))/2;
Imsc2=(m6(:,4)+m6(:,6))/2;
%Iemsc=(m5(:,9)+m5(:,10)+m5(:,11))/3; %voltage in excitation, main is supplied
Pmsc2=m6(:,7);
Qmsc2=m6(:,8);
Vesc1=(m7(:,1)+m7(:,3))/2;
Iesc1=(m7(:,4)+m7(:,6))/2;
%Vemsc=(m5(:,9)+m5(:,10)+m5(:,11))/3; %voltage in excitation, main is supplied
Pesc1=m7(:,7);
Qesc1=m7(:,8);
Vesc2=(m8(:,1)+m8(:,3))/2;
Iesc2=(m8(:,4)+m8(:,6))/2;
%Iemsc=(m5(:,9)+m5(:,10)+m5(:,11))/3; %voltage in excitation, main is supplied
Pesc2=m8(:,7);
Qesc2=m8(:,8);
em0=sqrt(Pm0.^2+Qm0.^2)./(3*Vm0.*Im0)-1;
ee0=sqrt(Pe0.^2+Qe0.^2)./(3*Ve0.*Ie0)-1;
176 Appendix
emsc1=sqrt(Pmsc1.^2+Qmsc1.^2)./(2*Vmsc1.*Imsc1)-1;
emsc2=sqrt(Pmsc2.^2+Qmsc2.^2)./(2*Vmsc2.*Imsc2)-1;
eesc1=sqrt(Pesc1.^2+Qesc1.^2)./(2*Vesc1.*Iesc1)-1;
eesc2=sqrt(Pesc2.^2+Qesc2.^2)./(2*Vesc2.*Iesc2)-1;
sem0=std(em0);
see0=std(ee0);
semsc1=std(emsc1);
semsc2=std(emsc2);
seesc1=std(eesc1);
seesc2=std(eesc2);
mem0=mean(em0);
mee0=mean(ee0);
memsc1=mean(emsc1);
memsc2=mean(emsc2);
meesc1=mean(eesc1);
meesc2=mean(eesc2);
wm0=exp(-(em0.^2)/(sem0^2+mem0^2+em^2));
we0=exp(-(ee0.^2)/(see0^2+mee0^2+em^2));
wmsc1=exp(-(emsc1.^2)/(semsc1^2+memsc1^2+em^2));
wmsc2=exp(-(emsc2.^2)/(semsc2^2+memsc2^2+em^2));
wesc1=exp(-(eesc1.^2)/(seesc1^2+meesc1^2+em^2));
wesc2=exp(-(eesc2.^2)/(seesc2^2+meesc2^2+em^2));
imv0=find(abs(em0)<=abs(mem0)+sem0); %valid data

imvsc1=find(abs(emsc1)<=abs(memsc1)+semsc1); %valid data
imvsc2=find(abs(emsc2)<=abs(memsc2)+semsc2);
iev0=find(abs(ee0)<=abs(mee0)+see0); %valid data
ievsc1=find(abs(eesc1)<=abs(meesc1)+seesc1); %valid data
ievsc2=find(abs(eesc2)<=abs(meesc2)+seesc2);
x
nm0=length(imv0);
ne0=length(iev0);
Vm0c=Vm0(imv0);
Vem0c=Vem0(imv0);
Im0c=Im0(imv0);
Pm0c=Pm0(imv0);
Qm0c=Qm0(imv0);
wm0c=wm0(imv0);
Ve0c=Ve0(iev0);
Vme0c=Vme0(iev0);
Ie0c=Ie0(iev0);
Pe0c=Pe0(iev0);
Qe0c=Qe0(iev0);
we0c=we0(iev0);
Vmsc1c=Vmsc1(imvsc1);
Imsc1c=Imsc1(imvsc1);
Pmsc1c=Pmsc1(imvsc1);
Qmsc1c=Qmsc1(imvsc1);
wmsc1c=wmsc1(imvsc1);
Vmsc2c=Vmsc2(imvsc2);
Imsc2c=Imsc2(imvsc2);
Pmsc2c=Pmsc2(imvsc2);
Qmsc2c=Qmsc2(imvsc2);
wmsc2c=wmsc2(imvsc2);
Vesc1c=Vesc1(ievsc1);
Iesc1c=Iesc1(ievsc1);
Pesc1c=Pesc1(ievsc1);
Qesc1c=Qesc1(ievsc1);
wesc1c=wesc1(ievsc1);
Vesc2c=Vesc2(ievsc2);
Iesc2c=Iesc2(ievsc2);
Pesc2c=Pesc2(ievsc2);
Qesc2c=Qesc2(ievsc2);
wesc2c=wesc2(ievsc2);
ke1=Vm0c./Vem0c;
ke2=Vme0c./Ve0c;
kemax=min(Vm0c./Vem0c)*(1+2*ev);
kemin=max(Vme0c./Ve0c)*(1-2*ev);
R1max=R1dc*(1+erdc);
R2max=R2dc*(1+erdc);
R1min=R1dc*(1-erdc);
R2min=R2dc*(1-erdc);
R10=3*Vm0.^2./Pm0;
Z10=Vm0c./Im0c;
X10=Qm0c./(3*Im0c.^2);
R1sc1=Pmsc1c./(2*Imsc1c.^2);
R2sc1=Pesc1c./(2*Iesc1c.^2);
R1sc2=Pmsc2c./(2*Imsc2c.^2);
178 Appendix
R2sc2=Pesc2c./(2*Iesc2c.^2);
Rrmax=max(R1sc1-R1min);
Rrmin=max(min(R1sc1)-R1max,0);
Rmmax=1/(1/max(R10)- min(Pm0(imv0)/3-
R1min*Im0(imv0).^2)/max(Vm0(imv0))^2)*(1+2*ev);
Rmmin=max(R10)*(1-2*ev)-R1max;
Z1sc1=Vmsc1(imvsc1)./Imsc1(imvsc1);
X1sc1=sqrt(Z1sc1.^2-R1sc1.^2);
Z2sc1=Vesc1c./Iesc1c;
Z1sc2=Vmsc2c./Imsc2c;
Z2sc2=Vesc2(ievsc2)./Iesc2(ievsc2);
X1max=min(X1sc2);
X2max=min(X2sc2);
X12max=max(X1sc1)-min(X1sc2);
Xrmax=X12max;
X1min=X1max/10;
X2min=X2max/10;
X12min=0;
Xrmin=0;
Xmmax=max(X10)*(1+2*ev);
Xmmin=max(X10)-X1max-X12max;
k0max=min(Pm0-3*R1min*Im0.^2)/W0;
k1max=k0max/W0;
k2max=k1max/W0;
k0min=0;
k1min=0;
k2min=0;
pt_fig;
set_ga;
par_gaOptim;
par_grafic
pt_fig2;
save_param;
toc
REFERENCES
[1] Wang D., Ma W., Xiao F., Zhang B., Liu D., and Hu A.: A novel stand-alone
dual stator-winding induction generator with static excitation regulation, IEEE
Transactions on Energy Conversion, Vol. 20, No. 4, December 2005, pp. 826-
835
[2] Alfredo R., Lipo M.T.: Dual stator winding induction machine drive, IEEE
Transactions on Industry Applications, Vol. 36, No. 5, September/October 2000,
pp. 1369-1379
[3] Naidu M., Walters J.: A 4-KW 42-V Induction Machine-based automotive
power generation system with a diode bridge rectifier and a PWM inverter, IEEE
Transactions on Industry Applications, Vol. 39, No. 5, September/October 2003,
pp. 1287-1293
[4] Tamrakar I., Malik O.P.: Power factor correction of induction motors using
PWM inverter fed auxiliary stator winding, IEEE Transactions on Energy
Conversion, Vol. 14, No. 3, September 1999, pp. 426-432
[5] Mirecki A., Roboam X., Richardeau F.: Architecture complexity and energy
efficiency of small wind turbines, IEEE Transactions on Industrial Electronics,
Vol. 54, No. 1, February 2007
[6] Chen Z., Guerrero J.M. , Blaabjerg F.: Review of the State of the Art of Power
Electronics for Wind Turbines, IEEE Transactions on Power Electronics, Vol.
24, No. 8, August 2009
[7] Dehghan S.M., Mohamadian M., Varjani A.Y.: A New Variable-Speed Wind
Energy Conversion System using Permanent-Magnet Synchronous Generator
and Z-Source Inverter, IEEE Transactions on Energy Conversion, Vol. 24, No.
3, September 2009
[8] Fukami T., Momiyama M., Shima K., Hanaoka R., and Takata S.: Steady-State
Analysis of a Dual-Winding Reluctance Generator With a Multiple-Barrier
Rotor, IEEE Transactions on Energy Conversion, Vol. 23, No. 2, June 2008, pp.
492-498
[9] Ojo O., Davidson I.E.: PWM-VSI Inverter-Assisted Stand-Alone Dual Stator
Winding Induction Generator, IEEE Transactions on Industry Applications, Vol.
36, No.6, November/December 2000, pp. 1604-1610
[10] Ojo O., Wu Z.: Modeling of a dual-Stator-Winding Induction Machine
Including the Effect of Main Flux Linkage Magnetic Saturation, IEEE
Transactions on Industry Applications, Vol. 44, No.4, July/August 2008, pp.
1099-1107
[11] Li Y., Hu Y., Huang W., Liu L., Zhang Y.: The Capacity Optimization for the
Static Excitation Controller of the Dual-Stator-Winding Induction Generator
180 References
Operating in a Wide Speed Range, IEEE Transactions on Industrial Electronics,
Vol. 56, No. 2, February 2009, pp. 530-541
[12] Roberts P.C., McMahon R.A., Wang X., Tavner P.J., Maciejowski J.M., Flack
T.J.: Equivalent circuit for the brushless doubly-fed machine (BDFM) including
parameter estimation, Proc. Inst. Electr. Eng.-Electr. Power Appl., vol. 152, no.
4, pp. 933-942, Jul. 2005
[13] Williamson S., Ferreira A.C., Wallace A.K.: Generalized theory of the
brushless doubly-fed machine-Part 1: Analysis, Part 2: Model verification and
performance, Proc. IEE-Elect. Power Applict., vol. 144, no. 2, pp. 111-121, pp.
123-129, Mar. 1997
[14] Boldea I.: Variable Speed Generators, CRC Press, Boca Raton, FL USA, 2006
[15] Budisan N.: Problems of induction generator systems, Politehnica Publishing
House, Timisoara, Romania, 2003
[16] Tutelea L., and Boldea I.: Optimal Design of Residential Brushless D.C.
Permanent Magnet Motors with FEM Validation, Proceedings of ACEMP, 2007,
pp. 435-439.
[17] R. Muñoz and T. A. Lipo, “Complex vector model of the squirrel cage
induction machine including instantaneous rotor bar currents,” IEEE Trans. Ind.
Appl., vol. 35, no. 6, pp. 1332–1340, Nov./Dec. 1999.
[18] R. Li, R. Spee, A. K. Wallace, and G. C. Alexander, “Synchronous drive
performance of brushless doubly-fed motors,” IEEE Trans. Ind. Appl., vol. 30,
no. 4, pp. 963–970, Jul./Aug. 1994.
[19] Seyed Mohammad Dehghan, Mustafa Mohamadian, and Ali Yazdian Varjani,”
A New Variable-Speed Wind Energy Conversion System using Permanent-
Magnet Synchronous Generator and Z-Source Inverter”, IEEE Transactions on
Energy Conversion, Vol. 24,No. 3, September 2009.
[20] B. Marungsri, N. Meeboon, A. Oonsivilai, Dynamic Model Identification of
Induction Motors using Intelligent Search Techniques with taking Core Loss
into Account Proceedings of the 6th WSEAS International Conference on Power
Systems, Lisbon, Portugal, September 22-24, 2006, pp. 108-115.
[21] M. Vasudevan, R. Arumugam, G. Venkatesan, and S. Paramasivam,
Estimation of stator resistance of induction motor for direct torque control,
WSEAS TRANSACTIONS on CIRCUITS and SYSTEMS Issue 2, Volume 2,
April 2003, pp. 327-332.
[22] R. Mittal, K.S. Sandhu, D.K. Jain, Controlled Operation of Variable Speed
Driven Permanent Magnet Synchronous Generator, WSEAS Transactions on
Systems, ISSN 1109-2777, Vol.8, Issue 2, 2009, pp. 189-199.
[23] B., Singh, Induction generator-a prospective’, Electric Machines and Power
Systems, Vol. 23, 1995, pp. 163-177.
[24] D.S. Zinger, and E. Muljadi, Annualized Wind Energy Improvement Using
Variable Speeds, IEEE Trans. on Industry Applications, Vol. 33, No.6, Nov/Dec
1997, pp. 1444-1447.
[25] R.C., Bansal, T.S., Bhatti, and D.P., Kothari, A bibliographical survey on
induction generators for application of non-conventional energy systems’, IEEE
Trans. Energy Conversion, Vol. 18, No. 3, 2003, pp. 433-439.
[26] K.S., Sandhu, and J., Dheeraj, A Simple Approach to Estimate the Steady-
State Performance of Self-Excited Induction Generator”, WSEAS Transactions
on Systems and Control, issue 3, volume 3, March 2008, pp 208-218.
[27] S., Sharma, and K.S., Sandhu, “Role of Reactive Power Source on Power
Quality of Three-phase Self-Excited Induction Generator”, WSEAS
Transactions on Power Systems, issue 4, volume 3, April 2008, pp. 216-225.
[28] Boldea, I., Nassar, S.A., The Induction Machine Handbook, CRC Press Boca
Raton FL USA, 2001.
[29] Madescu, G., Boldea, I., Miller, T.J.E., An Analytical Iterative Model (AIM)
for Induction Motor Design, Rec. of IEEE – IAS, Annual Meeting, 1996, vol. 1,
pp. 566 – 573.
[30] Chen, W.L., and Xie, C.Z., Active Voltage and Frequency Regulator Design
for a Wind Driven Induction Generator to Alleviate Transient Impacts on Power
Grid, IEEE Trans. on Ind. Electron., vol. 60, no. 8, August 2013, pp. 3165-
3175.
[31] Z. Wu, “An Investigation of Dual Stator Winding Induction Machines”, Phd
Dissertation, Tennessee Technological University, USA, December, 2006.
[32] Z. Wu, O. Ojo, “Coupled circuit model simulation and air gap field calculation
of a dual stator winding induction machine,” IEEE Proceedings-Electric Power
Applications, Vol. 153, No. 3, May 2006, pp. 387 – 400.
[33] P.C. Roberts, R.A. McMahon, P.J. Tavner, J.M. Maciejowski and T.J. Flack,
“Equivalent circuit for the brushless doubly fed machine (BDFM) including
parameter estimation and experimental verification”, IEE Proc. Electr. Power
Appl. Vol 152, No. 4, 2005 pp.933-942.
[34] C.B. Jacobina, A.M.N. Lima and A.L.R. de Medeiros, “Linear parameter
estimation of an induction machine”, Power Electronics Specialists Conference,
PESC '93 Record, 24th Annual IEEE Meeting, 1993, DOI
10.1109/PESC.1993.472030, pp. 913-918.
[35] S. Peresada, S. Kovbasa, D. Prystupa and S.E. Lyshevski, “Identification of
induction motor parameters adaptively controlling stator currents”, SICE
Proceedings of the 41st SICE Annual Conference, vol. 4, DOI
10.1109/SICE.2002.1195791, 2002, pp.2430-2434.
[36] S. Bachir, S. Tnani, J.C. Trigeassou and G. Champenois, “Diagnosis by
parameter estimation of stator and rotor faults occurring in induction machines”,
Industrial Electronics, IEEE Transactions on, vol. 53, issue 3, DOI
10.1109/TIE.2006.874258, 2006, pp. 963-973.
[37] A. Raie, V. Raschtchi, “Accurate identification of parameters, in winding
function model of induction motor, using genetic algorithm”, SICE Proc. 2002
Aug. 5-7, Osaka, pp. 2430-2434.
[38] J. A. Riveros Yepes, A. G. Barrero, F. Doval-Gandoy, J. Bogado, B. Lopez, O.
Jones, M. and E. Levi, “Parameter Identification of Multiphase Induction
182 References
Machines With Distributed Windings - Part 2”, Time-Domain Techniques, IEEE
Trans. on Energy Conversion, Vol. 27, N4. 4, December 2012, pp. 1067-1077.
[39] Tutelea, L.N., Deaconu, S.I., Boldea, I., and Budişan, N., “Design, control and
2D-FEM validation for an Double Stator Winding Induction Generator”, IECON
2013, November 10-13, Vienna, Austria, pp. 2732-2737.
[40] Vas, P., “Sensor less Vector and Direct Torque Control”, Oxford, U.K.:
Oxford Univ. Press, 1998.
[41] Jabri, A.K.A., and Alolah, A.I., “Limits on the performance of the three-phase
self-excited induction generators,” IEEE Transactions on Energy Conversion,
vol. 5, no. 2, Jun. 1990, pp. 350–356.
[42] Muljadi, E., and Lipo, T.A., “Series compensated PWM inverter with battery
supply applied to an isolated induction generator,” IEEE Transactions on
Industry Applications, vol. 30, no. 4, Jul./Aug. 1994, pp. 1073–1082.
[43] Bu, F., Huang, W., Hu, Y., and Shi, K. "An Integrated AC and DC Hybrid
Generation System Using Dual-Stator-Winding Induction Generator With Static
Excitation Controller," IEEE Transactions on Energy Conversion, vol. 27, no. 3,
Sept. 2012, pp. 810-812.
[44] Che, H.S., Levi, E., Jones, M., Duran, M.J., Hew, W.P., Rahim, N.”Operation
of a Six-Phase Induction Machine Using Series-Connected Machine-Side
Converters,” IEEE Transactions on Industrial Electronics, vol. 61, no. 1, Jan.
2014, pp. 164-176.
[45] Tutelea, L.N., Deaconu, S.I., Budişan, N. and Boldea, I.,”Double Staor
Winding Induction Generator for Wind and Hydro Applications: 2D-FEM
Analysis and Optimal Design”, 15th European Conference on Power Electronics
and Application, EPE ECCE Europe, 3-5 September 2013, Lille, France, 10 pp.
[46] Nouali, S., and Ouali, A.,”Multi-Layer neural network for sensorless MPPT
control for Wind Energy Conversion System using Doubly Fed Twin Stator
Induction Generator”, Systems Signal and Devices, 8th International Conference
on, 2011, Sousse, ISBN 978-1-4577-0413-0, pp. 7.
[47] Madescu, G., Trica, A., Budisan, N., Prostean, O., Biriescu, M., and Mot, M.,
”Performance optimization of low-speed induction generators for direct drive
wind turbines”, Electrical Power Conference, Montreal, Canada, October 2007,
ISBN 978-1-4244-1444-4, pp. 166-171.
[48] Panda, D., and Lipo, T.A., “Reduced switch count double converter fed wound
rotor induction machine drive for wind energy application”, Electric Machines
and Drives Conference, June 2003, vol.3, ISBN 0-7803-7817-2, pp. 1924-1931.
[49] Chitti Babu, B., Mohanty, K.B., and Poongothai, C., “Performance of
Double-Output Induction Generator for Wind Energy Conversion Systems”,
Emerging Trends in Engineering and Technology, July 2008, Nagpur,
Maharashtra, India, ISBN 978-0-7695-3267-7, pp. 933-938.
[50] Bu, F., Huang, W., Hu, Y., Shi, J., and Shi, K., “A Stand-Alone Dual Stator-
Winding Induction Generator Variable Frequency AC Power System,” IEEE
Transactions on Power Electronics, vol. 27, no. 1, Jan. 2012, pp. 10-13.
[51] Siegfried Heier, "Grid Integration of Wind Energy Conversion Systems," John
Wiley & Sons Ltd, 1998, ISBN 0-471-97143-X.
[52] Alejandro Rolan', Alvaro Luna, Gerardo Vazquez,Daniel Aguilar, Gustavo
Azevedo Modeling of a Variable Speed Wind Turbine with a Permanent Magnet
Synchronous Generator, IEEE International Symposium on Industrial
Electronics (ISlE 2009) Seoul Olympic Parktel, Seoul, Korea July 5-8, 2009.
[53] Bingchang Ni, Constantinos Sourkounis, Energy Yield and Power Fluctuation
of Different Control Methods for Wind Energy Converters, IEEE Transactions
on Industry Applications, vol. 47, no. 3, May / June 2011, pp.1480-1486.
[54] P. L. Alger, E. H. Freiburghouse, and D. D. Chase, “Double windings for
turbine alternators,” AIEE Trans., vol. 49, pp. 226–244, Jan. 1930.
[55] Bu, F., Huang, W., Hu, Y., and Shi, K., “An Excitation-Capacitor-Optimized
Dual Stator-Winding Induction Generator with the Static Excitation Controller
for Wind Power Application” IEEE Transaction on Energy Conversion, vol. 26,
no. 1, March. 2011, pp. 122-131.
[56] O. Ojo, I.E., Davidson, A dual stator winding induction generator with a four
switch inverter-battery scheme for control, Power Electronics Specialists
Conference, 2000, PESC 00, Vol. 1, pp. 230-234.
[57] Li Yong, Hu Yuwen, Liu Lingshun, Huang Wenxin, Chen Guanghui, Control
of a Novel Dual Stator-Winding Induction Generator for wide speed-range
operation, Power Electronics Specialists Conference, PESC 2007, pp. 2096 –
2101.
[58] D.M. Ionel, M.V. Cistelecan, T.J.E. Miller, M.I. McGilp, A New Analytical
Method for the Computation of Air-gap Reactance’s in 3 Phase Induction
Machines, Record of IEEE – IAS, Annual Meeting 1998, vol. 1, pp. 65 – 72.
[59] Hofman, H., Sanders, S.R., Sullivan, C.R., Stator-Flux-Based Vector Control
of Induction Machines in Magnetic Saturation, IEEE Transactions on Industry
Applications, vol.33, no.4, July/August 1997, pp. 935-942.
[60] Holz, J., Hangwen, P., Elimination of saturation Effects in Sensorless Position
Controlled Induction Motors, IEEE IAS Annual Meeting, Pittsburg, USA, 2002,
pp. 8.
[61] Melkebeek, J.A., Novotny, D.W., The Influence of Saturation on Induction
Machine Drive Dynamics, IEEE Trans. on Power Apparatus and Systems, vol.
103, no. 4, April 1984, pp. 707-714.
[62] Levi, E., Main Flux Saturation Modelling in Double-Cage and Deep-bar
Induction Machines, IEEE Transactions on Energy Conversion, vol.11, no. 2,
1996, pp. 305-311.
[63] Ojo, J.O., Consoli, A., Lipo, T.A., An Improved Model of Saturated Induction
Machines, IEEE Transactions on Industry Applications, vol.26, no.2
March/April, 1990, pp. 212-221.
[64] Wamkeue, R.E., Aguglia, D., Lakehal, M., Viarouge, P., Two-step Method for
Identification of non-linear Model of Induction Machine, IEEE Transaction on
Energy Conversion, vol. 22, 2007, pp. 801-809.
184 References
[65] T.A. Lipo, Introduction to AC Machine Design, vol.1, pp.84, WEMPEC –
University of Wisconsin, 1996.
[66] V. Ostovic, Dynamics of Saturated Electric Machines, Springer Verlag, New
York, 1989.
[67] Li Yong, Hu Yuwen, Huang Wenxin, Zhang Yong, Hao Zhenyang, Liu
Lingshun, Decoupling control of the Dual Stator-Winding Induction Generator
using SVM, Power Electronics Specialists Conference, PESC 2008, pp. 3366 –
3370.
[68] D. Uygun, G. Bal, and D. Sefa, “Linear model of a Novel 5-Phase Segment
Type Switched Reluctance Motor”, ELEKTRONIKA IR
ELEKTROTECHNIKA, ISSN 1392-1215, vol. 20, no.1, 2014, pp. 3-7.
[69] B. Kundrotas, A. Petrovas, R. Rinkeviciene, and A. Smilgevicius, “Research of
Six-Phase Induction Motor Windings”, ELEKTRONIKA IR
ELEKTROTECHNIKA, ISSN 1392-1215, vol. 20, no.1, 2014, pp. 15-18.
[70] H. Xiao, and F. Liu, “A Control Strategy of DFIG under Unbalanced Grid
Voltage”, ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN 1392-1215, vol.
19, no.9, 2013, pp. 15-20.
[71] M. H. Aliabadi, S.H. Hosseinian, S. J. Moghani, and M. Abedi “Multisided
Linear Induction Generator, Analytical Modeling, 3D-Finite Element Analysis
and Experimental Test ”, ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN
1392-1215, vol. 19, no.8, 2013, pp. 8-14.
[72] S. Lisauskas, D. Udris, and D. Uznys “Direct Torque Control of Induction
Drive Using Fuzzy Controller”, ELEKTRONIKA IR ELEKTROTECHNIKA,
ISSN 1392-1215, vol. 19, no.5, 2013, pp. 13-16.
[73] J. Buksnaitis, “Electromagnetical Efficiency of the Six-phase Winding”,
ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN 1392-1215, vol. 18, no. 3,
2012, pp. 3-6.
[74] Bu F, Hu Y, Huang W and Zhuang S, Parameter design and static performance
of Dual Stator-Winding Induction Generator Variable Frequency AC Generating
System with Capacitive and Inductive Loads, IEEE Transactions on Industrial
Electronics, ISSN 0278-0046, vol. 61, issue 8, 2014, pp. 3902-3914.
[75] Bu F, Hu Y, Huang W , Zhuang S and Shi K, Control Strategy and Dynamic
Performance of Dual Stator-Winding Induction Generator Variable Frequency
AC Generating System with Capacitive and Inductive Loads, IEEE Transactions
on Power Electronics, ISSN 0885-8993, vol. 29, issue 4, 2014, pp. 1681-1692.
[76] Bu F, Hu Y, Huang W, Zhuang S and Shi K, Wide-Speed-Range-Operation
Dual Stator-Winding Induction Generator DC Generating System for Wind
Power Applications, IEEE Transactions on Power Electronics, ISSN 0885-8993,
vol. 30, issue 2, 2015, pp. 561-573.
[77] Nian H and Song Y, Direct power control of doubly fed induction generator
under distorted grid voltage, IEEE Trans. Power Electron., vol. 29, no. 2,
2014, pp. 894 -905.
[78] Huang W, Hu Y, Bu F and K Shi, The magneto motive force of a novel dual
stator-winding induction generator, Proc. IEEE Energy Convers. Cong.
Expo., 2010, pp.3582 -3587.
[79] Levi E, Impact of iron loss on behaviour of vector controlled induction
machines, IEEE Trans. Ind. Appl., vol. 31, no. 6, 1995, pp. 1287 -1296.
[80] Tutelea L N, Deaconu S I, Popa G N, Reduced Cost Low Speed Wind or
Hydro Energy Conversion System with Twin Stator Windings Induction
Generators, 16th International Conference onPower Electronics and Motion
Control Conference and Exposition, PEMC, Antalya, Turcia, 21-24 September,
pp. 317-324.
[81] Tutelea L N, Deaconu S I, Popa G N, Control System for Producing Electricity
with Dual Stator Winding Cage-Rotor Induction Generator, National Electric
Drives Conference no. 17, October 9-10 2014, Reşiţa, România, Anals of the
Eftimie Murgu University, year XXI, no. 2 , 2014, vol.I, pp. 323-334.
[82] Tutelea L N, Boldea I, Muntean N, Deaconu S I, Modeling and Performance of
Novel Scheme Dual Winding Cage Rotor Variable Speed Induction Generator
with dc Link Power Delivery, IEEE ENERGY CONVERSION CONGRESS &
EXPOSITION, ECCE 2014, September 14-18, Pittsburgh, USA, pp. 271-278.
[83] Tutelea L N, Boldea I, Deaconu S I, Parameter optimal identification of dual
three phases stator winding induction machine, 14th International Conference on
Optimization of Electrical and Electronic Equipment OPTIM 2014, Braşov, 22-
24 May, pp. 231-238.
[84] Tutelea L N, Deaconu S I, Popa G N, Dual stator winding variable speed
asynchronous generator: optimal design and experiments, IOP Conf. Series:
Materials Science and Engineering 85 (2015) 012010, ISSN: 1757-899X, pp. 1-
9.
[85] Tutelea L N, Muntean N, Deaconu S I, Cunţan C D, Dual stator winding
variable speed asynchronous generator: magnetic equivalent circuit with
saturation, FEM analysis and experiments, IOP Conf. Series: Materials Science
and Engineering 106 (2016) 012029 ISSN: 1757-899X, pp. 1-10.
[86] Martin A D, Vitan D L, Tutelea L N, Muntean N, Double Stator Winding
Induction Generator for Renewable Energy Conversion System, Journal of
Electrical Engineering, vol. 15, no. 3, pp. 1-8.
[87] K. K. Mohapatra et al., “Independent field-oriented control of two split phase
induction motors from a single six-phase inverter,” IEEE Trans. Ind. Electron.,
vol. 52, no. 5, Oct. 2005, pp. 1372–1382.
[88] E. A. Klingshirn, “Harmonic filters for six-phase and other multi phase motors
on voltage source inverter,” IEEE Trans. Ind. Appl., vol. 1A-21, no. 4, May/Jun.
1985, pp. 588–593.
[89] Y. Zhao and T. A. Lipo, “Space vector PWM control of dual three-phase
induction machine using space vector decomposition,” IEEE. Trans. Ind. Appl.,
vol. 31, no. 5, Sep./Oct. 1995, pp. 1100–1109.
[90] L.H. Hansen, L. Hellen, F. Blaabjerg, E. Rithcie, S. Munk-Nielsen, H.
Bindner, P. Sørensen and B. Bak-Jensen, “Conceptual survey of Generators and
186 References
Power Electronics for Wind Turbines”, Risø-R-1205(EN), Risø National
Laboratory, Roskilden, Denmark, December 2001.
[91] S. Basak, C. Chakraborty, “Dual Stator Winding Induction Machine:
Problems, Progress, and Future Scope”, IEEE Trans. on Ind. Electronics, vol.62,
no.7, July 2015, pp. 4641-4652.
[92] T. Long, S. Shao, P. Malliband, E. Abdi, and R. A. McMahon, “Crowbarless
fault ride-through of the brushless doubly fed induction generator in a wind
turbine under symmetrical voltage dips,” IEEE Trans. Ind. Electron., vol. 60, no.
7, Jul. 2013, pp. 2833–2841.
[93] P. L. Alger, E. H. Freiburghouse, and D. D. Chase, “Double windings for
turbine alternators”, AIEE Trans., vol.49, Jan. 1930, pp. 226-244.
[94] I. Boldea, L. Tutelea, “Electrical machines – Steady State, Transients, and
Design with MATLAB”, CRC Press, 2010.
[95] David Meeker, Finite Element Method Magnetics, Version 4.2, User’s Manual,
October 25, 2015, http://www.femm.info/wiki/Documentation/
[96] David Meeker, Finite Element Method Magnetics: Octave FEMM Version 1.2
User’s Manual October 25, 2015, http://www.femm.info/wiki/Documentation/
View publication stats

Carte Tutelea Muntean Deaconu

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Carte Tutelea Muntean Deaconu

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Dual Stator Winding Induction Generator for Wind or Hydro Applications

Book · October 2017

Sorin Ioan Deaconu Lucian Tutelea

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Chapter 2. Dual stator windings induction generators (DSWIG):

Chapter 3. Analytical model of DSWIG............................................................... 18

Chapter 4. Equivalent magnetic circuit considering the saturation................. 23

Chapter 5. Preliminary and optimal design……………………………………. 28

Chapter 6. Parameter optimal identificationn.................................................... 38

Chapter 7. 2D-FEM analysis……………………………………………………..45

Chapter 8. Control strategies, the DSWIG mathematical model and

Chapter 9. Experimental work………………………………………………….. 93

Chapter 10. Design and simulation software...…………………………………122

Chapter 11. Conclusions………………………………………………………….125

In the context of rapid technical progress, electrical drives have to ensure

The dual-stator-winding squirrel-cage induction machine is the most recent

There is a large variety of excitation schemes for conventional and DSWIG

Fig. 2 Induction machine power generation schemes

Fig. 4 Proposed solutions for unpretentious loads [82].

Equations defining the reported parameters are listed below:

where V 2 is phase auxiliary voltage, V’ 2 is phase auxiliary reported voltage, I’ 2 is

R’f L’f Z2i

Fig. 8 Reduced equivalent circuit of the figure 7 [82].

The corresponding equations are:

The solution of the equation (17) is [83]:

Fig. 9 Equivalent phase scheme of DSWA in no load probe [85].

V mo R m  jX σm   I1mo  V 'eo  k e  V eo , (27)

V msc R m  jXσm   I1msc  V'esc  k e  Vesc . (31)

B1 flat (teeth saturation)

Considering the air-gap distribution of the flux density sinusoidal, sometimes

Fig. 12 One pole geometry of the DSWIG.

Usually, the magnetization curvatures H = f (B) are given by the producer up

Table I. Design parameters

Table II Preliminary parameters and machine design dimensions

It presents data entered into the program optimization, optimization variables

%Primary Dimension, Technical request and Technological limitation

Pn=3; % KW -rated power

For ease of practical realization of the experimental model, the laminations

Fig. 14 Main dimensions of stator and rotor from optimal design.

Stator slot dimension

Fig. 15 Stator slot dimensions from optimal design [84].

Fig. 16 Rotor slot dimensions from optimal design [84].

The general optimization problem can be divided into three quasi-independent

Flux density (T)

Fig. 22 Current versus torque [45].

The equivalent phase circuit for DSWIG is presented in figure 23 [83].

Fig. 23 Equivalent phase circuit of the DSWIG [83].

A. The Circuit Model

An acceptable solution for the over determined system of equations, knowing

B. The Measured Errors Reduction

C. The Boundaries of the Candidate Parameters Vector

The boundaries of the candidate parameter vector are computed from

Rm min  max  R10 1  2 v   R1max , (101)

Fig. 25 Magnetic field lines resulting from finite element analysis.

Fig. 34 Highlighting the three-phase main stator winding in two layers.

Fig. 35 Highlighting of the three-phase excitation stator winding in one layer.

Fig. 36 Highlighting cage rotor winding.

No load (zero rotor current) FEM investigations at standstill, with dc currents

Linkage flux versus current

8.1 Mathematical model of the control

The complex variable (space-phasor) model of IM is [28]:

Fig. 49 Main, excitation, d-q and - winding diagram [39].