Power Systems & Smart Energies (PSE) is dedicated to the design, modeling, exploitation and diagnostics of electrica
347 101 5MB
English Pages 113 [127] Year 2017
Table of contents :
 Synthesis of a Power Gyrator Based on Sliding Mode Control of two Cascaded Boost Converters Using a Single Sliding Surface
 Evaluating Power Converters using a WindSystem Simulator
 On line improvement of power system dynamic stability using ANFIS and NSGA II algorithms
 A Sliding Mode Control Approach Applied to a Photovoltaic System operated in MPPT
 Comparison of the IGCT and IGBT for the Modular Multilevel Converter in HVDC Applications
 Modeling of Double Star Induction Machine Including Magnetic Saturation and Skin effect
 Increasing the Torque Density of PermanentMagnet Synchronous Machines using Innovative Materials and Winding Technologies
Faouzi Derbel, Nabil Derbel, Olfa Kanoun (Eds.) Power Systems & Smart Energies
Advances in Systems, Signals and Devices

Edited by Olfa Kanoun, University of Chemnitz, Germany
Volume 3
Power Systems & Smart Energies  Edited by Faouzi Derbel, Nabil Derbel, Olfa Kanoun
Editors of this Volume Prof. Dr.Ing. Faouzi Derbel Leipzig University of Applied Sciences Chair of Smart Diagnostic and Online Monitoring Wächterstrasse 13 04107 Leipzig, Germany [email protected]
Prof. Dr.Ing. Olfa Kanoun Technische Universität Chemnitz Chair of Measurement and Sensor Technology Reichenhainer Strasse 70 09126 Chemnitz [email protected]
Prof. Dr.Eng. Nabil Derbel University of Sfax Sfax National Engineering School Control & Energy Management Laboratory 1173 BP, 3038 SFAX, Tunisia [email protected]
ISBN 9783110446159 eISBN (PDF) 9783110448412 eISBN (EPUB) 9783110446289 SetISBN 9783110448429 ISSN 23657493 eISSN 23657507 Library of Congress CataloginginPublication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliograﬁe; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2017 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Konvertus, Haarlem Printing and binding: CPI books GmbH, Leck ♾ Printed on acidfree paper Printed in Germany www.degruyter.com
Preface of the Volume Editor The third volume of the Series “Advances in Systems, Signals and Devices” (ASSD), contains international scientiﬁc articles devoted to the ﬁeld of power systems and smart energies. The scope of the volume encompasses all aspects of research, development and applications of the science and technology in these ﬁelds. The topics concern energy systems and energy transmission, renewable energy systems, hybrid renewable energy systems, photovoltaic systems, solar energy, wind energy, energy storage, batteries, thermal energy, combined heat and thermal power generation, electric machine design, electric machines modelling and control, electrical vehicles, technologies for electro mobility, special machines, variable speed drives, variable speed generating systems, automotive electrical systems, monitoring and diagnostics, electromagnetic compatibility, power systems, transformers, power electronics, topologies and control of power electronic converters. These ﬁelds are addressed by a separate volume of the series. All volumes are edited by a special editorial board made up by renowned scientist from all over the world. Authors are encouraged to submit novel contributions which include results of research or experimental work discussing new developments in the ﬁeld of power systems and smart energies. The series can be also addressed for editing special issues for novel developments in speciﬁc ﬁelds. Guest editors are encouraged to make proposals to the editor in chief of the corresponding main ﬁeld. The aim of this international series is to promote the international scientiﬁc progress in the ﬁelds of systems, signals and devices. It provides at the same time an opportunity to be informed about interesting results that were reported during the international SSD conferences. It is a big pleasure of ours to work together with the international editorial board consisting of renowned scientists in the ﬁeld of power systems and smart energies. The Editors Faouzi Derbel, Nabil Derbel and Olfa Kanoun
Advances in Systems, Signals and Devices Series Editor: Prof. Dr.Ing. Olfa Kanoun Technische Universität Chemnitz, Germany. [email protected]
Editors in Chief: Systems, Automation & Control Prof. Dr.Eng. Nabil Derbel ENIS, University of Sfax, Tunisia [email protected]
Power Systems & Smart Energies Prof. Dr.Ing. Faouzi Derbel Leipzig Univ. of Applied Sciences, Germany [email protected]
Communication, Signal Processing & Information Technology Prof. Dr.Ing. Faouzi Derbel Leipzig Univ. of Applied Sciences, Germany [email protected]
Sensors, Circuits & Instrumentation Systems Prof. Dr.Ing. Olfa Kanoun Technische Universität Chemnitz, Germany [email protected]
Editorial Board Members: Systems, Automation & Control Dumitru Baleanu, Çankaya University, Ankara, Turkey Ridha Ben Abdennour, Engineering School of Gabès, Tunisia Naceur Benhadj, Braïek, ESSTT, Tunis, Tunisia Mohamed Benrejeb, Engineering School of Tunis, Tunisia Riccardo Caponetto, Universita’ degli Studi di Catania, Italy Yang Quan Chen, Utah State University, Logan, USA Mohamed Chtourou, Engineering School of Sfax, Tunisia Boutaïeb Dahhou, Univ. Paul Sabatier Toulouse, France Gérard Favier, Université de Nice, France Florin G. Filip, Romanian Academy Bucharest Romania Dorin Isoc, Tech. Univ. of Cluj Napoca, Romania Pierre Melchior, Université de Bordeaux, France Faïçal Mnif, Sultan qabous Univ. Muscat, Oman Ahmet B. Özgüler, Bilkent University, Bilkent, Turkey Manabu Sano, Hiroshima City Univ. Hiroshima, Japan AbdulWahid Saif, King Fahd University, Saudi Arabia José A. Tenreiro Machado, Engineering Institute of Porto, Portugal Alexander Pozniak, Instituto Politecniko, National Mexico Herbert Werner, Univ. of Technology, Hamburg, German Ronald R. Yager, Mach. Intelligence Inst. Iona College USA Blas M. Vinagre, Univ. of Extremadura, Badajos, Spain Lotﬁ Zadeh, Univ. of California, Berkeley, CA, USA
Power Systems & Smart Energies Sylvain Allano, Ecole Normale Sup. de Cachan, France Ibrahim Badran, Philadelphia Univ., Amman, Jordan Ronnie Belmans, University of Leuven, Belgium Frdéric Bouillault, University of Paris XI, France Pascal Brochet, Ecole Centrale de Lille, France Mohamed Elleuch, Tunis Engineering School, Tunisia Mohamed B. A. Kamoun, Sfax Engineering School, Tunisia Mohamed R. Mékidèche, University of Jijel, Algeria Bernard Multon, Ecole Normale Sup. Cachan, France Francesco Parasiliti, University of L’Aquila, Italy Manuel Pérez,Donsión, University of Vigo, Spain Michel Poloujadoff, University of Paris VI, France Francesco Profumo, Politecnico di Torino, Italy Alfred Rufer, Ecole Polytech. Lausanne, Switzerland Junji Tamura, Kitami Institute of Technology, Japan
Communication, Signal Processing & Information Technology Til Aach, Achen University, Germany Kasim AlAubidy, Philadelphia Univ., Amman, Jordan Adel Alimi, Engineering School of Sfax, Tunisia Najoua Benamara, Engineering School of Sousse, Tunisia Ridha Bouallegue, Engineering School of Sousse, Tunisia Dominique Dallet, ENSEIRB, Bordeaux, France Mohamed Deriche, King Fahd University, Saudi Arabia Khalifa Djemal, Université d’Evry, Val d’Essonne, France Daniela Dragomirescu, LAAS, CNRS, Toulouse, France Khalil Drira, LAAS, CNRS, Toulouse, France Noureddine Ellouze, Engineering School of Tunis, Tunisia Faouzi Ghorbel, ENSI, Tunis, Tunisia Karl Holger, University of Paderborn, Germany Berthold Lankl, Univ. Bundeswehr, München, Germany George Moschytz, ETH Zürich, Switzerland Radu PopescuZeletin, Fraunhofer Inst. Fokus, Berlin, Germany Basel Solimane, ENST, Bretagne, France Philippe Vanheeghe, Ecole Centrale de Lille France
Sensors, Circuits & Instrumentation Systems Ali Boukabache, Univ. Paul, Sabatier, Toulouse, France Georg Brasseur, Graz University of Technology, Austria Serge Demidenko, Monash University, Selangor, Malaysia Gerhard Fischerauer, Universität Bayreuth, Germany Patrick Garda, Univ. Pierre & Marie Curie, Paris, France P. M. B. Silva Girão, Inst. Superior Técnico, Lisboa, Portugal Voicu Groza, University of Ottawa, Ottawa, Canada Volker Hans, University of Essen, Germany Aimé Lay Ekuakille, Università degli Studi di Lecce, Italy Mourad Loulou, Engineering School of Sfax, Tunisia Mohamed Masmoudi, Engineering School of Sfax, Tunisia Subha Mukhopadhyay, Massey University Turitea, New Zealand Fernando Puente León, Technical Univ. of München, Germany Leonard Reindl, Inst. Mikrosystemtec., Freiburg Germany Pavel Ripka, Tech. Univ. Praha, Czech Republic Abdulmotaleb El Saddik, SITE, Univ. Ottawa, Ontario, Canada Gordon Silverman, Manhattan College Riverdale, NY, USA Rached Tourki, Faculty of Sciences, Monastir, Tunisia Bernhard Zagar, Johannes Kepler Univ. of Linz, Austria
Contents Preface of the Volume Editor  V R. Haroun, A. El Aroudi, A. CidPastor and L. MartinezSalamero Synthesis of a Power Gyrator Based on Sliding Mode Control of two Cascaded Boost Converters Using a Single Sliding Surface  1 F. FloresBahamonde, H. ValderramaBlavi, J. A. Barrado Rodrigo, J. M. Bosque and A. LeonMasich Evaluating Power Converters using a WindSystem Simulator  19 A. Farah, T. Guesmi, H. Hadj Abdallah and A. Ouali On line improvement of power system dynamic stability using ANFIS and NSGA II algorithms  39 N. Khemiri, A. Khedher and F. Mimouni A Sliding Mode Control Approach Applied to a Photovoltaic System operated in MPPT  55 M. Buschendorf, J. Weber and S. Bernet Comparison of the IGCT and IGBT for the Modular Multilevel Converter in HVDC Applications  67 H. Kouki, M. Ben Fredj and H. Rehaoulia Modeling of Double Star Induction Machine Including Magnetic Saturation and Skin effect  83 M. Lindner, P. Braeuer and R. Werner Increasing the Torque Density of PermanentMagnet Synchronous Machines using Innovative Materials and Winding Technologies  97
R. Haroun, A. El Aroudi, A. CidPastor and L. MartinezSalamero
Synthesis of a Power Gyrator Based on Sliding Mode Control of two Cascaded Boost Converters Using a Single Sliding Surface Abstract: In this paper, a systematic method to synthesize a dc power gyrator, based on cascaded connection of two dcdc converters is introduced. The dc power gyrator is synthesized by means of a slidingmode control that imposes a proportionality between the input current of the ﬁrst stage and the output voltage of the second stage. Only one sliding surface is used to drive both switches of the system. The power gyrator based on cascaded connection of two dcdc converters can be used for obtaining high conversion ratio. A systematic procedure for designing the system is presented and its stability analysis is performed analytically and validated by numerical simulations using PSIM software. As an example of application, it is shown that a gyrator based on the cascade connection of two boost converters can be a good candidate for the impedance matching between a photovoltaic (PV) generator and a resistive load. Keywords: Power gyrator, Centralized sliding mode control, Impedance matching. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction The increase in energy sources demand and the related harmful greenhouse gases are topical problems which have prompted to make use of new clean renewable energy sources [1]. The existence of different nature of renewable energy sources (solar, wind, hydropower, etc...) is prompting to increase the use of distributed generation. New trends in the future distribution and generation system shows that the current centralized system will evolve to a new decentralized one in which the consumers will have a role of active stakeholders of the energy generation. In this new scenario, distributed generation systems based on a dc voltage bus seem to have a slight advantage over those based on an ac distributed bus [2]. The preference of a dc voltage bus instead of an ac distribution one is mainly due to the fact that most of the typical consumer loads are supplied in dc. Moreover, most of RES such as PV generators, fuel cells as well as storage batteries and super capacitors use also dc energy. In the dcbased nanogrid context, the future home
R. Haroun, A. El Aroudi, A. CidPastor and L. MartinezSalamero: Departament d’Enginyeria Electronica, Electrica i Automatica, Escola Tecnica Superior d’Enginyeria, Universitat Rovira i Virgili, Tarragona, Spain, emails: [email protected], [email protected], [email protected], [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 1–17. DOI 10.1515/9783110448412001
2  R. Haroun et al.
electric system is expected to have two dc voltage levels: a highvoltage dc (380 V) powering major home appliances and electric vehicle charging and a lowvoltage (48 V) for supplying computer loads, low power consumer electronics, lighting etc. [3–5]. Furthermore, the photovoltaic technologies have rapidly expanded during the last decade and it is foreseen that they will increase signiﬁcantly the proportion of solar energy use which can be considered as the most promising green energy of the new century due to its abundance [6, 7]. The major problem of energy generation from this important energy source is the optimal functioning of the PV panels. This optimization process is traditionally performed by using a Maximum Power Point Tracker (MPPT) [8].
ggyrator i1 Vg
+ −
i2 + v1 −
Boost dcdc converter
380 V Boost dcdc converter
+ v2 R −
g Fig. 1. Power gyrator based on two cascaded boost converters.
In addition, cascaded dcdc converters (Fig. 1) can be considered as an elemental building block for the future distribution generation systems. As an example, in [6], the authors examined the advantages, difficulties, and implementation issues of using a cascade connection of converters for a series string of PV panels. A typical problem in electrical power generation using PV systems is the high conversion ratio needed to obtain an output dc voltage about 400 V from a low dc voltage of the PV panel. The use of a single stage in performing this conversion ratio will imply working with high duty cycles and therefore will increase the losses which will jeopardize the system efficiency and will reduce the voltage conversion ratio. Therefore, cascaded stepup converters can be a good alternative in terms of efficiency, in order to obtain the desired output voltage [9]. The electrical architecture of power processing systems can be modeled by means of ideal canonical elements and analyzed using the laws governing the interconnection of twoport networks which belong to a class of ideal circuits named POPI (power output is equal to power input) [10]. In [9], the cascaded boostbased converters based gyrator has been synthesized using two sliding surfaces. In addition, the gyrator can be used for impedance matching between a PV panel and a battery as reported in [11]. In this paper, the gyrator based two cascaded boost converters will
Power Gyrator Based on SMC of two Cascaded Boost Converters

3
be used to step up the voltage of the PV panel to the load voltage (380 V). In order to facilitate the design procedure, the gyrator will be synthesized by SlidingMode Control (SMC) [12, 13]. This control method has many advantages for controlling these converters such as stability in front of variations of load line, robustness, good dynamic response and simple implementation. It will be shown that a simple strategy based on using only one switching surface for controlling both cascaded converters will be sufficient to achieve our goal. The rest of the paper is organized as follows: Section II gives an overview about the synthesis of power gyrators using SMC. Section III describes the system under study which consists of a power gyrator based on two cascaded boost converters. The ideal sliding dynamics, the mathematical continuoustime model, its equilibrium point and its stability analysis will be introduced in Section IV. The numerical simulations are performed in Section V. In Section VI, the system is used for impedance matching between a PV panel and a load resistance. Finally, some concluding remarks of this work are summarized in the last section.
2 Synthesis of power gyrators in SlidingMode The concept of power gyrator was introduced in [14, 15] where it was related to a general class of circuits named POPI (power output = power input) describing the ideal behavior of a switchedmode power converter as described in Fig. 2.
i1
i2
+
+
v1 (t)
v2(t)
−
− Gyrator
Fig. 2. Schematic of a power gyrator.
A power gyrator is a twoport structure characterized by any of the following two sets of equations I1 = gV2 , I2 = gV1
ggyrator
(1)
V1 = rI2 ,V2 = rI1
rgyrator
(2)
where I1 , I2 , V1 and V2 are the steady state values of the current and the voltage at the input and output ports respectively and g (r) is the gyrator conductance (resistance).
4  R. Haroun et al.
dcdc switching converter i1
i2
+ v1
+ v2
−
−
k1
HC s( x )
u 1
k2
s ( x ) = k1i1 − k2v2
Σ +
−
Fig. 3. Block diagram of a power ggyrator with controlled input current for power processing.
Equations (1) and (2) deﬁne the two types of gyrators. The power ggyrator can be synthesized as shown in the block diagram of Fig. 3. It consists of a switching converter controlled by means of a sliding mode regulation loop [12, 13], in which the switching manifold is the set = {xs(x) = 0}, where s(x) = k1 i1 + k2 v2 in such a way that, in steady state the following equation holds I1 = −(k2 /k1 )V2 = gV2 . Note that in this case the gyrator parameter is given by g = −k2 /k1 . Imposing a slidingmode regime requires that the input current i1 to be a continuous function of time, that implying the existence of a series inductor at the input port. It has to be pointed out that the presence of a Hysteretic Comparator (HC) in the feedback loop of the switching regulator of Fig. 3 will result in a variable switching frequency which will depend mainly on the hysteresis width h and the operating point [16]. The simplest converters with such a constraint are the boost converter and the fourth order structures like boost with output ﬁlter (BOF), buck with input ﬁlter (BIF), and Ćuk converter with galvanic isolation. It was shown in [17, 18] that both BIF and Ćuk converter, behaving as ggyrators with controlled output current, can exhibit stable ggyrator characteristics if capacitive damping are inserted and certain parametric conditions are satisﬁed. In particular, boost converters, behaving as ggyrators with controlled input current, can exhibit a sliding regime with unconditionally stable equilibrium point. Therefore, in this study boost converters have been chosen to be cascaded because of their higher efficiency when compared with other fourth order structures having the same sliding and stability characteristics.
Power Gyrator Based on SMC of two Cascaded Boost Converters
i L1
L1
+ −
5
i L2 L 2 +
Vg

S1
C1
+ S2
vc1
C2
−
vc2
R
−
×
+
HC
2h
g
−
Multiplier
Fig. 4. The schematic diagram for a power gyrator with controlled input current based on two cascaded boost converters.
3 GGyrator based on the cascade connection of two boost converters In this study, our main objective is to design a ggyrator based on two cascaded boost converters in order to fed a load resistance with a 380 V dc output voltage. For that, the two cascaded converters will be controlled by means of a single sliding surface in order to reduce the number of components of the system. Therefore, a single control variable is used to drive the switches of both subsystems. Figure 4 depicts the circuit description corresponding to ggyrator based on two cascaded boost converters. The gyrator characteristics can be implemented by means of a sliding mode regulation loop like the described in the previous section. The sliding surface imposes that the input current of the ﬁrst converter is proportional to the output voltage of the second converter. The sliding surface can be described with the switching function: s(x) = gv c2 − i L1 .
(3)
In steadystate s(x) = 0, i.e., I L1 = gV c2 . Furthermore, by considering that the two converters in Fig. 3 are ideal, one will have V g I L1 = V c2 I R and therefore, (2) will be automatically satisﬁed to obtain the following relationship between the output current and the input voltage I R = gV g
(4)
4 Ideal sliding dynamics 4.1 Switched model By applying standard KVLs and KCLs to the circuit depicted in Fig. 4. The two cascaded boost converters can be represented by the following differential equations
6  R. Haroun et al. di L1 dt di L2 dt dv c1 dt dv c2 dt
V g (1 − u)v c1 − L1 L1 v c1 (1 − u)v c2 − = L2 L2 (1 − u)i L1 i L2 − = C1 C1 (1 − u)i L2 v c2 − = C2 RC2 =
(5) (6) (7) (8)
where V g is a voltage source which is considered constant. All the other parameters that appear in (5)–(8) are shown in Fig. 4. The signal u is the control variable used to drive the switches of both converters. u = 1 during the period T ON and u = 0 during the period T OFF .
4.2 Equivalent control The ideal sliding mode model can be obtained by substituting the discontinuous control variable u, in (5)–(8), which is a binary signal belonging to the set {0,1} by its equivalent continuous variable u eq (x) that can take all the values between 0 and 1. This equivalent control variable is obtained by imposing that the trajectories are evolving on the switching manifold. To synthesize a ggyrator in the two cascaded boost converter stages, the switching function can be selected as follows: s(x) = gv c2 − ˙ = 0, where the i L1 . By imposing the invariance conditions [16], one has s(x) = s(x) overdot stands for the time derivative. Therefore, the dynamical behavior of i L1 is constrained by the following differential equation ˙ s(x) =g
dv c2 di L1 − =0 dt dt
(9)
From equations. (5), (8) and (9), the following expression is obtained for the equivalent control variable u eq (x) u eq = 1 −
V g RC2 + gv c2 L1 v c1 RC2 + gi L2 RL1
(10)
Note that u eq (x) must be bounded by the minimum and maximum value of u [12, 13], i.e 0 < u eq (x) < 1
(11)
Substituting (10) in equations (5)–(8) and taking into account (9), the following ideal sliding dynamics reducedorder model is obtained
Power Gyrator Based on SMC of two Cascaded Boost Converters
V g RC2 + gv c2 L1 di L2 v c1 − v = dt L2 RL2 (v c1 C2 + gi L2 L1 ) c2 dv c1 g(V g RC2 + gv c2 L1 )v c2 i L2 − = dt C1 RC1(v c1 C2 + gi L2 L1 ) V g RC2 + gv c2 L1 v dv c2 i L2 − c2 = dt RC2 RC2 (v c1 C2 + gi L2 L1 )

7
(12) (13) (14)
4.3 Equilibrium point Being a third order nonlinear system, the dynamical analysis of (12)–(14) is challenging. However, the linearization of the system near the operating point reveals that the system is stable with certain conditions as will be shown later. The equilibrium point can be obtained by forcing the time derivative of the state variables of the ideal sliding mode model to be null. From (12)–(14) and taking into account the sliding surface equation and that the input power equals the output power, the equilibrium point of the ideal sliding dynamics is given by ⎛ ⎞ g2 V g R ⎜ ⎟ ⎟ ⎞ ⎜ ⎛ ⎜ ⎟ ⎜ gV g Rg ⎟ I L1 ⎜ ⎟ ⎟ ⎜ ⎜ ⎟ * ⎜ I L2 ⎟ ⎜ ⎟ x =⎜ (15) ⎟= ⎟ ⎝ V c1 ⎠ ⎜ Rg V g ⎜ ⎟ ⎜ ⎟ V c2 ⎜ ⎟ ⎜ ⎟ ⎝ gV g R ⎠
It can be observed that I L1 = gV c2 which deﬁnes the steadystate gyrator behavior and that I L2 = gV c1 , which deﬁnes a steadystate Loss Free Resistor (LFR) characteristics that means that the input port of the second boost converter has natural LFR characteristics. The control law at the equilibrium point can be obtained by substituting (15) in (10). In doing so, one obtains the following steady state value U eq of the control variable u eq 1 U eq := u eq (x* ) = 1 − Rg
(16)
As we mentioned before in (11) that U eq is bounded between 0 and 1, the following condition should be fulﬁlled Rg > 1
(17)
8  R. Haroun et al.
Tab. 1. Parameter values used in this study. Vg (V)
V c1 (V)
V c2 (V)
L1 (μH)
L2 (mH)
C1 , C2 (μF)
g (S)
R (Ω)
h
fs (kHz)
18
80
380
200
2
10
0.0135
2600
0.1
100
4.4 Stability analysis In order to study the stability of the system, the nonlinear model (12)–(14) is linearized around the equilibrium point x* given by equation (15). The stability of the linearized system can be studied by using the Jacobian matrix J corresponding to (12)–(14) and evaluating it at the equilibrium point x* . This matrix can be expressed as follows ⎞ ⎛ 2C2 + g 2 L1 2g 2 L + C2 L1 g − 1 ⎟ ⎜ L2 C L2 C RgL2 C ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎜ 2g 2 L + C 2 C2 g g(2g L1 + C2 ) ⎟ ⎟ ⎜ 1 2 − − ⎟ ⎜ − J=⎜ (18) C1 C C1 C ⎟ RgC C 1 ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ g 1 1 ⎟ ⎜ − − ⎠ ⎝ RC RgC RgC
where C = C2 + g 2 L1 and ∆τ2 = C2 L2 − L1 C1 . The characteristic polynomial equation of the linearized system can be obtained using the Jacobian matrix det(J − sI) = 0, where I is the unitary matrix. Developing this equation, the characteristic polynomial can be written in the following form s3 +
L2 C1 + Rg∆τ2 2 2L2 g 2 + 2RgC + C1 2 s + s+ RL2 C1 C RgL2 C1 C RL2 C1 C
(19)
The stability of this system can be checked by using RouthHurwitz criterion to get the following stability conditions L2 C1 + Rg∆τ2 > 0 2
2
2 2
2
(L2 C1 + Rg∆τ )(2g L2 + C1 ) + 2R g C∆τ > 0
(20) (21)
Power Gyrator Based on SMC of two Cascaded Boost Converters

9
5 Numerical simulations In order to verify the theoretical results predicted in Section IV, the circuit depicted in Fig. 4 has been simulated by using PSIM software with the set of parameter values depicted in Tab. 1 that satisﬁes the stability conditions of Section IV. First, the validity of the ideal sliding dynamics model (12)–(14) will be checked using numerical simulation for the full order model. The system is simulated from two certain initial points P1 and P2 using the two different models. As shown in Fig. 5, the trajectories of the reducedorder model are in perfect agreement with the fullorder model for both cases. 400 Equilibrium point 350
vc2
300
250
200
P2
150
P1
100 0
1
2
iL2
3
4
Fig. 5. Trajectories obtained from the reducedorder ideal slidingmode dynamics model and from the fullorder switched model using PSIM starting from different initial conditions P1 and P2 in the state plane (i L2 , v c2 ).
Figure 6 shows the transient start up and steady state responses of the system from zero initial conditions. Note, that after a short transient time of 60 ms, the state variables reach their steady state values which are in agreement with (15). The cascade connection of the two converters is behaving as ggyrator in steady state and the second converter behaves as an LFR in the intermediate point of the two cascaded boost converters, due to the constraints imposed by the switching function s(x).
10  R. Haroun et al. 400 vc2 200 vc1 0 0
0.04
0.02
0.06
0.08
0.1
0.08
0.1
6 4
iL1
2
iL2
0 0
0.02
0.04
0.06 Time (s)
Fig. 6. The capacitor voltages v c2 , v c1 and the inductor currents i L1 , i L2 (respectively from up to down) showing the startup of the ggyrator with controlled input current based on two cascaded boost converters for g = 0.013 S.
Figure 7a shows the effect of the input voltage variation from 18 V to 20 V. By increasing the input voltage, the output capacitor voltages and the inductor currents increase. However, when the output power changes from 90 W (half load) to 180 W (full load)
500 400
vc2
300 200 vc1
100 0 0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.14
0.16
0.18
0.2
(a) 6 4
iL1
2
iL2
0 0.02
0.04
0.06
0.08
0.1
0.12
Time (s)
Fig. 7. The capacitor voltages v c2 , v c1 and the inductor currents i L1 , i L2 (respectively from up to down) for ggyrator with controlled input current based on two cascaded boost converters for g = 0.013 S and under (a) Input voltage change from 18 V to 20 V (b) Load change from 90 W to 180 W.
Power Gyrator Based on SMC of two Cascaded Boost Converters

11
400 300 vc2
200 100 0 0.02 (b)
vc1 0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.14
0.16
0.18
0.2
6 4
iL1
2 0 0.02
iL2 0.04
0.06
0.08
0.1 0.12 Time (s)
Fig. 7. (continued) The capacitor voltages v c2 , v c1 and the inductor currents i L1 , i L2 (respectively from up to down) for ggyrator with controlled input current based on two cascaded boost converters for g = 0.013 S and under (a) Input voltage change from 18 V to 20 V (b) Load change from 90 W to 180 W.
(the load R changes from 1600 to 800 Ω), the capacitor voltages and the inductor currents decrease as shown in Fig. 7b. This can also be deduced from (15).
6 Impedance matching between a PV panel and load using power gyrator based on cascaded boost converters In this section, the gyrator based on two cascaded boost converters of the previous sections will be supplied by a PV panel as illustrated in Fig. 8. The used PV panel is BP585 and the required voltage is 380 V at the output as in the previous section. The used hysteresis width is h = 0.1 A which results in a steadystate switching frequency f s of about 100 kHz. The gyrator has been used as interfacing element between the PV panel and the load. This can be considered as an example of application in impedance matching in PV systems when supplying a load resistance with 380 V. The circuit is simulated using PSIM and the parameter g for the control can be obtained from the MPPT controller to supply the two boost converters with the same control variable. An extremumseeking control MPPT is used to extract the maximum power from the PV panel [8]. The direct connection of the PV panel to the load resulting in the operating point A as shown in Fig. 9. The variation of the parameter g of the gyrator changes the
12  R. Haroun et al.
i L1
ip
L1
i L2 L 2 + S1
vp
C1
+
vc1
S2
C2
− PV Panel
HC
ip
−
Multiplier
×
+
2h
ip
R
−
MPPT
vp
vc2
g
Fig. 8. Impedance matching of a PV panel by means of two cascaded boost converters.
V Voc
fin (g2)
A P2
P
R
fin (g1) 1 g22R
1 g21 R
P1 g2 < g1
Isc
I
Fig. 9. Impedance matching between a PV panel and a resistive load.
operating point of the PV panel as illustrated in Fig. 9. Operating points P1 and P2 are corresponding respectively to conductances g1 and g2 with g2 < g1 . The goal of the matching is to ﬁnd an optimal value of the conductance g that leads to an intersection of the PV VI characteristic and f in (g) characteristic at the maximum power point which can be deﬁned with the operating point P in Fig. 9. Figure 10 shows the steady state behavior for the gyrator based on two cascaded boost converters supplied from a PV panel using an MPPT controller. It can be noticed that the output voltage and the output current for the PV panel are 180o out of phase. The frequency of the instantaneous power p is twice the frequency of current or voltage. Therefore, each half period of current or voltage, a maximum value of p is reached. The intermediate voltage v c1 has the same frequency as the voltage of the PV panel and its value is around 80 V. Finally, the output voltage v c2 has the same frequency of the input current and its averaged value is 380 V. The response of the gyrator based on two cascaded boost converters connected to the PV panel with an MPPT have been checked also under the change of temperature T and irradiance S. The PV i−v characteristic curve and the gyrator load line are depicted
 13
p(W)
Power Gyrator Based on SMC of two Cascaded Boost Converters
53.5 53
vp
ip
3.4 3.3 3.2 17 16.5 16 15.5
vc1
80 78 76
0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
vc2
372 371 370
0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 Time (s)
Fig. 10. Steady state waveforms of the power gyrator based on two cascaded boost converters and supplied from a PV panel with an MPPT controller.
4 4
3.5
3.5 2
3.3 g » 0.26(T = 45°C)
2.5 ip (A)
ip (A)
2.5 2
2
1.5
1.5
g » 0.25 (T = 25°C)
1
1
0.5
(a) 00
g » 0.25 S (S = 700W / m )
3.3
2
g » 0.21 S (S = 500W / m )
0.5
5
10
vp (V)
15
20
(b) 00
25
5
10
vp (V)
15
20
p(W)
55 60 p(W)
T = 25 °C
50
40
0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5
3.5
0 –20
ip
T = 25 °C
3
T = 45 °C 2.5
2.5
(c)
0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5
3.5
vp 3
S = 500W/m2
20
T = 45 °C 45
S = 700W/m2
0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5 Time (s)
2
(d)
S = 700W/m2 S = 500W/m2
0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5 Time (s)
Fig. 11. Response of the the gyrator based on two cascaded boost converters supplied by the PV panel with the MPPT controller under temperature and irradiation changes. (a): Temperature effect at S = 700 W/m2 . (b): Irradiance effect at T = 25o C. (c): Temperature effect at S = 700 W/m2 . (d): Irradiance effect at T = 25o C.
25
14  R. Haroun et al.
in Fig. 11a and Fig. 11b for the same step change in the temperature and irradiance respectively. It can be noticed that, when the temperature increases from 25 o C to 45 o C, the average conductance g changes from 0.25 S to 0.26 S to extract the maximum power as shown in Fig. 11a. When the irradiance changes from S = 700 W/m2 to S =500 W/m2 , the conductance g changes from 0.25 S to 0.21 S for achieving the Maximum Power Point (MPP) steady state for the new conditions as shown in Fig. 11b. Therefore, the steadystate of the system for both irradiance levels is oscillating around the MPP. This can also be observed in the corresponding waveforms of v p and p which are depicted in Fig. 11c, 11d respectively. It can be noticed that when the temperature increases, the power decreases as shown in Fig. 11c. Similarly, when the irradiance decreases, the power decreases as shown in Fig. 11d, while maintaining the operating point at the MPP for the both cases.
7 Conclusion In this paper, the gyrator concept has been used to connect two cascaded boost converters using a single sliding surface to reduce the number of components which will consequently decrease the cost and increase the efficiency. This system has been analyzed theoretically by means of numerical simulation using PSIM. The ideal sliding model has been obtained and validated using the fullorder switched model. Stability analysis has been carried out for the gyrator based on two cascaded boost converters and the conditions for sliding motions and for stability have been derived. Using the gyrator canonical element based on SMC adds simplicity for the stability analysis and the implementation. It has been deduced that the gyrator based two cascaded boost converters has a natural LFR in the intermediate point of the converters. It has also been shown that gyrator based cascaded boost converters can be used to solve the problem of achieving high voltage conversion ratio and at the same time for dc impedance matching in photovoltaic systems with potentially a better efficiency if only a single switching surface is used to drive the switches of the two cascaded converters. Acknowledgements: This work was supported by the Spanish MINECO under grants DPI201016481, DPI201016084 and CSD200900046.
Bibliography [1]
A. Cellatoglu and K. Balasubramanian. Renewable energy resources for residential applications in coastal areas: A modular approach. 42nd Southeastern Symp. on System Theory (SSST), 2010.
Power Gyrator Based on SMC of two Cascaded Boost Converters
[2] [3] [4] [5] [6] [7]
[8]
[9]
[10] [11]
[12] [13]
[14] [15] [16] [17]
[18]

15
S. Luo and I. Bataresh. A Review of distributed power systems part I: DC distributed power system. IEEE Aerospace and Electronic Systems Magazine, 20(8):5–16, 2005. D. Salomonsson and A. Sannino. Lowvoltage DC distribution system for commercial power systems with sensitive electronic loads. IEEE Trans. on Power Delivery, 22(3):1620– 1627, 2007. P. Karlsson and J. Svensson. DC bus voltage control for a distributed power system. IEEE Trans. on Power Electronics, 18(6):1405– 1412, 2003. M. Brenna, G. Lazaroiu and E. Tironi. High power quality and DG integrated low voltage dc distribution system. in IEEE Power Engineering Society General Meeting, 2006. G. R. Walker and P. C. Sernia. Cascaded DCDC converter connection of photovoltaic modules. IEEE Trans. on Power Electronics, 19(7):1130–1139, 2004. A. I. Bratcu, I. Munteanu, S. Bacha, D. Picault and B. Raison. Cascaded DCDC converter photovoltaic systems: power optimization issues. IEEE Trans. on Industrial Electronics, 58(2):403–411, 2011. R. Leyva, C. Alonso, I. Queinnec, A. CidPastor, D. Lagrange and L. MartinezSalamero. MPPT of photovoltaic systems using extremumseeking control. IEEE Trans. on Aerospace and Electronic Systems, 42(1):249–258, 2006. R. Haroun, A. CidPastor, A. El Aroudi, and L. MartinezSalamero. Synthesis of canonical elements for power processing in dc distribution systems using cascaded converters and slidingmode control. IEEE Trans. on Power Electronics, Early acsess 2013. S. Singer and R. Erickson. Canonical modeling of power processing circuits based on the POPI concept. IEEE Trans. on Power Electronics, 7(1):37–34, 1992. A. CidPastor, L. MartinezSalamero, C. Alonso, G. Schweitz and R. Leyva. DC Power Gyrator versus DC Power Transformer for Impedance Matching of a PV Array. 12th Int. Power Electronics and Motion Control Conf. (EPEPEMC), :1853– 1858, Aug. 2006. V. I. Utkin. Sliding modes and their application in variable structure systems. MIR Publishers, 1978. P. Mattavelli, L. Rossetto, G. Spiazzi, and P. Tenti. Generalpurpose slidingmode controller for DC/DC converter applications. 24th Annual IEEE Power Electronics Specialists Conf., (PESC), :609–615, 1993. S. Singer. Gyrators Application in Power Processing Circuits. IEEE Trans. on Industrial Electronics, 34(3):313–318, 1987. S. Singer. Lossfree gyrator realization. IEEE Trans. on Circuits and Systems, 35(1):26–34, 1988. H. SiraRamirez. Sliding motions in bilinear switched networks. IEEE Trans. on Circuits and Systems, 34(8):919–933, 1987. A. CidPastor, L. MartinezSalamero, C. Alonso, G. Schweitz, J. Calvente, and S. Singer. Classiﬁcation and synthesis of power gyrators. in IEE Proc. on Electric Power Applications, 153:802–808, 2006. A. Cid Pastor, Energy processing by means of power gyrators. PhD thesis, Technical University of Catalonia (UPC), Barcelona, July 2005.
16  R. Haroun et al.
Biographies Reham Haroun was born in Egypt in 1982. She obtained the graduate degree in power and electrical engineering from Aswan Faculty of Engineering, South Valley University, Aswan, Egypt, in 2004 and the Master degree from the same University in 2009 where she worked as lecture assistant during the period 2004–2009. During the same period, she was a member of Aswan Power Electronics Application Research Center (APEARC) group. Her research interests are in power electronics applications including dcdc switched power supply and ACDC Power Factor Correction (PFC) converters. She received her Ph.D degree at Universitat Rovira i Virgili, Tarragona, Spain in 2014. Abdelali El Aroudi was born in Tangier, Morocco, in 1973. He received the graduate degree in physical science from Faculté des sciences, Université Abdelmalek Essaadi, Tetouan, Morocco, in 1995, and the Ph.D. degree (hons) from Universitat Politècnica de Catalunya, Barcelona, Spain in 2000. During the period 1999–2001 he was a Visiting Professor at the Department of Electronics, Electrical Engineering and Automatic Control, Technical School of Universitat Rovira i Virgili (URV), Tarragona, Spain, where he became an associate professor in 2001 and a fulltime tenure Associate Professor in 2005. From September 2007 to January 2008 he was holding a visiting scholarship at the Department of Mathematics and Statistics, Universidad Nacional de Colombia, Manizales, conducting research on modeling of power Electronics circuits for energy management. From February 2008 to July 2008, he was a visiting scholar at the Centre de Recherche en Sciences et Technologies de Communications et de l’Informations (CReSTIC), Reims, France. He has participated in three Spanish national research projects and ﬁve cooperative international projects. His research interests are in the ﬁeld of structure and control of power conditioning systems for autonomous systems, power factor correction, stability problems, nonlinear phenomena, chaotic dynamics, bifurcations and control. He serves as usual reviewer for many scientiﬁc journals. He has published more than 150 papers in scientiﬁc journals and conference proceedings. He is a member of the GAEI research group (Universitat Rovira i Virgili) on Industrial Electronics and Automatic Control whose main research ﬁelds are power conditioning for vehicles, satellites and renewable energy. He has given invited talks in several universities in Europe, South America and Africa.
Power Gyrator Based on SMC of two Cascaded Boost Converters

17
Angel CidPastor graduated as Ingeniero en Electrónica Industrial in 1999 and as Ingeniero en Automàtica y Electrónica Industrial in 2002 at Universitat Rovira i Virgili, Tarragona, Spain. He received the M.S. degree in design of microelectronics and microsystems circuits in 2003 from Institut National des Sciences Appliquées, Toulouse, France. He received the Ph.D. degree from Universitat Politècnica de Catalunya, Barcelona, Spain, and from Institut National des Sciences Appliquées, LAASCNRS Toulouse, France in 2005 and 2006, respectively. He is currently an Associated Professor at the Departament d’Enginyeria Electrònica, Elèctrica i Automàtica, Escola Tècnica Superior d’Enginyeria, Universitat Rovira i Virgili, Tarragona, Spain. His research interests are in the ﬁeld of power electronics and renewable energy systems. Luis MartinezSalamero received the Ingeniero de Telecomunicación and the doctorate degrees from the Universidad Politécnica de Cataluña, Barcelona, Spain in 1978 and 1984, respectively. From 1978 to 1992, he taught circuit theory, analog electronics and power processing at Escuela Técnica Superior de Ingenieros de Telecomunicación de Barcelona. During the academic year 1992–1993 he was visiting professor at the Center for Solid State Power Conditioning and Control, Deparment of Electrical Engineering, Duke University, Durham, NC. He is currently a Full Professor at the Departamento de Ingeniería Electrónica, Eléctrica y Automática, Escuela Técnica Superior de Ingeniera, Universidad Rovira i Virgili, Tarragona, Spain. During the academic years 2003–2004 and 2010–2011, he was a Visiting Scholar at the Laboratoire d’Architecture et d’Analyse des Systèmes (LAAS) of the Research National Center (CNRS), Toulouse, France. His research interest are in the ﬁeld of structure and control of power conditioning systems for autonomous systems. He has published a great number of papers in scientiﬁc journals and conference proceedings and holds a US patent on the electric energy distribution in vehicles by means of a bidirectional dcto dc switching converter. He is the director of the GAEI, research group on Industrial Electronics and Automatic Control whose main research ﬁelds are power conditioning for vehicles, satellites and renewable energy. Dr MartínezSalamero was Guest Editor of the IEEE Trans. on Circuits And Systems (1997) for the special issue on Simulation, Theory and Design of SwitchedAnalog Networks. He has been distinguished lecturer of the IEEE Circuits and Systems Society in the period 2001–2002.
F. FloresBahamonde, H. ValderramaBlavi, J. A. Barrado Rodrigo, J. M. Bosque and A. LeonMasich
Evaluating Power Converters using a WindSystem Simulator Abstract: Distributed systems become an important solution to the generation of green energy integrating different renewable energy sources. The randomness of this kind of sources is counteracted by storage elements, injecting the energy into the grid according to a given power proﬁle. Integration of wind generator to a DCbus microgrid requires the study of threephase rectiﬁers to inject the energy, delivered by the generator, into the DCbus. Different converter topologies can be proposed for both power factor capability, and maximum power point tracking (MPPT) algorithm. To optimize the overall system, a windsystem simulator is required to make repeatable experiments for testing the performance of three different power rectiﬁers. In addition, MPPT algorithms can also be examined. This paper considers the development of such experimental tool using the graphical environment Labview. This simulator is programmed to drive,by means of a torque openloop, an induction machine mechanically coupled to the wind generator. Finally, the evaluations of the threephase PFC rectiﬁers are carried out, and a classical MPPT algorithm is also implemented to test the simulator. Keywords: Wind energy, Windsystem simulator, PFC rectiﬁers, Threephase LFR, MPPT algorithm. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction The increasing demand on electric energy and a global concern on fossilfuel emission levels encourage the development of renewablebased gridconnected systems. The optimization on the energy production requires continuous improvements in energy sources, conversion stages, and transport networks. In this context, distributed generation systems (DGS) have been widely used in power distribution, and recently have been proposed to integrate different renewable energy sources contributing to generate green energy with lower transport losses. The use of distributed conﬁguration has different advantages, such as the minimization of harmonics and electromagnetic interferences (EMI), redundancy, reliability and
F. FloresBahamonde, H. ValderramaBlavi, J. A. Barrado Rodrigo, J. M. Bosque and A. LeonMasich: DEEEA, Universitat Rovira i Virgili, Av. Països Catalans 26, Tarragona, Spain, emails: freddy.ﬂ[email protected], [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 19–38. DOI 10.1515/9783110448412002
20  F. FloresBahamonde et al.
standardization of the power architectures [1, 2]. The connection of different renewable sources to a regulated DCbus can include storage devices to counteract the randomness of the energy produced,allowing the injection of energy into the grid according to a given proﬁle [3]. Figure 1 illustrates a distributed generation system under development. This is based on a variable DCbus (270370 V) ascore of the system. The plant is tied to the grid in a single connection point (PCC) by a 6 kW inverter. All the generators, storage elements, and loads are connected to the DCbus through an adaptor circuit [4].
DC BUS 270370 V 15´12 V 250 Ah
ADAPTOR BATTERY BANK
LOADS
ADAPTORS 3 Phase 400 V
INVERTER MAIN GRID
SOURCES
ADAPTORS 120 mF 400 v
CAPACITOR BANK
Fig. 1. DCMicrogrid.
The design of power converters with power factor correction (PFC) capability, high efficiency and low cost is still an important research subject in power conversion systems. In this context, the integration of wind power sources to the distributed generation system requires the study of AC/DC converters to optimize the energy injected to a high voltage regulated DCbus. In the literature, different topologies applied to the conversion process of wind turbine can be found. For instance, the sixswitch rectiﬁer [5, 6], used as a ﬁrst stage in backtoback twolevel conﬁguration, and the Vienna [7] rectiﬁer used in NPC threelevel structures [8], are commonly power stages used for AC/DC conversion in high power wind systems. Nevertheless, other structures for low power application, such as the singlephase AC/DC converter composed of an uncontrolled rectiﬁer and a boost converter stage, are commonly used also in some applications [9]. Moreover, some authors have proposed to use three singlephase rectiﬁers, one per phase, as a suitable solution. Thus, the power is split in three stages offering modularity and easy implementation [10, 11].
Power converters using windsystem simulator
 21
The main idea behind this study is to evaluate the performance, efficiency, and complexity of the different threephase PFC rectiﬁers in order to optimize the efficiency of the system maximizing the energy delivered by the generator. For that, repeatable experiments emulating different weather conditions, resulting in different power levels, must be carried out According to that, a wind system simulator is proposed to evaluate AC/DC converters for connecting a wind generator to a distributed system. Besides, the simulator tool can be also used to evaluate the performance of different MPPT algorithms. The whole system is illustrated in Fig. 2. The system proposed is divided in two different parts. The ﬁrst one shows the wind simulator that consist of a wind workbench composed of a wind generator coupled mechanically to an induction motor, and driven by a variable frequency drive (VFD). The main idea is to control the wind workbench by means of Labview, in that way that the workbench behaves as a wind generator for a given certain wind speed proﬁle turbine power coefficient C p (λ, β).
Wind Simulator
SmartGrid Wind Integration DCLink
Wind
Threephase PFC rectifier
GS
C
SmartGrid DCbus 270v370v
Vdc(a,b,c) Va,b,c ia,b,c
MPPT algorithm
Fig. 2. Wind Generator Simulator Block Diagram.
The second part of the work consists in testing the simulator, connecting the wind generator to an emulated DCbus using different threephase rectiﬁer. Besides, a MPPT algorithm will be implemented to extract the maximum power of the wind generator. The remaining part of the paper is organized as follows. In Section 2, mathematical details concerning the wind system simulator are given. Next, in Section 3, the implementation of the wind simulatoris explained. In section 4, different rectiﬁers are evaluated. First, the circuits of the different threephase rectiﬁers are given. Next, an example of a classical MPPT algorithm is programed in a PIC controller to drive the rectiﬁer. In Section 5, some preliminary experimental results are given, and ﬁnally in Section 6, some conclusions are summarized.
22  F. FloresBahamonde et al.
2 Wind generator simulator To design a wind simulator is necessary to consider important aspects, such as wind energy power ﬂuctuations, transient effects, and rotor performance. The importance of these aspects, come from the need to determine how much energy can be extracted from the wind. In this context, it is well known that the aerodynamic system of a wind turbine is composed mainly of the rotor blades. The air speed v is reduced by the rotor, and the kinetic energy absorbed from the air E kin is transformed in mechanical power P mec . By [12–14], the kinetic energy from a given cross section of air can be calculated in (1), where ρ is the air density, A the area of the rotor blades, and v wind the wind speed.
t dE kin =
1 ρA(v wind t)v2wind 2
→ P wind =
dE kin dt
(1)
0
Applying the derivative to the energy expression shown in (1), the power available in the wind can be easily deduced: P wind =
1 ρAv3wind 2
(2)
For an air stream,the mechanical power extracted from the wind by an energy transducer is expressed as the difference between the power available in the wind, before and after the transducer, as shown in (3), where v1 represents the air stream speed before, and v2 represents the air stream speed after the energy transducer: P mec =
1 ρA(v21 − v22 )(v1 + v2 ) 4
(3)
Equation (3) can be expressed as (4), where C p is the ratio of the mechanical power extracted by the converter, called power coefficient: P mec = C p (λ, β)P wind
(4)
This coefficient is a nonlinear expression that depends on the tipspeed ratio λ (5), and the pitch angle β. The power coefficient represents the power percentage that is actually converted in mechanical power. This concept was introduced by Albert Betz in 1920, who also demonstrated the existence of a physical upperlimit for that value. λ=
ωm R , v
C p < 0.59
(5)
The mathematical expression of the rotor coefficient is the key for wind turbine simulator. In this context different graphical and numerical expressions for power
Power converters using windsystem simulator  23
and torque coefficient aims to characterize verticalaxis or horizontalaxis wind rotors [12–14]. For this reason, different studies have been developed to obtain an approximation of C p (λ, β), as a result of experimental tests [15], mathematically [16] or analytically from ﬂuid dynamics theory applied to a given rotor type. In fact, the expression of C p (λ, β) (6) includes several coefficients (c1 to c6 ), that must be calculated for each real turbine, and depend on some aspects, such as the rotor design, the number and shape of blades, the weight, stiffness and son on. On the other hand, such rotor design aspects cause conversion losses, and therefore the C p (λ, β) in a real turbine will be always lower than shown in (5). To test the wind simulator, a certain expression of C p (λ, β) is required. As the generator used in our system come from a horizontal turbine with three blades, we propose the expression (6) with the coefficients deﬁned as c1 = 0.5, c2 = 116, c3 = 0.5, c4 = 0, c5 = 5 and c6 = 21 [9]: c2 c6 x (6) − c3 β − c4 β − c5 exp − C p = c1 λi λi where: λi =
0° 1° 3° 5° 7° 10° 12° 15°
0.4
2´103
0.3
(7)
3 m/s 4 m/s 5 m/s 6 m/s 7 m/s 8 m/s 9 m/s 10 m/s 11 m/s 12 m/s
P (W)
C (l , b )
1 1 0.035 − λ + 0.08 β3 + 1
1´103
0.2
0.1
0
0
1.5
3
4.5
6
7.5 l
(a)
9
10.5
12 13.5
15
0 0
200
400
600 V (rpm)
800
1´103
(b)
Fig. 3. (a): C p versus λ for different pitch angles β. (b): P mec in terms of rotor speed and wind speed.
Figure 3a shows the characteristic curves for the power coefficient C p (λ, β) for different values of λ and β. These curves are obtained by substituting (7) in (6) and using the length of the rotor blades of the generator (R = 1.35 m). In the same way, from (4) and (6) is possible to obtain the waveforms for the power extracted in function of the rotor speed in rpm for different wind velocities, between 3m/s and 12m/s, observed
24  F. FloresBahamonde et al.
in Fig. 3b. Wind turbine aerodynamic equations show that the power available in the wind P wind depends on the cube of the wind speed. By other hand, the power really extracted from the wind P mec depends also on the rotor angular speed ω, and the pitch angle of the rotor blade β. Thus, the wind simulator has to control the workbench to obtain an output power from the generator following the curves depicted in Fig. 3b.
3 Wind simulator implementation In this section the implementation of the wind simulator is described, using equations (6) and (7). Figure 4 illustrate the workbench that consist of a permanent magnet synchronous machine, from a horizontal axis turbine, that is mechanically coupled to an induction motor, and driven by a commercial variable frequency drive (VFD). The main idea of the wind simulator is that using a standard scalar openloop torque control the VFD drives the induction machine emulating the wind turbine mechanical characteristics, i.e. emulate the turbine shaft [17].
Wind Simulator
VFD
M
GS ω
τ Wind Turbine Model
Pmec = Cp ( λ, β )Pwind τ = Pmec /w
Fig. 4. Wind workbench emulator.
The simulator will have three input variables: the wind speed v, the rotor angular speed ω, and the pitch angle of the rotor blades β. The ﬂowchart in Fig. 5 shows how the wind speed, and the rotor angular speed are used to calculate the power available in the wind P wind , the power coefficient C p , and ﬁnally a torque reference for the induction motor. Therefore, using the graphic platform Labview, the torque reference τ is calculated according to Fig. 5, and is sent to the VFD, previously programmed using sensorless Vectorial Control setting.
Power converters using windsystem simulator  25
Wind speed V
Rotor speed W
Power Wind
Tipspeed ratio ωmR λ= v
1 Pwind = ρAv 3wind 2
Power coefficient Cp ( λ, β ) –c c λ Cp = c1 ( 2 –c3 β–c4βx –c5 ) e λi
6
i
Mecanic Power extracted Pmec = Cp ( λ, β )Pwind
Torque τ = Pmec /W
Fig. 5. Torque calculation Flowchart.
To initialize the program, the ﬁrst step is the introduction of the input variables as the wind speed and the rotor blades pitch angle, and then, the calculation of the rotor speed that is measured by an encoder continuously. The communication between the variable frequency drive and the encoder with the computer is realized by the PCI6024E data acquisition card from National Instruments. The ﬂowchart shown in Fig. 6 illustrates the program design in Labview, where, the ﬁrst step is to initialize the input variables. After, a given proﬁle of wind can be programmed and charged in the code. This proﬁle can be previously generated by statistic methods according to a given climate proﬁle in a geographic area. Once the proﬁle is charged, a subroutine is triggered for calculating the rotor speed. In this subroutine, each 20 ms, the encoder pulses are counted, and the rotor speed is calculated and ﬁltered by the software to eliminate measurement noise. On the other hand,once the rotor speed ω is calculated, different subroutines are executed in a concurrent form. According to Fig. 6, from the left side, the ﬁrst subroutine calculates the torque reference as is described in Fig. 5, and sent it to the VFD through the DAQ assistant (digital analog toolbox of Labview). The next subroutine is designed to detect any changes in the wind speed. If any change occurs, the simulator calculates and draws the new power curves as shown in Fig. 7.
26  F. FloresBahamonde et al.
Inicialization Manual Fuctionality
Manual or Automatic
Y
Manual/Automatic Selection
N N
On Y
Enconder adquisition (DAQ) Velocity calculation (rpm) 20 ms
N
N
Wind Changed
Wind Turbine Model (torque calculation)
200 ms
Y
Torque signal output (DAQ)
Y
New power curve calculation and drawn
Graphic of Cp and rotor velocity
Fig. 6. Flowchart of the Labviewprogrammed code.
9 m/s
Extracted Power (w)
8 m/s
7 m/s
6 m/s 5 m/s 4 m/s 3 m/s Rotor Speed (Rpm)
Fig. 7. Mechanical power curves obtained from Labview.
N
1000ms
Y
Power mesurement and Actual position
Power converters using windsystem simulator
 27
The next subroutine shows the information of different variables to the user, for instance: the rotor speed, power available, mechanical power and the power coefﬁcient. The last routine is devoted to refresh the power extracted from the wind, and measured in the output of the generator. This information is graphically shown over the power curve drawn for a given wind speed. This is considered important information, because shows the actual point where the system is working, and also it allows to observe graphically when the maximum power point is achieved.
4 ACDC conversion and MPPT algorithm To evaluate the wind simulator implemented in the previous section, we connect different threephase rectiﬁers with power factor correction (PFC) at the output of the synchronous generator. This section, is divided in two parts. The ﬁrst is dedicated to the explanation of the topologies used to extract the power of the wind turbine. The second part is devoted to a MPPT algorithm implemented to extract the maximum power of the wind turbine.
4.1 PFC BoostBased Rectiﬁers To evaluate the emulator, three different rectiﬁer topologies will be connected to the output of the wind generator. All the energy delivered is injected to a highvoltage DCbus emulated by a power supply, as is depicted in Fig. 8. Figure 9 illustrates the rectiﬁers. A rectiﬁer composed of three modular singlephase rectiﬁers is shown in Fig. 9a, the sixswitch rectiﬁer in Fig. 9b, and the Vienna rectiﬁer in Fig. 9c. The three rectiﬁers are controlled to behave as a lossfree resistor (LFR) by means of sliding control mode [18]. This means that, imposing a sliding surface to the inductor current, the threephase rectiﬁer is seen by the generator as three equal resistances, one per phase, obtaining a good power factor correction. As the currents have sinusoidal waveforms, the winding losses are reduced and the torque vibrations are eliminated. The switching surface, shown in (8), must induce the sliding motions in each phase of the rectiﬁer, where g (expressed in Ω−1 ) is the input conductance of each phase, and V j is the input voltage of that j th phase. At permanent regime is reached S(x) = 0, and then i Lj = gV j . Consequently the current will track its respective input voltage [10, 11, 18] (for j = A, B, C): S j (x) = i Lj − gV j = 0
(8)
28  F. FloresBahamonde et al.
Then we can calculate the power delivered by each phase as: P j (t) = V j i j
(9)
Solving (9), and assuming permanent regimen conditions is applied (8), we obtain ⎧ 2 2 ⎪ ⎨ P A (t) = gV m sin ωt 2 (10) sin2 ωt + 2π P B (t) = gV m 3 ⎪ ⎩ P (t) = gV 2 sin2 ωt − 2π m C 3 where V m is the peak voltage of the sinus wave. On the other hand, an assuming an ideal converter without losses, the output power can be expressed as P0 = P A + P B + P C =
Output DCLink
i0
3 2 gV 2 m
(11)
rbus
+ v0 –
+ Vbus –
Cb
RL
Vb
+ –
Fig. 8. Emulated busDC.
Applying some manipulations is easy to deduce that the current injected into the DCbus can be expressed as: I bus =
2 P0 3gV m = V bus 2V bus
(12)
Equation (11) shows that the surface expressed in (8) performs simultaneously a power factor correction for each phase, allowing also, a direct control of the power extracted from the wind turbine. This power can be controlled adjusting the input conductance (g) of each phase. This conductance that now is controlled externally by an arbitrary voltage reference, in the future can be controlled by the MPPT algorithm.
4.2 MPPT algorithm Although the power extracted from the wind depends on the cube of the wind speed, and the power coefficient C p (λ, β) depends on both, rotor speed and wind speed, it is not convenient to develop a MPPT algorithm sensing the wind speed, as such sensors are slow and imprecise [14]. For this reason, MPPT algorithms normally are based on electrical variables, for example, the rotor power.
Power converters using windsystem simulator
 29
Output DCLink N
N N
+
–
N (a)
Output DCLink
N
(b) Output DCLink
(c)
N
Fig. 9. (a) Modular singlephase rectiﬁer. (b) Sixswitch rectiﬁer. (c) Vienna rectiﬁer.
According to [19, 20], if the curves depicted in Fig. 10a are taken into account, it can be deduced that the maximum power can be found by means of (8). The algorithm operation for different wind speed can be observed in Fig. 10b, where v2 > v3 > v1 . How is demonstrated in [20], applying some mathematical manipulations, expression (13) becomes (14), where V gen is the output turbine voltage, ω the rotor speed, and ω e the voltage phase angle speed generated by the turbine: dP mec =0 dω
(13)
30  F. FloresBahamonde et al.
2500
Pmax 12m/s P(Vrpm,3) 3 P(Vrpm,4) 2´10 P(Vrpm,5) Pmax 11m/s P(Vrpm,6) P(Vrpm,7) P(Vrpm,8) Pmax 10m/s P(Vrpm,9) 3 P(Vrpm,1) 1´10 Pmax 10m/s P(Vrpm,11) P(Vrpm,12)
0 0
(a)
0
200
400
600
50
800
Vrpm
1´10 1´10
3
3
3
1´10
800
P(Vrpm,8) P(Vrpm,7) P(Vrpm,5)
600
V2
V3
400
200 V1
0
(b)
0
200
400 Vrpm
600
800
Fig. 10. (a): Maximum power points on the wind power curve. (b): MPPT algorithm performance.
where: dP mec dP mec dV gen dω e × × = dω dV gen dω e dω
(14)
Through (13) and (14), (15) is obtained and is possible ﬁnd the maximum power controlling the output power of the wind turbine. Therefore, the maximum power extracted is achieved controlling V gen , thus avoiding the use of mechanical variables as ω e , and obtaining a better performance, speciﬁcally in precision terms. Thus: dP mec =0 dω
→
dP mec =0 dV gen
(15)
Power converters using windsystem simulator
 31
The algorithm used in this work is shown in the ﬂowchart given in Fig. 11. The algorithm is initialized with an arbitrary voltage reference V dc , where V dc is a proportional value of the conductance given by V dc = kg. Next, the measurement of the voltage and current in the generator output is realized by the controller, and a cyclic process begins until the maximum power point is reached. The expression V dc represents the DCvoltage level introduced at the rectiﬁer topology to vary the conductance of each phase. Therefore, and increment or decrement of V dc produces directly and increment or decrement of the rectiﬁer conductance. Note that, the output voltage is imposed by the DCbus, thus,the currents and voltage signals using by the algorithm are measured at the rectiﬁer input and have a rectiﬁed sinusoidal waveform.Besides, the voltage of the generator has a range of frequency which varies between 20 − 80 Hz and consequently between 40 − 160 Hz for the rectiﬁed waveforms.Whatever be the frequency then umbers of samples have to be sufficiently high to calculate the mean value of each semicycle. On the other
Inicilization Reference Vdc ( i )
Measurement Vgen(k) and igen(k)
Calculation PVmec(k) = Vgen(k) · igen(k)
N
N
Vgen(k) ³ Vgen(k–1)
Vdc(k) = Vdc(k–1) – DVdc
Y
Pmec(k) ³ Pmec(k–1)
Y
N
Vgen(k) ³ Vgen(k–1)
Vdc(k) = Vdc(k–1) + DVdc
Fig. 11. Classical MPPT algorithm ﬂowchart.
Y
Vdc(k) = Vdc(k–1) – DVdc
32  F. FloresBahamonde et al.
hand, as well known, it is mandatory to realize the conductance changes, i.e. power changes, in the end of each semicycle. Therefore, using a dsPIC30F2020 controller, the algorithm described in Fig. 11 was implemented, where 85 samples of eachsemi cycle of i gen and V gen are accumulated to calculate the mean value. Moreover, taking into account that the mechanical system have a time constant to reach the steady state, the power calculated used by the algorithm P mek(k) is the result of averaging the last 5 power samples, this means, that, the averaging of each 5 semicycles of the signal. The number of samples used to calculate the power value P mek(k) can be adapted by the user to adjust the delay needed for stabilize the MPPT algorithm [20, 21]. Finally, once the algorithm has found the MPP the system oscillate around this point.
5 Experimental results To verify the wind simulator design, and prove its functionality, different experiments have been realized to evaluate its performance. The main idea of the wind simulator is to study the performance of both, PFC rectiﬁer converters, and the MPPT algorithms. Therefore, the ﬁrst experiment is the connection of the rectiﬁers shown in Fig. 9. The converters are connected to the output of the wind turbine, and also connected to a dcbus composed of a capacitors bank, a power supply and a load, Fig. 8. The picture shown in Fig. 12 depicts the wind workbench, whereas the Fig. 13a, 13b and 13c shown the steady state waveforms for current of the rectiﬁers of the
Fig. 12. Workbench connected to a threephase rectiﬁer.
Power converters using windsystem simulator  33
V0
i0
iLA
iLB
iLC
(a)
v0 iLA
iLB
iLC
i0
(b)
iLA
iLB
iLC
VAN
(c)
Fig. 13. (a): Steady Bus experiment for modular 3phase rectiﬁer. (b): Steady Bus experiment for Viennaswitch rectiﬁer. (c): Steady Bus experiment for sixswitch rectiﬁer.
34  F. FloresBahamonde et al.
8 Wind Speed (m/s)
Wind Speed (m/s)
8 6 4 2
Extracted Power (w)
300 200 100
2
400 300 200 100
0 0,5
0 0,5
0,4
0,4
0,3
Cp
Extracted Power (w) Cp
400
0,2 0,1 0 04:56:50 09/05
0,3 0,2 0,1 0 04:50:21 09/05
04:56:51 09/05 Time
Time
(a)
Wind Speed (m/s)
4
0 500
0 500
Extracted Power (w)
6
04:50:22 09/05
(b) 8 7 6 5 4 3 500 400 300 200 100 0 0,5 0,4
Cp
0,3 0,2 0,1 0 03:35:57 09/03
03:37:09 09/03 Time
(c) Fig. 14. (a): Constant conductance and variable wind speed. (b): Constant wind speed and variable conductance. (c): Given proﬁle wind speed and MPPT algorithm.
Power converters using windsystem simulator  35
Fig. 9a, 9b and 9c respectively. All the rectiﬁers are connected to a approximately 300 V dcbus. As has been demonstrated in the theoretical analysis, by means of adjusting the input conductance the power extracted from the wind turbine can be controlled. In addition, the rectiﬁer conductance can be modiﬁed manually with the aid of an external DC power supply. Then, if a DCvoltage excursion is carried out, it is possible to explore the power curves shown in Fig. 7. The next experiment is shown in Fig. 14a. Keeping constant the conductance values, a wind proﬁle is programmed as depicted in the upper part of that ﬁgure. The corresponding extracted power is depicted in the central part of the ﬁgure, and ﬁnally, in the inferior part of the ﬁgure, the power coefficient is shown. A similar experiment is realized and depicted in Fig. 14b. Now, the conductance is randomly varied for a given wind speed (5 m/s). This experiment demonstrates that adjusting the rectiﬁer conductance is possible to control the amount of power extracted for a certain given wind speed. Finally, the experiment shown in Fig. 14c is the most important of all.Here the manual conductance adjustment is replaced by the MPPT controller explained in Section III. In the upper part of the ﬁgure a random wind proﬁle is given. Next, in the central part of such ﬁgure, we can observe the extracted power, and ﬁnally the inferior part the power coefficient C p is shown. As we can realize, the power coefficient is always around the maximum point, with the exception associated to the windspeed transients. This last experiment demonstrates the correct behavior of the MPPT algorithm, proving the importance of the wind simulator as a platform tool for repeatable experiments.
6 Conclusions The developed work in this project is a contribution for the study of renewable energy systems, speciﬁcally for the wind energy. The main idea is to design a simulation platform to realize repeatable experiments to study the performance of both, the conversion chain converters, and the MPPT algorithm. After programming a speciﬁc C p (λ, β) curve, the user by means of the Labview graphic interface, can introduce a wind proﬁle and then observe the P − ω curve, the power extracted, and the power coefficient C p (λ, β) of the wind turbine. Besides, measuring the power output is possible to situate and observe the actual generator operation point in the power curve P − ω. By other hand, the behavior of the simulator has been veriﬁed connecting a threephase rectiﬁer at the output of the wind turbine, developing also a MPPT
36  F. FloresBahamonde et al.
algorithm to extract the maximum power point or equivalently to operate at a constant value of the power coefficient C p (λ, β). The wind simulator will be used to compare the behavior of different rectiﬁer topologies with power factor correction, and to evaluate the global conversion chain. Acknowledgement: This work has been partially sponsored by the Spanish Ministry of Research and Science, under grants: DPI200914713C0302, and Consolider RUE CSD200900046.
Bibliography [1] [2]
[3]
[4]
[5]
[6] [7]
[8]
[9]
[10]
[11]
[12] [13]
S. Luo and I. Batarseh. A review of distributed power systems. Part I: DC distributed power system. IEEE Aerospace Electronic System Magasine, 21(62):5–14, June 2006. K. Kurohane, T. Senjyu, A. Yona, N. Urasaki, E.B. Muhando and T. Funabashi. A high quality power supply system with DC smart grid. Transmission and Distribution Conf. and Exposition, 2010 IEEE PES, :1–6, April 19–22, 2010. C. Millais and L. Colasimone. Large Scale Integration of Wind Energy in the European Power Supply: Analysis Issues and Recommendations. Report of the European Wind Energy Association (EWEA), 2005. H. ValderramaBlavi, J.M. Bosque, F. Guinjoan, L. Marroyo and L. MartinezSalamero. Power Adaptor Device for Domestic DC Microgrids Based on Commercial MPPT Inverters. IEEE Trans. on Industrial Electronics, 60(3):1191–1203, March 2013. H. Zhang and B. Tan. Simulation Research on ThreePhase SixSwitch PWM Rectiﬁer with One Cycle Control. 2nd Int. Conf. on Intelligent Computation Technology and Automation, ICICTA ’09., 2:244–247, October 1011, 2009. M. Liserre, R. Cardenas, M. Molinas and J. Rodriguez. Overview of MultiMW Wind Turbines and Wind Parks. IEEE Trans. on Industrial Electronics, 58(4):1081–1095, April 2011. Hao Chen and D.C. Aliprantis. Analysis of SquirrelCage Induction Generator with Vienna Rectiﬁer for Wind Energy Conversion System. IEEE Trans. on Energy Conversion, 26(3):967–975, September 2011. E.J. Bueno, S. Cobreces, F.J. Rodriguez, A. Hernandez and F. Espinosa. Design of a BacktoBack NPC Converter Interface for Wind Turbines with SquirrelCage Induction Generator. IEEE Trans. on Energy Conversion, 23(3):932–945, September 2008. Y.Y. Xia, J.E. Fletcher, S.J. Finney, K.H. Ahmed and B.W. Williams. Torque ripple analysis and reduction for wind energy conversion systems using uncontrolled rectiﬁer and boost converter. Renewable Power Generation, IET, 5(5):377–386, September 2011. F. FloresBahamonde, H. ValderramaBlavi, J.M. Bosque and L.MartinezSalamero. Modularbased PFC for low power threephase wind generator. 7th Int. Conf.Workshop Compatibility and Power Electronics (CPE), :125–130, June 13, 2011. F. FloresBahamonde, H. ValderramaBlavi, J.M. BosqueMoncusi, L. MartinezSalamero, A. LeonMasich and J.A. Barrado. SinglePhase PFC for ThreePhase Wind Generator, a Modular Approach. 915147: Przeglad Elektrotechniczny. 88(1):56–60. S. Heier. Grid Integration of Wind Energy Conversion Systems. Wiley, 1998. Z. Lubosny. Wind Turbine Operation in Electric Power Systems. Berlin, Germany: Springer, 2003.
Power converters using windsystem simulator
 37
[14] E. Hau. Wind Turbines: Fundamentals, Technologies, Application, Economics. 2nd edition. Springer 2005. [15] HsiangChun Lu and LeRen ChangChien. Use of wind turbine emulator for the WECS development. Int. Power Electronics Conf. (IPEC), 3188–3195, June 21–24, 2010. [16] A. Monroy and L. AlvarezIcaza. Realtime identiﬁcation of wind turbine rotor power coefficient. 45th IEEE Conf. Decision and Control, 3690–3695, December 13–15, 2006. [17] R. Teodorescu and F. Blaabjerg. Flexible control of small wind turbines with grid failure detection operating in standalone and gridconnected mode. IEEE Trans. on Power Electronics 19(5):1323–1332, September 2004. [18] A. CidPastor, L. MartinezSalamero, A. El Aroudi, R. Giral, J. Calvente and R. Leyva. Synthesis of lossfree resistors based on slidingmode control and its applications in power processing. Control Engineering Practice, 21(5):689–699, May 2013. [19] D.A. Caixeta, G.C. Guimaraes, M.L.R. Chaves, J.C. Oliveira and A.F. Bonelli. Maximization of variable speed wind turbine power including the inertia effect. 11th Int. Conf. on Electrical Power Quality and Utilisation (EPQU), :1–6, October 17–19, 2011. [20] E. Koutroulis and K. Kalaitzakis. Design of a maximum power tracking system for windenergyconversion applications. IEEE Trans. on Industrial Electronics, 53(2):486–494, April 2006. [21] R. Leyva, C. Alonso, I. Queinnec, A. CidPastor, D. Lagrange and L. MartinezSalamero. MPPT of photovoltaic systems using extremum – seeking control. IEEE Trans. on Aerospace and Electronic Systems, 42(1):249–258, January 2006.
Biographies Freddy FloresBahamonde was born in Osorno, Chile, in 1983. He received the B.S in electronical engineering, in 2007 from Universidad Tecnològica Metropolitana de Chile in Santiago de Chile. In 2009 received the M.S degree from Universitat Rovira i Virgili in Tarragona, Spain. From 2010 he is a member of the GAEI research group (Universitat Rovira i Virgili) on Industrial and Automatic Control whose main research ﬁelds are power electronics, especiﬁcally AC/DC converters for wind energy integration. He is currently working towards the Ph.D degree at Universitat Rovira i Virgili, from Tarragona, Spain. Hugo ValderramaBlavi received the Ingeniero and Ph.D. degrees from the Universitat Politècnica de Catalunya, Barcelona, Spain, in 1995 and 2001, respectively. He is currently an Associate Professor in the Departament d’Enginyeria Electrònica, Elèctrica I Automàtica, Universitat Rovira i Virgili, Tarragona, Spain. During the academic year 20012002, he was a Visiting Scholar at Laboratoire d’Automatique et Analyse des Systèmes, Centre National de la Recherche Scientiﬁque, Toulouse, France. His current research interests are power electronics, renewable energies, silicon carbide devices, and nonlinear control.
38  F. FloresBahamonde et al.
José Antonio Barrado Rodrigo received the electronic engineering degree from the Universitat de Barcelona (UB), Spain, in 2000 and the Ph.D. degree in automatic control from Universitat Politècnica de Catalunya (UPC), Barcelona, Spain, in 2008. He is currently an Associate Professor with the Department of Electronic Engineering and Automatic Control of the Universitat Rovira i Virgili (URV), Tarragona, Spain. His research interests include analysis, modeling and control of electric generators and power converters applied to renewable energy systems.
Josep M. Bosque received the Enginyer Tècnic Industrial en Electrònica Industrial degree and the Enginyer en Electrònica master degree from the Universitat Rovira i Virgili (URV), Tarragona, pain, in 2005 and 2009, respectively, where he is currently working toward the Ph.D. degree. Since 004, he has been a Research Technician with the Automatics and Industrial Electronics Group, URV. is research interests are power electronics and renewable energy.
Antonio LeonMasich was born in Lleida, Spain, in 1986. He received the B.S and the M.S degrees in electronical engineering, in 2009 and 2011 respectively from Universitat Rovira i Virgili in Tarragona, Spain. From 2009 he is a member of the GAEI research group (Universitat Rovira i Virgili) on Industrial and Automatic Control whose main research ﬁelds are highgain converters and highvoltage converters for electronic ballast using Silicon carbide devices. He is currently working towards the Ph.D degree at Universitat Rovira i Virgili, from Tarragona, Spain.
A. Farah, T. Guesmi, H. Hadj Abdallah and A. Ouali
On line improvement of power system dynamic stability using ANFIS and NSGA II algorithms Abstract: This paper investigates the use of an adaptive power system stabilizer (PSS) for improving the dynamic stability of a power system. Adaptive network based fuzzy inference systems (ANFIS) and the second version of nondominated sorting genetic algorithms (NSGAII) is employed to select the optimal parameters of the controller for different loading conditions. Firstly genetic algorithms are used to tune stabilizer parameters on a wide range of loading conditions to create a data base. Two eigenvaluebased objective functions are considered to place the closedloop system eigenvalues in the Dshape sector. Then, the relationship between these operating points and the corresponding stabilizer parameters is learned by the ANFIS. The proposed stabilizer has been tested by performing non linear simulations and eigenvalue analysis using single machine inﬁnite bus (SMIB) model. The results show the effectiveness and the robustness of the proposed stabilizer to provide efficient damping in realtime. Keywords: Power system stability, PSS, NSGAII, ANFIS. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction The stability of power systems is a challenging problem in electric system operation. This problem can be deﬁned as the ability of the power system to return to normal operating state when subject to disturbance [1]. Currently, power system stabilizers (PSS) are routinely used to improve the dynamic stability of power systems by controlling the excitation of the generator. In general, the function of the PSS is to produce component of electrical torque in phase with the rotor speed deviations. The conventional PSS (CPSS) is widely presented as a leadlag compensator [2, 4]. In [1, 5], the authors have presented a comprehensive study of the effects of the parameters of this leadlag model. Recently, considerable researches have been focused on the designing and using of new adequate damping sources [1–6]. The robust design of the PSS is evaluated by several operating conditions [3].
A. Farah, T. Guesmi, H. Hadj Abdallah and A. Ouali: University of Sfax, National Engineering School of SfaxTunisia, Control & Energy Management Laboratory, email, [email protected], tawﬁ[email protected], [email protected], [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 39–53. DOI 10.1515/9783110448412003
40  A. Farah et al.
The design problem is formulated as an optimization problem. Multiple conventional methods are used to solve this design problem, such as eigenvalues assignment, mathematical programming, and gradient procedure. These techniques are iterative and consume an important computing time. Also, it may converge to local optimum. To overcome these disadvantages, evolutionary algorithms are becoming the most popular in this type of problem. In this context, a second version of nondominated sorting genetic algorithms (NSGAII) [8] is adapted in this paper, to generate the Pareto optimal solutions. Larsen and Swann [2] have demonstrated that the PSS can be very well tuned to an operating point and provide satisfactory damping over a certain range around the design point. Therefore, the stabilizers cannot guarantee the stability and good performance, if the operating conditions change. So, it is necessary to determine the optimum stabilizer parameters at each operating point, which does not allow an online decision. To overcome these difficulties, novel intelligent techniques based on fuzzy logic and artiﬁcial neural network is used in this paper. This online design is based on multi objective evolutionary algorithms and ANFIS and it is done on two different stages. The ﬁrst step consists to adjust controller parameters using an improved version of nondominated sorting genetic algorithms (NSGAII) [7–9] for a wide range of operating conditions. The closedloop system eigenvalues should be placed in the Dshape sector. The second step is based on ANFIS training with the collected inputoutput data pairs which are stocked in the ﬁrst step. The input data are the operating conditions and the outputs are the controller parameters. To assess the effectiveness of the proposed stabilizers, their performance has been tested on a weakly connected power system. Eigenvalue analysis and nonlinear simulations are carried out. This paper is organized as follows: Section 2 describes the small signal modeling of the Single Machine Inﬁnite Bus (SMIB) with PSS. The optimization problem has been introduced in section 3. In section 4 the general theory of NSGAII is presented. The ANFIS is introduced in section 5. Simulations and results are provided and discussed in section 6 and conclusions are given in VII.
2 System modeling A SMIB system, as shown in Fig. 1, is considered for the damping control design. The equivalent line impedance is Z = R + jX. The generator G is equipped with a PSS and it has a local load Y L = G + jB. The preliminary system data are given in the appendix. In Fig. 1, V t and V0 are the generator terminal and the inﬁnite bus voltages, respectively.
Power system dynamic stability 
R + jX
Vt
G
∞
YL
Exciter
KA 1 + sT A
ANFIS
Vref Δ UPSS
+
41
K PSS
1 + sT 1P 1 + sT 2P
1 + sT 3P 1 + sT 4P
sT WP 1 + sT WP
Operatin g Conditions P Q
Δω
− Vt
Fig. 1. SMIB system.
2.1 Generator model In this study, the generator is represented by the thirdorder model [2, 9, 14, 15], which are, the two motion equations and the generator internal voltage equation. pδ = ω b (ω − 1) P m − P e − D(ω − 1) pω = M E − (x fd d − x d )i d − E q PEq = T d0
(1) (2) (3)
where δ and ω are rotor angle and speed, respectively. ω b is the base frequency expressed in rad/sec. In equation (2), P m and P e are the mechanical power input and the electrical power output of the generator, respectively. These powers are in per unit (p.u.). D and M are the damping coefficient and inertia constant respectively. For equation (3), E fd and Eq are the ﬁeld and the internal voltages, respectively. They are in per unit. i d is the daxis armature current xd and x d are the daxis transient reactance and the daxis reactance of the generator, respectively. T d0 is the open circuit ﬁeld time constant. The electrical power P e can be expressed by the daxis and qaxis components of the internal voltage V t and the armature current i, as follows [14, 16]: Pe = Vd id + Vq iq
(4)
Vd = Xq iq
(5)
Vq = V t2
=
Eq − X d i d V d2 + V q2
(6) (7)
X q is the qaxis reactance of the generator. Using equations (4–6) the electrical power can be written as: (8) P e = Eq i q + (X q − X d )i d i q
42  A. Farah et al.
2.2 Structure of excitation system with PSS The IEEE typeST1 excitation system with PSS shown in Fig. 1 is considered in this paper [2, 14], where K A and T A are the gain and the time constants of the excitation system, respectively. V ref is the reference voltage and U PSS is the stabilizer signal. The PSS representation consists of a gain K PSS , a washout block with the time constant T WP and two leadlag blocks. From the excitation system block, the ﬁeld voltage E fd can be expressed as: pE fd =
K A (V ref − V t + U PSS ) − E fd TA
(9)
2.3 PhilipsHeffron model with PSS Using previous equations, the linearized form of P e , E q and V t can be expressed by the following equations: ∆P e = K1 ∆δ + K2 ∆Eq
(10)
∆E q =
K3 ∆Eq
+ K4 ∆δ
(11)
∆V t =
K5 ∆δ + K6 ∆Eq
(12)
∂E q ∂E q ∂P e ∂P e ∂V t ∂V t , K3 = , K4 = , K2 = , K5 = , K6 = ∂δ ∂Eq ∂Eq ∂δ ∂δ ∂Eq
(13)
where: K1 =
These coefficients depend on the operating conditions. Thus, the linearized SMIB model with PSS can be represented in the statespace form by: X˙ = AX + BU, where: ⎡ ⎤ 0 0 0 ω0 ⎢ ⎥ K K ⎢ − K1 − D − 2 0 ⎥ ⎢ ⎥ M M M ⎢ ⎥ ⎢ A=⎢ K3 1 ⎥ K4 ⎥ 0 − − ⎢ ⎥ T d0 T d0 T d0 ⎢ ⎥ ⎣ K K K A K6 1 ⎦ A 5 − 0 − − TA TA TA (14) ⎤ ⎡ ⎤ ⎡ 0 ∆δ ⎢ 0 ⎥ ⎢ ∆ω ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ B=⎢ ⎥ , U = ∆U PSS ⎢ 0 ⎥, X = ⎢ ⎣ ∆E q ⎦ ⎣ K ⎦ A ∆E fd TA
Power system dynamic stability 
43
3 Optimization problem formulation In this study, the problem of tuning parameters of the PSS controller that stabilize the system is converted to a multi objective optimization problem. As given in [2, 17], two eigenvaluebased objective functions are considered. The ﬁrst one consists to shift the closedloop eigenvalues in to the leftside of the line deﬁned by σ = σ0 , as shown in Fig. 2(a). This function is expressed by J1 in equation (15). In equation (16), J2 deﬁnes the second objective function. It will place the closedloop eigenvalues in a wedgeshape sector corresponding to ξ ij ≥ ξ0 , as shown in Fig. 2(b) As consequence the maximum overshoot is limited. (σ0 − σ i ) (15) J1 = σ i ≥σ0
J2 =
(ξ0 − ξ i )
(16)
ξ i ≤ξ0
where σ i and ξ i are respectively, real part and damping ratio of the ith eigenvalue corresponding to an operating point. Therefore, the coordinated design problem is aimed to minimize simultaneously J1 and J2 by respecting the adjustable parameter bounds. So, the closedloop eigenvalues will be placed in the Dshape sector shown in Fig. 2(c). It is important to recognize that only the unstable or lightly damped electromechanical modes are relocated. The adjustable parameter bounds are given by the following equations: max K min PSS ≤ K PSS ≤ K PSS
(17)
min max T1P ≤ T1P ≤ T1P
(18)
min max T3P ≤ T3P ≤ T3P
(19)
The washout time constant T W and the two lead lag time constants of PSS are usually prespeciﬁed, as follows [2]: T W = 5s, T2P = T4P = 0.1s. Other constraint can be added, which is the electromechanical modes frequency limits [17]. ωmin ≤ ω ≤ ωmax
(20)
44  A. Farah et al. jω
jω
σi ≤ σ0 σ
ξi ≤ ξ0
jω
ξ0
ξi ≤ ξ0
σ
σi ≤ σ0
σ0
(a)
(b)
ξ0 σ σ0
(c)
Fig. 2. Region of eigenvalues location for different objective functions.
4 Implementation of the NSGAII approach Multi objective evolutionary algorithms which use no dominated sorting and sharing, such as, NSGA and NPGA (niched Pareto genetic algorithm) have been criticized for their, high computational complexity, absence of elitism and need for specifying the sharing parameter. Thus, an improved version of NSGA, called NSGAII [7, 8] is proposed in this paper. In this approach, the sharing function approach is replaced with a crowded comparison. Initially, an offspring population Q t is created from the parent population P t at the tth generation. After, a combined population R t is formed. Rt = Pt ∪ Qt
(21)
R t is sorted into different no domination levels F j . So, we can write: Rt =
r
Fj
(22)
j=1
To offer a higher precision with reduced CPU time, this proposed algorithm has been implemented using realcoded genetic algorithm [7]. So, a chromosome X corresponding to a decision variable is represented as a string of real values x i , i.e. X = x1 x2 ...x lch where, l ch is the chromosome size and x i is a real number within its lower limit a i and upper limit b i i.e. x i ∈ [a i , b i ]. A nonuniform arithmetic crossover is used. Thus, for two individuals having as chromosomes respectively X and Y and after generating a random number α ∈ [0, 1], the crossover operator can provide two chromosomes X and Y with a probability P C as follows: X = αX + (1 − α)Y (23) Y = (1 − α)X + αY
Power system dynamic stability 
45
Moreover, the nonuniform mutation operator has been employed. So, at the tth generation, a parameter x i of the chromosome X will be transformed to other parameter xi with a probability P m as follows: if τ = 0 x i + ∆(t, b i − x i ) (24) xi = x i + ∆(t, x i − a i ) if τ = 1 with:
∆(t, y) = y 1 − r 1 −
t gmax
β (25)
and where τ a binary is number, r is a random number and gmax is the maximum number of generations.
5 ANFIS approach ANFIS was originally suggested in [18], where the ANFIS architecture was presented to model nonlinear functions, identify nonlinear components control system and also, to predict a chaotic time series. The ANFIS is composed of ﬁve layers. A description of each layer is presented in [18]. The TakagiSugeno (TS) fuzzy rules are linear combinations of linear inputs [10–13]. Selection of initial number of membership functions is an important step in the ANFIS application [12]. In [19], the authors have determined this number by trial and error. They demonstrated that this method was not effective because it is based on a grid partition and it causes an explosion of the number of rules when the inputs number is large. So, they have proposed other method based on a clustering algorithm. The objective of clustering is to generate a concise representation of a system’s behavior by dividing the data space into clusters. Several clustering methods are used in literature [20]. In this paper, a procedure based on subtractive clustering algorithm [20] is used to generate the initial fuzzy inference system (FIS) structure.
6 Simulation results To evaluate the effectiveness and robustness of the proposed controller, their performances have been examined on a SMIB system given by Fig. 1. A 6cycle fault at the inﬁnite bus at the end of one transmission line was been considered. This fault has been cleared without line tripping. The system data is given in the appendix. The design process of this controller has been done on two steps, which are data preparation and training phases. All time
46  A. Farah et al.
simulations are carried out using nonlinear model. In this study, lower and upper limits of the controller gains KPSS are, 10 and 100. However, minimum and maximum values of T1P , T3P , are respectively 0.05 and 1.
1.8
8
δ [rad]
ANFIS PSS NSGAII PSS
1.7
x 10−3 Δω [pu]
ANFIS PSS NSGAII PSS
6
1.6
4
1.5
2
1.4 0
1.3
−2
1.2
−4
1.1 1 (a)
Time [s] 0
1.4
1
2
3
4
5
6
−6 (b) 0 5
Pe[pu]
ANFIS PSS NSGAII PSS
1.2
4
Time [s] 1
2
3
4
Efd[pu]
5
6
ANFIS PSS NSGAII PSS
3 1
2 1
0.8
0 0.6
−1
0.4
−2 −3
0.2 0 (c) 0
−4
Time [s] 1
2
3
4
5
6
−5 (d) 0
Time [s] 1
2
3
4
5
6
Fig. 3. (a): Rotor angle response at nominal loading. (b): Speed deviation response at nominal loading. (c): Electrical power response at nominal loading. (d): Field voltage variation at nominal loading.
6.1 Data preparation Firstly, the NSGAII algorithm is used to tune the PSS on a wide range of operating conditions, according to the objective functions deﬁned by equations (15) and (16). The generator real and reactive power outputs are ranged respectively, from 0.3 to 1.1 and 0.01 to 0.4 p.u. The collected inputoutput data pairs are stocked in a training set. Inputs are deﬁned by the operating conditions and outputs are the corresponding PSS parameters. In this application, the training set size is 1044 inputoutput data pairs.
47
Power system dynamic stability 
2.3
8
δ[rad]
2.2
ANFIS PSS NSGAII PSS
x 10−3 Δω [pu]
ANFIS PSS NSGAII PSS
6
2.1 2
4
1.9
2
1.8 1.7
0
1.6
−2
1.5 −4
1.4
Time [s]
1.3 0
(a) 1.4
1
2
3
4
5
6
5
Pe[pu]
ANFIS PSS NSGAII PSS
1.2
Time [s]
−6 (b) 0
1
2
3
4
Efd[pu]
5
6
ANFIS PSS NSGAII PSS
4 3
1
2
0.8
1 0.6
0
0.4
−1
0.2 0
(c) 0
−2 Time [s] 1
2
3
4
5
6
−3 (d) 0
Time [s] 1
2
3
4
5
Fig. 4. (a): Rotor angle response at heavy loading. (b): Speed deviation response at heavy loading. (c): Electrical power response at heavy loading. (d): Field voltage variation at heavy loading.
In order to evaluate the effectiveness of the NSGAII algorithm, the optimal parameters of the proposed controller corresponding to a three loading conditions given in Tab. 1, are shown in Tab. 2.
Tab. 1. Loading conditions. Loading condition
P [pu]
Q [pu]
δ0 [rad.]
Nominal loading Light loading Heavy loading
1 0.7 1.1
0.015 0.015 0.27
1.1871 0.8240 1.4234
Figures 3–5 depict the nonlinear time domain simulations at the loading conditions with a sixcycle fault disturbance. It can be clearly seen that the proposed stabilizer design able to provide a good damping characteristics.
6
48  A. Farah et al.
1.15
δ[rad]
4
ANFIS PSS NSGAII PSS
1.1
x 10−3 Δω [pu]
ANFIS PSS NSGAII PSS
3
1.05 2
1
1
0.95 0.9
0
0.85
−1 −2
0.8 0.75 (a)
Time [s] 0
5
1
2
3
4
Efd[pu]
4
5
6
1
ANFIS PSS NSGAII PSS
0.9
3
Time [s] 0
1
2
3
4
Pe[pu]
5
6
ANFIS PSS NSGAII PSS
0.8
2 1
0.7
0
0.6
−1
0.5
−2
0.4
−3
0.3
−4
0.2
Time [s]
−5 (c)
−3 (b)
0
1
2
3
4
5
6
Time [s]
0.1 (d) 0
1
2
3
4
5
Fig. 5. (a): Rotor angle response at light loading. (b): Speed deviation response at light loading. (c): Electrical power response at light loading. (d): Field voltage variation at light loading.
Tab. 2. Optimal parameters settings of the proposed stabilizer for various loading conditions.
Nominal loading Light loading Heavy loading
K PSS
T1P
T2P
T3P
T4P
32.1539 41.8617 39.5598
0.1373 0.1371 0.0952
0.1000 0.1000 0.1000
0.1253 0.1330 0.1633
0.1000 0.1000 0.1000
6.2 Training phase The initial (FIS) is trained using ANFIS, to reach the least possible error between the desired output and the FIS output through the training set. A combination of leastsquares and backpropagation gradient descent methods are used. Figure 6 shows real checking data and training data corresponding to K P SS. It can be clearly seen that the ANFIS output has a good approximation.
6
Power system dynamic stability 
100
49
Kpss
90 80 70 60 50 40 30 20
Cheking data Training data
Samples 0
20
40
60
80
100
120
Fig. 6. ANFIS prediction and checking data.
6.3 Validation of the proposed stabilizer To demonstrate the effectiveness of the proposed approach, closedloop system eigenvalues at three loading conditions are computed. 25 20
ANFIS eigenvalues NSGAII eigenvalues
15 10 5 0 −5 −10 −15 −20 −25 −20
−15
−10
−5
0
Fig. 7. Dshape sector at nominal condition.
Table 3 gives the electromechanical modes eigenvalues and their damping ratios without and with the proposed controller, at these operating conditions. It is clear that the open loop system is unstable because of the negative damping ratios. Figures 7–9 shows that all eigenvalues of the proposed stabilizer are in the Dshape sector and they are close to those obtained by the NSGAII design.
50  A. Farah et al.
Tab. 3. System electromechanical modes without and with control. (a): Nominal loading. (b): Light loading. (c): Heavy loading.
(a) (b) (c)
No control
NSGAII optimized stabilizer parameters
0.2949 ±j 4.9576, (−0.0594) 0.1026 ±j 5.3755, (−0.0191) 0.4786 ±j 4.3491, (−0.109)
−5.0454 ±j 2.5537, (0.8922) −4.9364 ±j 2.8004, (0.8698) −4.0462 ±j 1.8165, (0.9123)
25 20
ANFIS eigenvalues NSGAII eigenvalues
15 10 5 0 −5 −10 −15 −20 −25 −20
−15
−10
−5
0
−5
0
Fig. 8. Dshape sector at light condition. 25 20
ANFIS eigenvalues NSGAII eigenvalues
15 10 5 0 −5 −10 −15 −20 −25 −20
−15
−10
Fig. 9. Dshape sector at heavy condition.
Power system dynamic stability 
51
7 Conclusion In this study, an adaptive power system stabilizer based on adaptive network based fuzzy inference system (ANFIS) and second version of nondominated sorting genetic algorithms (NSGAII) is proposed. NSGAII is used to collect the training set composed by loading conditions and corresponding stabilizer parameters. Two eigenvaluebased objective functions are considered to place the closedloop system eigenvalues in the Dshape sector. The approach effectiveness is validated on single machine inﬁnite bus. The nonlinear simulation results and eigenvalue analysis for various operating conditions show that the proposed stabilizer is able to provide a good damping and improves the overall system performance on realtime. The proposed approach can be applied to real time stability for multimachine power system.
Appendix: System data Generator data: Generator
M = 9.26s X q = 0.55pu P e = 1pu
D=0 X d = 0.19pu Q e = 0.015pu
x d = 0.973pu = 7.76s T d0 δ0 = 67.61
Exciter data: Exciter
K A = 50
T A = 0.05s
Line data: Transmission line
R = −0.034pu G = 0.249pu
X = 0.997pu B = 0.262pu
Bibliography [1] [2]
[3]
P. Kundur. Power System Stability and Control. McGrawHill, 1994. E.V. Larsen and D.A. Swann. Applying power system stabilizer, Part I: General concept, Part II: Performance and tunning concepts, Parts III: Practical considerations. IEEE Trans. on Power system apparatus and systems, PAS100(6):3017–3024, June 1981. M. Kashki, M. A. Abido and Y. L. AbdelMagid. Pole placement approach for robust optimum design of PSS and TCSCbased stabilizers using reinforcement learning automata. Electrical Engineering, 91:383–394, 2010.
52  A. Farah et al.
[4] [5]
[6]
[7] [8] [9]
[10]
[11]
[12] [13]
[14] [15]
[16]
[17]
[18] [19]
[20]
A.D. Del Rosso, C.A. Canizares and V.M. Dona. A study of TCSC controller design for power system stability improvement. IEEE Trans. Power Systems, 18(4):1487–1496, 2003. Y. L. AbdelMagid and M. A. Abido. Robust Coordinated Design of Excitation and TCSCbased Stabilizers using Genetic Algorithms. Int. J. of Electrical Power & Energy Systems, 69(2–3):129–141, 2004. S. Panda, N.P. Padhy and R.N. Patel. Robust Coordinated Design of PSS and TCSC using PSO Technique for Power System Stability Enhancement. J. Electrical Systems, 3(2):109–123, 2007. K. Deb. A Fast and Elitist Multiobjective Genetic Algorithm: NSGAII. IEEE Trans. On Evolutionary Computation, 6(2):182–197, April 2002. N. Srinvas and K. Deb. Multiobjective function optimization using nondominated sorting genetic algorithms. Evolutionary Computation, 2(3):221–248, 1994. C. M Fonseca and P. J. Fleming. Genetic algorithms for multiobjective optimization: formulation, discussion and generalization. 5th Int. Conf. on Genetic Algorithms, San Mateo California, :416423, 1993. S.M. Radaideh, I.M. Nejdawi and M.H. Mushtafa. Design of power system stabilizers using two level fuzzy and adaptive neurofuzzy inference systems. Electrical Power and Energy Systems, 35:47–56, 2012. A.M.A. Haidar, M.W. Mustafa, F.A.F. Ibrahim and I.A. Ahmed. Transient stability evaluation of electrical power system using generalized regression neural networks. Applied Soft Computing, 11:3558–3570, 2011. S. R. Khuntia and S. Panda. Simulation study for automatic generation control of a multiarea power system by ANFIS approach. Applied Soft Computing, 12:333–341, 2012. S. R. Khuntia and S. Panda. ANFIS approach for SSSC controller design for the improvement of transient stability performance. Mathematical and Computer Modelling, 57:289–300, 2013. S. Panda. Differential evolutionary algorithm for TCSCbased controller design. Simulation modelling practice and theory, 17:1618–1634, 2009 H. Shayeghi, A. Safari and H.A. Shayanfar. PSS and TCSC damping controller coordinated design using PSO in multimachine power system. Energy Conversion and Management, 51(12):2930–37, 2010. V. Mukherjee and S.P. Ghoshal. Comparison of intelligent fuzzy based AGC coordinated PID controlled and PSS controlled AVR system. Int. J. of Electrical Power & Energy Systems, 29(9):679–689, 2007. H. Yassami, A. Darabi and S.M.R. Raﬁei. Power system stabilizer design using Strength Pareto multiobjective optimization approach. Electric Power Systems Research, 80:838–846, 2010. JS.R. Jang. ANFIS: Adaptivenetworkbase fuzzy inference system. IEEE Trans. On power Systems, Man & Cybernetics, 23(3):665–685, 1993. J. FraileArdanuy and P. J. Zuﬁria. Design and comparison of adaptive power system stabilizers based on neural fuzzy networks and genetic algorithms. Neurocomputing, 70:2902–2912, 2007 S. Chiu. Fuzzy Model Identiﬁcation Based on Cluster Estimation. J. of Intelligent & Fuzzy Systems, 2(3):267–278, 1994.
Power system dynamic stability 
53
Biographies Anouar Farah received the M.S. degree in Higher School of Sciences and Techniques of Tunis from Tunis I University, Tunisia, in 1996 and Master degree in 2010. He is currently a doctoral student in the National Engineering School of Sfax, Tunisia. His research interests power system stability, FACTS, optimization techniques applied to power system and intelligent control.
Tawﬁk Guesmi received the electrical engineering degree (1999) and the M.Sc (2002) and Ph.D. (2007) in electrical engineering from National Engineers School of Sfax, Tunisia. He is currently an associate professor in the biomedical Institute of TunisTunisia. His current research interests intelligent techniques applications in electrical power systems.
Hsan Hadj Abdallah, He received his Ph.D in electrical engineering from the Higher School of Sciences and Techniques of Tunis from Tunis I University, Tunisia, in 1991. He is currently a professor in the Department of Electrical Engineering of National School of Engineering of SfaxTunisia. His current research interests electrical power systems (EPS), the dispatching and the stability of EPS, wind energy, and intelligent techniques applications in EPS.
Abderrazak Ouali received his Engineering Diploma in Science physics from Tunis University in 1974. The Ph.D in automatic and informatics from 7 Paris University, in 1977. He is currently a professor in the Department of Electrical Engineering of National School of Engineering of SfaxTunisia. His current research interests control and electric power systems.
N. Khemiri, A. Khedher and F. Mimouni
A Sliding Mode Control Approach Applied to a Photovoltaic System operated in MPPT Abstract: In this paper, the modeling and control of a photovoltaic system is presented. A maximum power point tracking (MPPT) algorithm is adopted to maximize the photovoltaic output power. The proposed strategy is based on the sliding mode control is robust to environment changes (irradiance and temperature). Simulation results using Matlab/Simulink have shown good performances of the photovoltaic system operated in MPPT. Keywords: Photovoltaic system, MPPT, sliding mode control, environment changes. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction Used to produce electrical energy, renewable energy sources such as solar photovoltaic energy is fastgrowing. The photovoltaic generator is composed of a various number of solar cells connected like series and parallels. However, the performance of photovoltaic depends on irradiation, temperature and load impedance [1]. Several literatures deal with the problem concerning the search of optimal operating point by using some maximum power point tracking (MPPT) methods in order to extract the maximum energy from the PV modules [2–4]. Therefore, the basic structure of photovoltaic system presented can contain the following components: solar PV array, with a number of series/parallel interconnected solar modules, a DC/DC boost converter, a load and a control system, illustrated by Fig. 1. The DCDC boost converters are extensively used in photovoltaic generating systems as an interface between the photovoltaic array and the load, which allows tracking of maximum power point. There are several methods of PV conversion with maximum power point tracking (MPPT) [5, 6]. Some of the methods use perturbation and observation method [7], incremental conductance method [8], constant voltage method [9], backstepping technique and sliding mode control (SMC) [10].
N. Khemiri, F. Mimouni: University of Monastir, National Engineering School of Monastir (ENIM) Research unit ESIER, Monastir, Tunisia, emails: [email protected], [email protected] A. Khedher: University of Sousse, National Engineering School of Sousse (ENISo), Sousse, Tunisia. A. Khedher: University of Sfax, National Engineering School of Sfax (ENIS), Research unit RELEV, Sfax, Tunisia, emails: [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 55–66. DOI 10.1515/9783110448412004
56  N. Khemiri et al.
Sliding mode control offers a very good way to implement a control action which exploits the inherent variable structure nature of DCDC converters [11]. Sliding mode control is one of the effective nonlinear robust control approaches. This mode occurs on switching surface, and the system remains insensitive to parameter variations and disturbance. This paper is organized as follows. In the second section, we have modeled the photovoltaic system. In the third section, we have presented and modeled the DCDC boost converter. The fourth section is devoted to the sliding mode control for boost converter. In the last section, some simulation results are shown interesting obtained control performances of the PV in terms of the robustness against environment changes.
IL
T
Load E Vpv
α MPPT Control
Fig. 1. Basic structure of photovoltaic system.
2 Modeling of photovoltaic systems The electrical equivalent circuit of a solar cell is given by Fig. 2.
Ipv
IL
Ish
Rsh Vpv
Iph
Vd
Rsh
Fig. 2. Equivalent circuit model of PV.
Sliding mode control for photovoltaic systems

The voltage current characteristic equation of a solar cell is given by [12]: V pv + R s I pv V pv + R s I pv −1 − I pv = I ph − I s exp VT R sh
57
(1)
where R s is relatively small and R sh is relatively large, which are neglected in the equation in order to simplify the simulation. The obtained ideal equivalent circuit of the PV cell is given by Fig. 3 [13].
Id Iph
Ipv
Vd
Vpv
Fig. 3. Simpliﬁed electric model of a photovoltaic.
We have rewritten equation (1) as follows: V pv −1 I pv = I ph − I s exp VT
(2)
In this section, we study the inﬂuence of irradiance and temperature on the photovoltaic generator. The currentvoltage I pv (V pv ), the powervoltage P pv (V pv ) curves under different irradiance and temperature, are illustrated by Fig. 4 and Fig. 5, respectively. These ﬁgures show that the open circuit voltage is dominated by the temperature and irradiance has an inﬂuence on the short circuit current. We conclude that high temperatures and low irradiance will reduce the capacity of power conversion.
3 Mathematical model of boost DCDC converter Figure 6 shows the conﬁguration of a boost DCDC converter used to interface the photovoltaic generator with the load. The statespace averaged model of the boost converter is given by [13]: x˙ 1 = −a2 (1 − α)x2 + a2 V pv (3) x˙ 2 = −bx2 + a1 (1 − α)x1 where: x1 = I L , x2 = V s , a1 =
1 1 1 ,b= ,a = and α is the duty cycle. C 2 L pv RC
58  N. Khemiri et al.
10
× 104 1000 W/m2
160
1000 W/m2
9
140
8 800 W/m2 7
100
Ppv(W)
Ipv(A)
120
600 W/m2
800 W/m2
6 600 W/m2
5 80 4 60
(a)
40 0
3
100
200
300
500
400 Vpv(V)
600
700
800
(b)
2 100
200
300
400 500 Vpv(V)
600
700
800
Fig. 4. (a): Model I pv (V pv ), (b): Model P pv (V pv ) curves for various irradiation levels: (E = 1000, 800 and 600 W/m2 ; and T = 298°K).
160
12
× 104
11 140
10
283°K
9
120
8 Ppv(W)
Ipv(A)
298°K 100 323°K
283°K 7 298°K 6
80
323°K
5 4
60
3
(a)
40 0
100
200
300
400 500 Vpv(V)
600
700
800
900
(b)
2 100
200
300
400
500 Vpv(V)
600
700
800
900
Fig. 5. (a): Model I pv (V pv ), (b): Model P pv (V pv ) curves for various irradiation levels: (E = 1000W/m2 , and T = 283°K, 298°K and 323°K).
Ipv
Lpv
IL
Is C
Cpv
Vpv
k
α
R Vs
Fig. 6. Boost Converter Circuit.
The general form of the nonlinear time invariant system is given by: x˙ = f (x) + g(x)α
(4)
Sliding mode control for photovoltaic systems

59
4 Sliding mode control of boost converter The sliding mode control is insensitive to externals changes. In this study, we introduce the concept of the control. The sliding surface is chosen as follows [14]: s=
∂P pv ∂I pv
(5)
2 . The sliding mode is obtained by the following condition [14]: with: P pv = R pv I pv ∂R pv =0 (6) s = I pv 2R pv − I pv ∂I pv
In this condition, the system operates at maximum power point of the PV generator. Based on the observation of the sliding surface with respect to the duty cycle and power photovoltaic, given by Fig. 7, we ﬁnd that the sliding surface is negative for a big duty cycle and is positive for a small duty cycle.
104
11
dPpv/dlpv=0
10 9
s>0
s0
800
900 α 0 if s < 0
(7)
60  N. Khemiri et al.
So, we can deduce the expression of the control duty cycle α which is composed of two terms α eq and α n , of which who guarantee and depended in sign of can be regarded as an effort to reach the maximum power: α = α eq + α n
(8)
The calculation of requires the application of the approach studied by [15], the equivalent control is given according to the condition deﬁning the sliding mode. s˙ =
∂s ∂x
T x˙
(9)
The derivative of the sliding surface can be written:
∂s s˙ = ∂x
T [f (x) + g(x)α eq ]
(10)
Then, we deduce: α eq = −
∂s ∂x ∂s ∂x
T f (x) T
= 1− g(x)
V eq Vs
(11)
We choose the nonlinear law form by using the method of study by Gao [16]: α n = −ksn sign(s)
(12)
where k is a positive constant scalar and 0 < n < 1. The control α is:
α=
1 α eq + α n 0
if if if
α eq + α n > 1 0 ≤ α eq + α n ≤ 1 α eq + α n < 0
(13)
The SMC approach is based on the Lyapunov stability theory: s s˙ ≤ 0 The derivative of the sliding surface is given by: ∂2 R pv ∂R pv ˙s = 3 + I pv −a2 (1 − α)x2 + a2 V pv 2 ∂I pv ∂I pv
(14)
(15)
Sliding mode control for photovoltaic systems
V pv , we can write: I pv ⎧ ∂R pv ∂ V pv ⎪ ⎪ = = ⎪ ⎨ ∂I pv I pv I pv ⎪ ∂R pv ∂2 R pv ∂ ⎪ ⎪ = = ⎩ 2 ∂I pv ∂I pv ∂I pv

61
From R pv =
1 ∂V pv V pv × − 2 I pv ∂I pv I pv V pv 1 ∂2 V pv 2 ∂V pv × − 2 × +2 3 2 I pv ∂I pv I pv ∂I pv I pv
By (2), the PV voltage can be rewritten as follows: I ph + I s − I pv V pv = V T ln Is and:
⎧ ∂V pv ⎪ ⎪ ⎨ ∂I pv ⎪ ∂2 V pv ⎪ ⎩ 2 ∂I pv
= =
−V T
I0 0
0
i[1] < 0
min[1] ×PV,cond
(1 − min[1] ) ×PV,cond 0 0
min[1] ×PV,cond 0
0 0 (1 − min[1] ) ×PV,cond
72  M. Buschendorf et al.
For positive phase leg current, devices S1 and D2 are blocking and do not conduct current. For negative phase leg current, devices S2 and D1 are blocking. The distribution of the conduction losses amongst the semiconductor devices also depends on drive factor min[z] since this determines the actual phase leg voltage in relation to the maximum phase leg voltage.
2.2.2 Switching losses The switching losses are also computed as commonly done, by integrating the mean switching energy over one switching period, scaled by the ratio of the actual to nominal voltage, and multiplied by the switching frequency: 1 PV,sS = fswitch T1
t
0 +T 1
!
" u Eon (t) + Eoff (t) c(t) dt Unom
(6)
t0
1 PV,sD = fswitch T1
t
0 +T 1
Erec (t)
uc(t) dt Unom
(7)
t0
For an IGCT and its inverse diode, the switching energies are given in the device datasheet and scaled as described by (6) and (7). For IGBT modules, the switching energies Eon , Eoff , and Erec were determined by experiment. In switching tests the energies for different operating conditions were measured and a function E = f (U, I) was determined which approximates the relations. A general switching frequency cannot be given, because of the nondeterministic switch control. For this investigation, an average switching frequency of switch fswitch = 150 Hz is assumed.
2.2.3 Losses in clamp circuit Additionally to the conduction and switching losses of the semiconductors, the IGCT has a relevant amount of losses in the mandatory clamp as shown in Fig. 4. For the calculation the prior losses are those caused by the energy stored in the clamp inductivity, which has to be wasted every switching period within the clamp resistor. These losses are calculated as done in [11] per 1 PV,Cl = fswitch T1
t
0 +T 1
t0
LCl 2 il (t) dt. 2
(8)
IGCT and IGBT for MMC in HVDC Applications  73
LCl
Rcl uc
CSM
Lσ S1
D1
D2
S2
Dcl Ccl
il
Fig. 4. Clamp circuit for IGCT conﬁguration.
2.2.4 Thermal model Based on the determined semiconductor losses a thermal model is applied to calculate the junction temperatures of the semiconductor devices. It is adapted as described in [12] for thermal equivalent circuits and follows the physical cauer model, and is simpliﬁed by neglecting the thermal capacities and only using the thermal resistors. Hence, the model enables the calculation of the average junction temperatures. The model is not applicable for transient processes as instantaneous power variations. Thermal circuits for the IGCT and the IGBT including their inverse diodes are shown in Fig. 5. In the thermal circuits, the resistors represent the thermal interfaces between: junction and housing (SG), housing and heatsink (GK), and heatsink and environment (KU), from left to right. In the IGCT converter, the diode and switch are separately installed and the thermal resistors between housing and cooling are consequently also distinct. Because of the module housing, IGBT chips and diode chips are stronger thermally coupled and therefore the thermal resistors between housing and cooling are coupled and drawn as one single element. The current sources represent device power losses and the voltage source represent the temperature difference to a reference temperature. The junction temperature is the voltage drop across the current source. Pv,valve
Rth,SG,Valve
Pv,valve
Rth,GK,Valve
Rth,SG,Valve
Rth,KU Pv,Diode
Rth,SG,Diode
Rth,GK Pv,Diode
Rth,GK,Diode
Rth,KU
Rth,SG,Diode
cooling
Fig. 5. Stationary thermal models for a) IGCT PressPack and b) IGBT module.
cooling
74  M. Buschendorf et al.
2.2.5 Simulation parameters For this investigation, an IGCT for medium switching frequencies and a suitable inverse diode with 4.5 kV nominal voltage and high current rating was chosen. For comparison, a 4.5 kV IGBT module was chosen as well. The selected devices are – ABB Semiconductors, IGCT 5SHY 35 L 4500 [13] with, – Inﬁneon, inverse diode D1961SH [14] and – Mitsubishi, IGBT CM1200HC90R [15]. The simulation parameters are summarized in Tab. 3.
Tab. 3. Simulation parameters of the HVDC converter. Udc
fac
ωadc
LCl
CCl
RCl
600 kV
50 Hz
2πfac
3 μH
15 μF
1Ω
3 Losses and junction temperature of the HVDCMMC The simulation is made for one representative submodule and scaled up for the entire converter. This is justiﬁed since all submodules experience practically a similar loading. This is achieved by an algorithm which selects submodules for activation/deactivation with the goal of maintaining equal submodule capacitor voltages within a phase leg [16, 17].
3.1 Junction temperature Figure 6 shows the average junction temperature in dependence on the converter power for the two most heavily loaded semiconductors in the submodule: diode 2 and switch 2 for all realistic drive factors, 0 < min[z] > 1. It can be seen that the IGCT converter can achieve a maximum power of 2000 MW, whereas the IGBT exceeds its maximum junction temperature at approximately 1700 MW. The active semiconductor chip area is similar for both, which is achieved by simulating two IGBTs in parallel, a technique commonly found in the literature, for example in [18, 19]. It is apparent that, in the chosen application, the inverse diode is the limiting device for the IGCT converter as well as for the IGBT converter. The asymmetrical junction temperature is a result of the chosen drive factor which is much smaller than one to ensure safe operation.
IGCT and IGBT for MMC in HVDC Applications  75
ϑj in °C D2 (IGBT)
rectifier
160
D2 (IGCT)
inverter
140 120
IGBT2
IGCT2 D2 (IGBT)
100 80 60 40 20 0 3000
2500
2000
1500
1000
500
0
–500
–1000 –1500 –2000 –2500 –3000 P in MW
Fig. 6. Junction temperature as a function of the converter power; Q = 350 MVar for P < 1000 MW, Q = 500 MVar for P ≥ 1000 MW, fswitch = 150 Hz, Udc = 600 kV.
3.2 Semiconductor losses The reason for the higher possible converter power for the IGCT converter is superior thermal behavior and lower semiconductor losses.
3.2.1 Losses in different operating conditions The total semiconductor losses in dependence of the converter power are shown in Fig. 7. In rectiﬁer mode, semiconductor losses are higher for the IGCT converter compared to the IGBT converter. The slope of the curves are nearly the same, with only a vertical offset making IGCT losses 5.0% higher than for the IGBT converter. The reason for that is the almost same dependence between current and losses in both used inverse diodes, which outweigh the switch losses. Another interesting fact is noticeable in the inverter mode, where the slope of the IGCT curve is signiﬁcantly lower. At no load, the IGCT converter still has higher losses but as the power level increases, the losses for the IGCT converter fall below those for the IGBT converter starting around 900MW. This is due of the substantially lower conduction losses of the IGCTs, and will be explained in detail in the following section.
76  M. Buschendorf et al. Pv, normalized
inverter
rectifier 100% 80% 60% 40% 20%
IGBT
2000
1750
1500
1250
1000
750
500
250
0
–250
–500
–750
–1000
–1250
–1500
–1750
–2000
–2250
0%
P in MW IGCT
Fig. 7. Normalized semiconductor losses; Q = 350 MVar for P < 1000 MW, Q = 500 MVar for P ≥ 1000 MW, fswitch = 150 Hz, Udc = 600 kV.
3.2.2 Loss distribution Figure 8 shows the losses of one submodule at one speciﬁc operating point, speciﬁed by transferred power P = 850 MW, Q = 350 MVar, switching frequency fswitch = 150 Hz, and dcvoltage Udc = 600 kV, which are split into its components for the IGCT submodule as well as the IGBT conﬁguration.
%
rectifier inverter
S2 S1
100
D2 D1 S2
Clamp Eoff
S1
75
S2 S1 D2
Eon
D1
50 S2
conduction losses S1
25
S1
0 IGCT
IGBT
IGCT
IGBT
Fig. 8. Semiconductor losses of a MMC submodule divided in its components; P = 850 MW, Q = 350 MVar, fswitch = 150 Hz, Udc = 600 kV, normalized to IGCT losses in rectiﬁer operation.
IGCT and IGBT for MMC in HVDC Applications 
77
Figures 9 and 10 show in style of Fig. 8 the losses for the operating points P = 1700 MW, Q = 500 MVar, fswitch = 150 Hz, Udc = 600 kV and P = 2000MW, Q = 500 MVar, fswitch = 150 Hz, Udc = 600 kV where the result for the IGBT in Fig. 10 is only theoretical because of exceed junction temperature, as mentioned in Section 3.1.
%
rectifier
S2 S1
inverter
D2 D1 S2 S1 S2 S1 D2 D1
100
75
50 S2
Clamp
Eoff Eon
conduction losses
25 S1
0 IGCT
IGBT
IGCT
IGBT
Fig. 9. Semiconductor losses of a MMC submodule divided in its components; P = 1700 MW. Q = 500 MVar, fswitch = 150 Hz, Udc = 600 kV, normalized to IGCT losses in rectiﬁer operation.
% rectifier
S2 S1
inverter
D2 D1 S2 S1 S2
100
S1 D2
75
Clamp Eoff Eon
D1
50 S2
conduction losses
25 S1
0 IGCT
IGBT
IGCT
IGBT
Fig. 10. Semiconductor losses of a MMC submodule divided in its components; P = 2000 MW, Q = 500 MVar, fswitch = 150 Hz, Udc = 600 kV, normalized to IGCT losses in rectiﬁer operation.
78  M. Buschendorf et al.
As expected from Fig. 6, diode 2 for rectiﬁer mode and the IGCT 2 / IGBT 2 for inverter mode are the most stressed devices in a submodule. The second interesting effect is the distribution between conduction and switching losses. The IGCT conﬁguration has substantially lower conduction losses compared to the IGBT conﬁguration at the considered operating point. The switching losses for IGCT, especially for the IGCT inverse diode, are distinctly higher. Additionally the losses in the mandatory clamp circuit increase the overall losses for the IGCT converter. Finally, the overall semiconductor losses for an ACDCAC transmission system (including both inverter and rectiﬁer losses) depending on converter power is shown in Fig. 11. Losses given by peripheral components, such as cooling, electronics, transmission losses, etc., are not taken into account.
105%
105%
105%
100%
100%
100%
95%
95%
95%
90%
90%
90%
85%
85%
85% 80%
80%
80% IGCT 850
IGBT 850
IGCT 1700
IGBT 1700
IGCT 2000
IGBT 2000
Fig. 11. Total semiconductor losses as a sum of inverter and rectiﬁer losses; fswitch = 150 Hz, Udc = 600 kV, a) P = 850 MW, Q = 350 MVar, b) P = 1700 MW, Q = 500 MVar, c) P = 2000 MW, Q = 500 MVar.
For transmission of 850 MW, as shown in Fig. 11 a), the IGBT implementation of the converter has nearly 4% lower semiconductor losses compared to the IGCT. This is the breakeven point between IGCT and IGBT converter losses in inverter operation mode. Below this amount of converter output power the IGBT system has superior efficiency. Figure 11 b) is the point of maximum output power for the IGBT converter (1700 MW), and here the IGBT solution has nearly 4.5 % higher losses. This shows the advantages of the low conduction losses of IGCT at high currents, despite the higher losses in the inverse diode and the additionally losses in the clamp circuit. This effect increases up to 6 % by increasing the output power up to 2000 MW in Fig. 11 c) (where the IGBT already has a too high junction temperature, which could in practice be solved by using a more powerful cooling system). The breakeven point between IGCT and IGBT regarding all semiconductor losses is approximately 1200 MW for the investigated devices.
IGCT and IGBT for MMC in HVDC Applications 
79
4 Conclusions In this paper, a mathematical model for the MMC was described, and for the semiconductor losses and the junction temperature, which was then used for comparative analysis. Converter losses and junction temperatures for an MMC of a hypothetical HVDC transmission system were investigated, for which suitable high power semiconductors were selected. 4.5kV IGCTs and 4.5kV IGBT modules were chosen for the MMC submodules. The comparison of junction temperatures, semiconductor losses and loss distributions indicate that for high power, high current transmission systems the IGCT is an attractive alternative compared to IGBT modules. The IGCT converter enables up to 4% higher efficiency (for high current ratings), and higher power throughput capability for equally rated semiconductors. Especially when HVDC transmission lines are used to strengthen a heavilyloaded AC grid, as is done in San Francisco by Trans Bay Cable, the momentary power throughput is usually above 70% of capacity. In cases such as this, the use of IGCTs offers a signiﬁcant advantage. However, IGBT converters have lower losses in applications which usually work at low power conditions, or with strongly ﬂuctuating power like in connections to wind parks or photovoltaic ﬁelds.
Nomenclature uC[z,m] uSM ukl[1]...[6] Udc uuN , uvN , uwN iSM i[1]...[6] Idc Ip Iq Uac ωac fac fswitch Lz S1 S2 D1 D2 m n z
Capacitor voltage of the mth capacitor in the phase leg z Submodule voltage Voltage on the submodule terminals Direct current voltage Phase voltages at the ac side Current trough the submodule Phase leg currents Direct current Amplitude of the real part of the phase current Amplitude of the imaginary part of the phase current Amplitude of the alternating current voltage Angular frequency of the alternating current Frequency of the alternating current Average switching frequency Phase inductor Upper IGBT in the submodule conﬁguration Lower IGBT in the submodule conﬁguration Upper diode in the submodule conﬁguration Lower diode in the submodule conﬁguration Momentary submodule Count of submodules Momentary phase leg
80  M. Buschendorf et al.
mon[z] min[z] t PV,cond t0 T1 ucond icond PV,sS PV,sD Eon Eoff uc Unom Erec PV,Cl LCl il CCl DCl RCl Lσ CSM P Q ϑj ϑj,max Rth
Count of “switched on” submodules in the phase leg z Drive factor Time Conduction losses Starting time for integration Time period of integration Semiconductor voltage during conduction Semiconductor current during conduction Switching losses of IGBTs Switching losses of diodes Turn on energy Turn off energy Voltage of the submodule capacitor Nominal voltage of semiconductor Reverse recovery energy Clamp losses Clamp inductance Load current Clamp capacitor Clamp diode Clamp resistor Stray inductance Submodule capacity Converter output power Converter reactive power Junction temperature Maximum junction temperature Thermal resistor
Bibliography [1] [2] [3] [4] [5] [6] [7]
A. Kumar, D. Wu, and R. Hartings. Experience from First 800 kV HVDC Test Installation International Conference on Power Systems, Bangalore, India, 2007. M.P. Bahrman and B.K. Johnson The ABCs of HVDC Transmission Technology. IEEE Power & Energy Magazine, Vol 5. No. 2, March/April 2007. J. Dorn, H. Huang, D. Retzmann. Novel voltage source converters for HVDC and FACTS applications. Conf. CIGRE Symposium, Osaka, CDRom, 2007. S. Henry, A.M. Denis, P. Panciatici. Feasibility study of offshore HVDC grids. IEEE Power and Energy Society General Meeting, MN, 2010. J. Dorn, H. Huang, D. Retzmann. A new Multilevel VoltageSourced Converter Topology for HVDC Applications. Cigre, Paris, 2008. S.S. Fazel, Investigation and Comparison of MultiLevel Converters for Medium Voltage Applications. PhD thesis, TU Berlin, Berlin, 2007. R. Marquardt, A. Lesnicar, J. Hildinger. Modulares Stromrichterkonzept für Netzkupplungsanwendungen bei hohen Spannungen. ETGFachtagung, Bad Nauheim, 2002.
IGCT and IGBT for MMC in HVDC Applications  81
[8] [9]
[10]
[11] [12] [13] [14] [15] [16]
[17]
[18]
[19]
P.K. Steimer, O. Apeldoorn, B. Ødegård, B. Bernet, T. Brückner. Very High Power IGCT PEBB technology. Power Electronics Specialists Conference, Recife, Brasil, 2005. S. Bernet, R. Teichmann, A. Zuckerberger, P.K. Steimer. Comparison of HighPower IGBT’s and HardDriven GTO’s for HighPower Inverters IEEE Trans on industry applications, Vol. 35, No. 2, 1999 S. Rohner, J. Weber, S. Bernet. Continuous Model of Modular Multilevel Converter and Experimental Veriﬁcation. IEEE Energy Conversion Congress & Exposition (ECCE2011), Phoenix, Arizona, USA, September 2011. S. Tschirley. Automatisierte messtechnische Charakterisierung von 10kV Integrierten Gatekommutierten Thyristoren (IGCTs). PhD thesis, TU Berlin, Berlin, 2007. A. Wintrich, U. Nicolai, W. Tursky, T. Reimann. Applikationshandbuch Leistungshalbleiter ISLE Verlag, Germany, 2010. ABB Semiconductors IGCT 5SHY 35 L 4500. datasheet, ABB Semiconductors. Inﬁneon Technologies AG Schnelle beschaltungslose Diode D1961SH datasheet, Inﬁneon Technologies AG. Mitsubishi Electric Corporation High Voltage IGBT CM1200HC90R datasheet, Mitsubishi Electric Corporation. S. Rohner, S. Bernet, M. Hiller, R. Sommer. Modulation, Losses, and Semiconductor Requirements of Modular Multilevel Converters. IEEE Transactions on Power Electronics, 57(8): 2633–2642, 2010. H. Akagi, M. Hagiwara, R. Maeda. Theoretical Analysis and Control of the Multilevel Cascade Converter Based on DoubleStar ChopperCells. The 2010 International Power Electronics Conference, Tokyo, Japan, 2010. J. Nelson, G. Venkataramanan, B. Beihoff. Investigation of parallel operation of IGBTs. Conference Record of the Industry Applications Conference, 37th IAS Annual Meeting, vol. 4, pp. 2585–2591, 2002. R. Alvarez, K. Fink, S. Bernet. Simulation and experimental investigation of parallel connected IGBTs. Industrial Technology (ICIT) IEEE International Conference on Industrial Technology, Viña del Mar, Chile, 2010.
Biographies Martin Buschendorf received his Dipl.Ing degree in electrical engineering from the Technische Universität Dresden, Germany, in 2011. He is currently working on his PhD thesis in the area of characterization of power semiconductor devices for HVDC applications. His main ﬁelds of research include the construction of a test bench, the calculation of stray inductances for dcbusbars and measurements of the switching behavior of power IGBTs.
82  M. Buschendorf et al. Jens Weber received the Dipl.Ing. and Dr.Ing. degrees in electrical engineering from the Technische Universität Dresden, Germany, in 1999 and 2007, respectively. He is currently with the Professur Leistungselektronik, Elektrotechnisches Institut, Technische Universität Dresden, where he leads an industryfunded research project dealing with power supplies for high output voltages. He is also involved in several projects funded by the German Federal Ministry of Education and Research (BMBF) or industry. His research interests include the modelling of power electronics circuits, nonlinear phenomena in switchedmode power supplies and nonlinear control of power electronic systems. Steffen Bernet received the diploma degree from Dresden University of Technology in 1990 and the Ph.D. degree from Ilmenau University of Technology in 1995, both in electrical engineering. During 1995 and 1996, he worked as Postdoc in the ECE Department of the University of Wisconsin Madison. In 1996, he joined ABB Corporate Research, Heidelberg (Germany) where he led the Electrical Drive Systems Group. From 1999 to 2000 he was subprogram manager responsible for the ABB research in the areas “Power Electronics Systems”, “Drives” and “Electric Machines”. From 2001 to 2007 he was Professor for Power Electronics at Berlin University of Technology. Since June 2007 he has been Professor at Dresden University of Technology. During the past twenty years, Dr. Bernet has conducted comprehensive research on power semiconductors, static power converters and ac motor drives.
H. Kouki, M. Ben Fredj and H. Rehaoulia
Modeling of Double Star Induction Machine Including Magnetic Saturation and Skin effect Abstract: The aim of this paper is to present a d − q accurate mathematical model of the double star induction machine (DSIM), whatever the electrical shift between the two stator stars. The proposed model takes into account the effect of magnetic saturation, the stator mutual leakage inductance between two stars and the skin effect. Obtained simulation and experimental results conﬁrm the validity and the performances of the proposed model, especially at startup. Keywords: Double star induction machine, modeling, main ﬂux saturation, stator mutual leakage inductance, skin effect. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction Machines with more than three phases have been applied for years in industrial applications, with their performances considered primarily in the high power ﬁeld [1–3]. The main advantage of increasing stator phase number is that on the one hand it allows the reduction of the size of the components in power modulators of energy, and on the other hand it provides better tolerance as well as greater reliability [4–6]. Among multiphase machines, the double star induction machine whose angular shift between the two stars 0°, 30° or 60° is the most popular in industrial applications [7]. Such machine, in addition to the power segmentation and redundancy it carries, has the advantage of reducing torque pulsations, rotor losses and current harmonics [8, 9]. In fact, the introduction of magnetic nonlinearities in the electrical equations has always been a topical issue for polyphase machines. Indeed, taking into account the saturation is not simply dictated by the desire to improve the results, but it can sometimes be a necessity [10, 11]. The presence of the mutual leakage inductance between two stars is due to the fact that their windings share the same slots [12–14] and are therefore mutually coupled. The mutual leakage coupling depends on the winding pitch and the angle shift between the two stator winding sets. Nevertheless, there have been some studies where the stator mutual leakage coupling has been neglected [15–17].
H. Kouki, M. Ben Fredj and H. Rehaoulia: University of Tunis, Institute of Sciences and Technology of Tunis, Electrical department, Tunis, emails: [email protected]; [email protected]; [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 83–95. DOI 10.1515/9783110448412006
84  H. Kouki et al.
Moreover, if the machine is fed by a voltage inverter, the rotor currents contain some speciﬁc harmonics. The frequency of these harmonic currents can have very high values. Consequently, the current distribution is not uniform in the rotor bars, which is known as the skin effect phenomenon. The skin effect increases the rotor resistance and decreases the rotor leakage inductance. We can use the Foster model to describe the dependency of rotor impedance with the effects induced by the variation of frequency. Foster models are usually referred to lumped parameter models RL circuits obtained by frequency identiﬁcation measures. Parameters variations can be determined using correction factors for the leakage inductance and resistance. Nevertheless, other studies provide analytical expressions for the resistance and leakage inductance versus rotor slip, which makes it easier to take into account the skin effect [18]. Furthermore, various models have been developed to study the behavior of electrical machines. There are current models and mixed models. In the present study, we have applied the model of the current taking into account the nonlinearity of the magnetic circuit and the skin effect. In this paper, the transient performance of the double star induction machine during startup has been determined. Finally, the validation of the model has been achieved by an experimental application.
2 Test and design machines Experimental results are obtained by means of the test bench of Fig. 1, where necessary requirements for machine supply and signals measurements are provided.
Fig. 1. Test bench.
Modeling of double star induction machine 
85
Figure 2 shows the development panoramic coil of two stator windings of the learning machine, which are spread over 24 slots. Frontal connections of winding 1 of phase 1 are heavy solid line, those of phase two and phase three are respectively strong dotted then strongly mixed lines. Concerning frontal connections of winding 2 of phases 1, 2 and 3 are in continuous, dotted and mixed lines. The angle θ mec represents the mechanical shift between the two stars, which belongs to the set {0/P, 30/P, 60/P} with P is the number of pole pairs.
Fig. 2. Panoramic development of stator windings of DSIM, 0.5 kW, 4 poles, 24 slots and two beams by slot. (i): winding 1. (ii): winding 2. (iii): mechanical shift between the two stars.
86  H. Kouki et al.
3 Transient model of double star induction machine As shown in Fig. 3, the machine has two stator windings sets (a, b, c) and (a , b , c ) spatially shifted by α, with isolated neutral points and an equivalent threephase squirrelcage rotor. Angles θ and (θ − α) denote the rotor position respectively to star 1 and 2. For this machine, we adopt the following assumptions: – Stator windings are sinusoidally distributed, – windings are identical within each three phase set.
Star 2
Vb
Va ̓
ib
ia ̓ Θ̓
Star 1
α
̓
Va
Vc
Vb
ia
V ̓c ic
ic ̓
ib ̓
Fig. 3. Dual star induction machine.
Characteristic equations of a DSIM in a common reference frame are: ⎧ ⎪ dλ ⎪ ⎪ v s1 = R s i s1 + s1 + ω a λ s1 ⎪ ⎪ dt ⎪ ⎨ dλ s2 v s2 = R s i s2 + + ω a λ s2 ⎪ dt ⎪ ⎪ ⎪ ⎪ dλ r ⎪ ⎩0 = Rr ir + + (ω a − ω m )λ r dt
(1)
with: J dω m × = T em − T load P dt
(2)
Modeling of double star induction machine 
dθ a dθ ωa = is the reference speed, and ω m = is the electrical rotor speed. dt dt Flux expressions are: ⎧ ⎪ λ = (l s + l sm )i s1 + l sm i s2 + λ m ⎪ ⎨ s1 λ s2 = l sm i s1 + (l s + l sm )i s2 + λ m ⎪ ⎪ ⎩λ = l i +λ r
r r
87
(3)
m
where: λm = Lm im
(4)
i m = i s1 + i s2 + i r
(5)
As shown in Fig. 4 and Fig. 5, in the electric equivalent circuit of DSIM, the stator mutual leakage inductance is included in the common branch of the two stars [19]. It is more convenient to separate the space vector equations in d − q ones. The d−axis equations are: ⎧ ⎪ ⎨λ ds1 = (l s + l sm )i ds1 + l sm i ds2 + L m i dm λ ds2 = l sm i ds2 + (l s + l sm )i ds2 + L m i dm (6) ⎪ ⎩ λ dr = l r i dr + L m i dm and the q−axis equations are: ⎧ ⎪ λ = ⎪ ⎨ qs1 λ qs2 = ⎪ ⎪ ⎩λ = qr
ids1
Rs
ls
(l s + l sm )i qs1 + l sm i qs2 + L m i qm l r i qr + L m i qm
ωa λqs1 lsm
ids2
Vds1
Rs
ls
(7)
l sm i qs2 + (l s + l sm )i qs2 + L m i qm
ωm λqr
l'r
ωa λqs2 idm
Lm Vds2
Fig. 4. daxis equivalent circuit of DSIM in an arbitrary reference frame.
R'r g
88  H. Kouki et al.
iqs1
Rs
ls
ωa λds1 lsm
Rs
iqs2
ls
ωm λdr
l'r
ωa λds2 iqm
Vqs1
Lm Vqs2
R'r g
Fig. 5. qaxis equivalent circuit of DSIM in an arbitrary reference frame.
4 Introduction of magnetic saturation effects The adopted philosophy considers the self ﬂux of a coil as the superposition of a leakage’s ﬂux and a useful ﬂux traversing the iron. As a consequence, leakage inductances become constant. This implies that only the main ﬂux is subject to magnetic saturation. Static and dynamic mutual inductances are deﬁned respectively, as follows: ⎧ λm ⎪ = ⎨L m im (8) dλ ⎪ m ⎩L = mdy di m
d λm
λdm im
idm β idm
q
λqm iqm
Fig. 6. Flux and current magnetizing.
According to Fig. 6, the d − q components of the magnetizing ﬂux and current are: λ md = λ m cos β i md = i m cos β (9) λ mq = λ m sin β i mq = i m sin β with: Lm =
λ m λ md λ mq = = im i md i mq
(10)
Modeling of double star induction machine 
89
For modeling a saturated squirrel cage double star induction machine in d − q axis, the proposed method is based mainly on the winding model of the current. Since (i ds1 , i qs1 , i ds2 , i qs2 , i dm , i qm ) constitute the selected state variables, the rotor ﬂux and current and their time derivatives in the primitive dq set of Equation (1), must be written as their functions. The d − q components of (λ s1 , λ s2 , λ r ) are normally written in terms of the winding currents in the system of Equations (6) and (7). Derived stator and rotor linkage ﬂuxes in Equations 6 and 7 leads to the time derivative of the magnetizing inductance L m . The leakage inductances in Equation (3) are assumed to be constant, only the dλ m has to be described by means of the main ﬂux λ m is subject to saturation. Thus, dt winding currents. Now let’s write: dL m dL m di m × = dt di m dt From Equation (4), L m =
λm im
(11) #
and after deriving i m =
i2dm + i2qm , we get, respectively
(12) and (13): dL m L mdy − L m = di m im di m i dm di dm i qm di qm × × = + dt im dt im dt di qm di dm = cos β + sin β dt dt di qs1 di qs2 di ds1 di = cos β + sin β + cos β ds2 + sin β dt dt dt dt di qr di dr + cos β + sin β dt dt
(12)
(13)
Argument β is the angular position of space vector i m (or λ m ) with respect to daxis so that: β = arctan
i qm i dm
(14)
Finally, we have: dλ ds1 dt dλ qs1 dt dλ ds2 dt dλ qs2 dt
di qm di ds1 di di + l sm ds2 + L d dm + L dq dt dt dt dt di qs2 di qs1 di qm di dm = (l s + l sm ) + l sm + L dq + Lq dt dt dt dt di qm di ds2 di ds1 di dm = l sm + (l s + l sm ) + Ld + L dq dt dt dt dt di qs1 di qs2 di qm di dm = l sm + (l s + l sm ) + L dq + Lq dt dt dt dt = (l s + l sm )
(15) (16) (17) (18)
90  H. Kouki et al. di qm di dλ dr di di = −l r ds1 − l r ds2 + (l r + L d ) dm + L dq dt dt dt dt dt di qs1 di qs2 di qm dλ qr di dm = −l r − lr + L dq + (l r + L q ) dt dt dt dt dt
(19) (20)
with: L dq = (L mdy − L m ) cos β sin β
(21) 2
(22)
2
(23)
L d = L mdy − (L mdy − L m ) sin β L q = L mdy − (L mdy − L m ) cos β
Using the equation V = A X˙ + BX, with X is a vector formed by d − q components of the winding currents and X˙ its time derivative, matrices A and B are: ⎡
0 l s + l sm ⎢0 l s + l sm ⎢ ⎢l 0 ⎢ sm A=⎢ ⎢0 l sm ⎢ ⎣−l r 0 0 −l r ⎡ −c1 0 Rs ⎢c1 R d s 1 ⎢ ⎢0 −d R s ⎢ 1 B=⎢ ⎢d1 0 c1 ⎢ ⎣−R r a1 −R r −a1 −R r −a1
l sm 0 l s + l sm 0 −l r 0 −d1 0 −c1 Rs a1 −R r
0
Ld l sm L dq 0 Ld l s + l sm L dq 0 lr + Ld −l r L dq ⎤ 0 −ω a L m ⎥ ωa Lm 0 ⎥ 0 −ω a L m ⎥ ⎥ ⎥ ⎥ ωa Lm 0 ⎥ Rr −b1 ⎦ b1 Rr
L dq Lq L dq Lq L dq lr + Lq
⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦
(24)
(25)
with a1 = (ω a − ω m )l r , b1 = (ω a − ω m )(l r + L m ), c1 = ω a (l s + l sm ) and d1 = ω a l sm . We can notice that the elements of matrix A depend on the saturation. Moreover, it contains all kinds of magnetic coupling along the d−axis (L d ), q−axis (L q ), and the d − q axis (L dq ).
5 Introduction of skin effect Rotor resistance varies not only by thermal effects, but also by the skin effect. Indeed, skin effect effectively increases rotor resistance and decreases rotor leakage inductance. Therefore, in order to take into account the skin effect we generally use the following linear equations: $ s − sn (26) R r = R rdc + (R rl − R rdc ) 1 − sn
Modeling of double star induction machine 
X r = X rdc + (X rl − X rdc )
% s − s &2 n
1 − sn
91
(27)
where: R rl : rotor resistance at 50 Hz, determined from rotorblocked test. R rdc : rotor resistance at zero (dc value) X rl : rotor leakage reactance at 50 Hz, determined from rotorblocked test. X rdc : rotor leakage reactance at synchronous speed.
6 Experimental and simulation results In order to validate the developed dynamic model of the double star induction machine, an experimental study was carried out by using a 6phases, 500W, 220V/380V; 4poles. The machine is supplied directly from the power distribution network. Figure 7 gives a comparison between the measured and simulated torque values at runup from rest. The torque peak values computed from model with saturation and skin effect are higher than computed values using conventional model. Moreover, the time taken to reach the steady state for accurate model is very close to that of experimental results.
10 linear model experimental results
8
with saturation and sking effect
Tem (N .m)
6
4
2
0 t(s) –2 0
0.05
0.1
0.15
0.2
0.25
Fig. 7. Comparison between measured and simulated transient torque during runup.
Furthermore, Fig. 8 shows the comparison between simulation and experimental results of rotor speed in transient regime at startup on load operation. We also note that at startup, if the saturation and the skin effect are taken into consideration, the
92  H. Kouki et al.
rotor speed of the DSIM is closer to the measured rotor speed than it is if they are ignored.
300
ωm ( rd .s–1 )
with saturation and skin effect experimental results
200
linear model
100
t(s) 0 0
0.1
0.2
0.25
Fig. 8. Comparison between measured and simulated transient rotor speed during runup.
From the mechanical characteristics, we can conﬁrm that the impact of the saturation is clearer in transient regime and especially at startup. Figure 9 illustrates the comparison between simulation and experimental results of the stator current of DSIM for the same operation. It can be observed that there’s
5 with saturation and skin effect experimental results linear model
3
1 0 –1
–3 t (s) –5
0
0.1
Fig. 9. Stator current of DSIM during runup.
0.2
0.25
Modeling of double star induction machine 
93
a little difference at the ﬁrst peak value. Although this difference appears, the saturation model still gives more accurate results for the stator current than it is for the conventional machine model. However, skin effect phenomena induce variation in rotor parameters. Considering the skin effect into rotor parameters, this difference can be slightly reduced. We note that taking into account the magnetic saturation and skin effect improves the accuracy of predicting the performance of the machine in transient state. However, there are still differences between the results of simulations and measurements especially in the early peaks. This can be explained by the fact that we have ignored the effect of saturation in the leakage ﬂux.
7 Conclusion A dynamic model of the double star induction machine that takes into account the magnetic saturation, the stator mutual leakage inductance between two stars and the skin effect is proposed in this paper. A comparison between simulation and experimental results shows that the proposed model is closer to measurement results than in the linear model, this, is validated for mechanical and electrical characteristics of the machine. It is shown that if we take into account the magnetic saturation and the skin effect, this will improve the accuracy and the performances of the prediction of the double star induction machine and particularly in transient operation and at startup.
Bibliography [1]
[2]
[3]
[4]
[5]
Farag K. AboElyousr and G. H. Rim. Performance Evaluation of AC/DC PWM Converter for 12phase StandAlone PMSG with Maximum Power Extraction. Int. Conf. on Electrical Machines and Systems (ICEMS), Incheon, :82–87, December 2010. J.M. Apsley, S. Williamson, A.C. Smith and M. Barnes. Induction motor performance as a function of phase number. IEE Proc. Electric Power Applications, 153(6):898–904, November 2006. Z. Oudjebour. Stabilization by New control technique of the input DC voltages of ﬁvelevel diode  Clamped inverters. Application to double star induction machine. 2nd Int. Symp. on Environment Friendly Energies and Applications (EFEA), 2012, :541–544. A. Tessarolo. Beneﬁts of increasing the number of stator phases in terms of winding construction technology in high power electric machines. 5th IET Int. Conf. on Power Electronics, Machines and Drives (PEMD), :16, 2010. S. Williamson and S. Smith. Fault tolerance in multiphase propulsion motors. J. of Marine Engineering and Technology, 4(5):3–7, March 2004.
94  H. Kouki et al.
[6]
[7]
[8] [9] [10] [11]
[12] [13] [14] [15] [16] [17] [18]
[19]
G. K. Singh and V. Pant. Analysis of multiphase induction machine under fault condition in a phaseredundant ac drive system. Int. J. Electric Machines Power Systems, 28(6):577–590, June 2000. S. Guizani, F. Ben Ammar. The eigenvalues analysis of the double star induction machine supplied by redundant voltage source inverter. Int. Review of Electrical Engineering (I.R.E.E), 3(2), April 2008. M.T. Mohammad and J.E. Fletcher. Fivephase Permanent Magnet Machines, Advantages and applications. 5th IET Int. Conf. on Power Electronics, Machines and Drives (PEMD), :1–5, 2010. A. Khedher and M. F. Mimouni. Sensorlessadaptive DTC of double star induction motor. Energy Conversion and Management, 51(12):2878–2892, December 2010. H. Rehaoulia, H. Henao and H. Capolino. G.A.: Modeling of synchronous machines with magnetic saturation. Electric Power Systems Research, 77(5):652–659, July 2007. Tu. Xiaoping, A. Louis Dessaint, Roger Champagne and K. AlHaddad. Transient Modeling of SquirrelCage Induction Machine Considering AirGap Flux Saturation Harmonics. IEEE Trans. on Industrial. Electronics, 55(7):2798–2809, July 2008. G. K. Singh. Multiphase induction machine drive research  a survey. Electric Power Systems Research, 61(2):139–147, March 2002. E. Levi. Multiphase electric machine for variable speed applications. IEEE Trans. Industrial. Electronics, 55(5):1893–1909, May 2008. T. A. Lipo. A dq Model for Six Phase Induction Machines. Int. Conf. on Electrical Machines, :860–867, 1980. H. Razik. Modelling of double star induction motor for diagnosis purpose. Conf. on Electric Machines and Drives, :907–912, 2003. R. H. Nelson and P. C. Krause. Induction machine analysis for arbitrary displacement between multiple winding sets. IEEE Trans. Power Apparatus Systems, 93(3):841–848, 1974. V. Pant, G.K. Singh and S.N. Singh. Modeling of a multiphase induction machine under fault condition. IEEE Int. Conf. Power Electrical Drives, :92–97, Hong Kong, 1999. N. Erdogan, T. Assaf, R. Grise and M. Aubouroug. An Accurate 3phase Induction Machine Model Including Skin Effect and Saturation for Transient Studies. Electrical Machines and Systems, ICEMS 2003, 2:646–649. H. Kouki, M. Ben Fredj and H. Rehaoulia. Effect of the stator mutual leakage inductance for low and high power applications of dual star induction machine. Int. Review on Modelling and Simulations (I.RE.MO.S.), 5(2), April 2012.
Biographies Hajer Kouki received the B.S. degree in Electrical Engineering and Power Electronics from the ESSTT (Institute of Sciences and Technology) in Tunis, Tunisia, in 2005, the M.S. degree in electric systems from the ENIT (National Institute of Engineering of Tunis), in 2007. Since 2009, she’s an Assistant in ISSTEG in the Department of Electrical Engineering. Her main research interests are the modeling, simulation, of electrical machines and power electronics.
Modeling of double star induction machine 
95
Mouldi Ben Fredj received the master’s degree in electrical and electronic automatic in 1985 and the doctoral degree in april 1989, from the University of Science and Technology of Lille France. Since 2000 he is with the ESSTT (Institute of Sciences and Technology of Tunis). His main research interests are power electronics and control of electrical machines: modeling and simulation of association’s electrical machines static converters, renewable energy sources.
Habib Rehaoulia is a Professor of electrical engineering. He received the B.E. degree in 1978, the M.S. degree in 1980, the doctoral degree in 1983 and the habilitation degree in 2007, all from the ENSIT (National high school of engineers), University of Tunis, Tunisia. He joined the teaching staff of the ENSET (lately ENSIT) in 1978. During his career, he was on leave for several months at WEMPEC (University of Madison Wisconsin USA), ENSIEG (University of Grenoble France), “Electrotechnic Lab.” (University of Paris VI France), and CREA (University of Picardie France). His main research interests are analysis and modeling of electrical machines.
M. Lindner, P. Braeuer and R. Werner
Increasing the Torque Density of PermanentMagnet Synchronous Machines using Innovative Materials and Winding Technologies Abstract: Due to the progress of system integration, the increase of energy efficiency and device mobility it is worthwhile to investigate alternative materials for electrical machines and methods of producing electric windings. In this article the inﬂuence of innovative materials on the machine behaviour as well as the machine design are discussed. This includes FeCo, FeNi and SMC. For each of these materials design guidelines are deﬁned. Beyond this the screen printing technology – so far used since a couple of years especially for producing cheap windings for smallpower electrical motors – is generalized to a wide range of machine sizes. A possible beneﬁt over classic copper wire windings regarding torque density is determined. Keywords: screenprinting, softmagnetic material, ironcobalt, SMC, utilization, electric machines, manufacturing. Mathematics Subject Classiﬁcation 2010: 65C05, 62M20, 93E11, 62F15, 86A22
1 Introduction With electrical machines penetrating increasingly more domains of everyday life new demands have been raised for them. Drives for handheld devices have to be sized as small as possible to achieve a high mobility or some ﬂexibility to distribute the mass. Examples are hand tools for craftsmen, medical technology or household purposes. Even more important is to reduce the weight of moving motors. Lower driven masses require lower acceleration torque and hence yet lower motor mass. [1] Furthermore, the system efficiency can be increased. The main focus in this respect is the transportation sector. Electrical and hybrid cars, space and aircrafts but even conventional cars with hundreds of auxiliary actuators. [2] Further examples are conveying systems, stair lifts and robots. Conventional machine design technology limits the torque density, described by the utilization factor C = T/V with torque T and machine volume V, to a value of approx. 10 kN m−2 [3] for aircooled permanent magnet synchronous motors (PMSM). This magnitude results from a maximum Lorentz force product – air gap ﬂux density
M. Lindner, P. Braeuer and R. Werner: Chair of Electrical Energy Conversion Systems and Drives, Chemnitz University of Technology, 09107 Chemnitz, Germany, [email protected] De Gruyter Oldenbourg, ASSD – Advances in Systems, Signals and Devices, Volume 3, 2017, pp. 97–113. DOI 10.1515/9783110448412007
98  M. Lindner et al.
Bg times orthogonal stator current Isq . The achievable ﬂux density deeply depends on the material and conﬁguration of the magnets. However, another great inﬂuence is exerted by the magnetization curve of the soft magnetic material and the width of ﬂux conducting sections as teeth and yoke. The current limit depends on the maximum current density of the winding, hence it’s resistance and surface area, as well as the crosssection of the slots. The latter highly interrelates to the previously mentioned width of soft magnetic sections with radial ﬂux machines. Furthermore, to guarantee the winding temperature the tolerable current density depends on the electric loading via the motor’s electrical stress, which is deﬁned e. g. by the magnetic core loss and the thermal conductivity of the core. A magnitude rather inﬂuencing the volume is the geometry of the end winding, which produces unproﬁtable room decreasing the utilization factor. Summarized, conventional machine designs are mainly limited by the material properties and production constraints of the soft magnetic core and the windings. One possibility to avoid those constraints is to adopt alternative core materials, which will be investigated in section 2. A second approach directly deduced from the previous analysis utilizes a novel winding technology, which is described in section 3. The achievable improvements with both proposals are summarized in section 4. All calculations were performed analytically in order to achieve general predictions as well as direct knowledge about the inﬂuences of single parameters instead of special solutions numerical calculations deliver. Furthermore, well proven analytical models provide a quality regarding overall machine predictions as good as numerical approaches do – merely local values differ in higher quantity.
2 Inﬂuence of Innovative Materials 2.1 Material Selection Soft magnetic materials differ – beside other physical and mechanical properties – in their magnetic polarization J and speciﬁc loss p. These parameters strongly determine the behaviour of the electrical machine. The higher the polarization the smaller the required soft magnetic cross section with a constant magnetic ﬂux. This leads to a smaller overall machine volume or more design space for nonferrous sections as slots or permanent magnets. Furthermore, the lower the speciﬁc loss the higher the machine efficiency and the lower the rated temperature rise. The standard ferromagnetic material in electrical machine construction are sheets from an ironsilicon alloy (FeSi). [4] A much higher saturation polarization and lower speciﬁc loss can be achieved with ironcobalt (FeCo). Alloys from ironnickel (FeNi) have even lower loss but also suffer from lower saturation. In addition to the alloy different delivery forms and production processes can be considered. Soft magnetic
Innovative Materials and Winding Technologies  99
2.5 1.5
p in W/kg
J in T
10
FeSi FeCo FeNi SMC
2.0 1.0 0.5 0.0
50 Hz
400
1 kHz
8
300
6
200
4
100
2 0
1
10
100 H in A/m
1000
10000
(a) Magnetic polarization
0 0
1
2 0 B in T
1
2
(b) Speciﬁc magnetic loss at different frequencies
Fig. 1. Characteristics of the considered materials. FeSi: Trafoperm N4; FeCo: Vacoflux 50; FeNi: Permenorm 5000V5 (all Vacuumschmelze GmbH & Co. KG [7]); SMC: Somaloy 700 3P (Höganäs AB [8]).
composites (SMC) are powders typically made from pure iron [5]. They allow less eddy currents but cause higher hysteresis loss and lower saturation compared to sheets. Further details on those material properties and their origin can be found in [6]. Fig. 1 compares the described characteristics of magnetic polarization and speciﬁc loss based on representative grades. A superior behaviour show nanocrystalline materials. They have high saturation polarizations and very low speciﬁc loss. However, currently, they can only be produced in thin and brittle ribbons hardly applicable to machine construction. Therefore, they will not be considered consecutively.
2.2 Effects on the machine design The determination of materials’ effects on the utilization factor of PMSM is deduced from [3] and [9] and was further enhanced. In oder to reasonably compare the results minimizing the inﬂuence of disregarded parameters, as e. g. stray, several predeﬁnitions need to be established: (PD 1) The main geometries and parameters as outer machine diameter and length, inner diameter, the rotor as well as the number of slots, poles and windings remain unchanged. (PD 2) The air gap is the highest magnetic resistance in the magnetic circuit. (PD 3) From (PD 1) and (PD 2) follows that the permanent magnets remain in the same operation point. To support this assumption the magnetic ﬁeld is kept constant in the stator sections of the magnetic circuit. (PD 4) From (PD 1) and (PD 3) follows that the magnetic ﬂux and induction in the air gap are constant.
100  M. Lindner et al.
(PD 5) The tooth ﬂanks are parallel. (PD 6) Armature reaction, thus effects of the stator on the rotor, are neglected. (PD 7) The rated temperature rise of the winding is constant. The calculation method discussed consecutively works with normalized machines. Instead of establishing precise numerical quantities the coefficients of modiﬁed values are determined. Thus, the results are transferable to most machine sizes. Fig. 2 shows the applied calculation sequence.
Reference PMSM with FeSi
Material properties
Tooth/yoke induction
Tooth/yoke crosssection
Tooth/yoke mass
Slotcrosssection
Thermal resistances
Tooth/yoke coreloss
I2R loss, Current
Torque, Utilization factor
Fig. 2. Concept chart of calculating the change of utilization factor with new ferromagnetic materials.
A magnetization curve as seen in Fig. 1a allows to directly gain the ratio of magnetic inductions B/BFeSi at constant ﬁeld strength according to (PD 3). This can be done both for the stator tooth and yoke. With constant air gap ﬂux Φg , (PD 4), the crosssections have to change consequently. A AFeSi
=
Φg BFeSi = · B Φg
B BFeSi
−1 at
H = const.
(1)
This directly applies to the ratio of tooth widths wt /wt, FeSi and yoke heights hy /hy, FeSi with constant machine length according to (PD 1). Thus, the stator mass changes considering additionally the varying mass density of the materials.
Innovative Materials and Winding Technologies  101
With (PD 1) changing the tooth width and yoke height modiﬁes the slot width ws and height hs and hence the crosssection of copper ACu at constant slot ﬁll factor, which is inverse to the ohmic resistance of the stator winding Rs . With a trapezoidal slot this is: −1 Rs ACu As ws hs = ≈ = · . (2) Rs, FeSi ACu, Fesi As, Fesi ws, FeSi hs, FeSi By means of the loss data in Fig. 1(b) ratios of the speciﬁc core loss for the new material (Pc /Pc, FeSi ) can be achieved taking into account the change of induction in tooth and yoke. Furthermore, with different geometries and materialspeciﬁc thermal conductivities the thermal resistances of the stator sections change, e. g. of the yoke above the slot: −1 Rth ys hy λth ws = · · . (3) Rth ys, FeSi hy, FeSi λth, FeSi ws, FeSi The thermal behaviour of the considered machines can be estimated with the lumped parameter thermal network in Fig. 3. In doing so the heat ﬂow across the air gap and in axial direction is neglected for the beneﬁt of lightweight relationships. This is reasonable because of few expected inﬂuences due to (PD 1), a high thermal resistivity of air and the axially consistent structure. [10]
Rthys
Rthyr Pcyt
Pcys
Rthsy
Rtht
P12Ry P12Rt Rthst ʋs
P12R
Pct
Fig. 3. Simple lumped parameter thermal network.
The keynote to thermal considerations is based on (PD 7), thus a constant slot temperature with changing core power losses and thermal resistances. Since the same
102  M. Lindner et al.
slot materials are supposed a maximum insulation temperature is crucial to machine design. With the assumptions of (a) homogeneous loss distribution in the core sections and (b) equal speciﬁc heat transfer through slot side and bottom a circuit analysis leads to the ratio of I2R loss (PI2R /PI2R, FeSi ). The thermal resistance of the slot was estimated for enamelinsulated round wires with rated diameters of 1 mm. [11, 12] Finally, the new phase current Iph results from the change of ohmic loss and resistance. With (PD 4) the torque T scales proportionally. The same applies for the utilization factor C with (PD 1). ' −1 Iph T Rs PI2R C = = = · (4) CFeSi TFeSi Iph, FeSi PI2R, FeSi Rs, FeSi Most machine parameters can be directly calculated. However, the determination of utilization variation requires assessing three quantities. The frequency f as well as the ﬁeld strengths in yoke Hy and tooth Ht have to be set in order to identify changes in the material characteristics Fig. 1. Furthermore seven ratios of the reference FeSi machine are required. The tooth to slot width (wt, FeSi /ws, FeSi ), yoke to slot height (hy, FeSi /hs, FeSi ) and yoke height to yoke diameter (hy, FeSi /dy, FeSi ) help to deﬁne the change of slot area. Additionally, both the tooth core loss to I2R loss (Pc t, FeSi /PI2R, FeSi ), the yoke core loss to I2R loss (Pc y, FeSi /PI2R, FeSi ), the I2R loss dissipating over the tooth ﬂank to overall I2R loss (PI2R t, FeSi /PI2R, FeSi ) and the slot height to slot width (hs, FeSi /ws, FeSi ) are needed to estimate the change of I2R loss. All those parameters were varied reasonably in order to ﬁnd dependencies of the utilization factor. The examined ranges are shown in Tab. 1. Finally, rules of thumb about the optimal usage as well as the expected changes are developed and mentioned in section 4.
Tab. 1. Variation of reference machine parameters. min f /Hz Ht /A m−1 Hy /A m−1 wt, FeSi /ws, FeSi hy, FeSi /hs, FeSi hy, FeSi /dy, FeSi hs, FeSi /ws, FeSi Pc t, FeSi /PI2R, FeSi Pc y, FeSi /PI2R, FeSi PI2R t, FeSi /PI2R, FeSi
max
50 1000 500 10, 000 500 10, 000 0.5 1.5 0.5 1.5 0.03 0.075 0.5 1.5 0.1 1.0 0.1 1.0 0.3 0.9
Innovative Materials and Winding Technologies  103
3 Inﬂuence of Innovative Winding Technologies 3.1 Conventional Windings Beside new materials also new technologies especially for the design of the winding can be useful to increase the torque density of electrical machines. In classic terms electric motors and generators with windings comprised of copper or aluminium are known. These are produced in a wide range of variants. Especially energy converters with distinct phase windings, commutator windings, windings with distinct poles and tooth coils should be mentioned [9], [13]. The winding is typically produced by coiling isolated copper wire. Depending on the type of winding, the process of coiling is one of the most expensive steps during fabrication of electrical machines. Furthermore, beside permanent magnets the winding is the temperature limiting element of an motor because of its isolation.
3.2 Screen Printed Windings The screen printing process has its origin in the printing industry. Since the 1960s it is used for printing conductive, insulating and resistive layers in the electronics industry [14]. The development in the area of printing forms as well as the availability of new material systems and more precise screen printing machines in the last years are the reason for distinct better producible layers and structures. Actually, screen printing is used for printing aerials for RFID smart labels [15]. Since at least two years it is possible to utilize this procedure in drive engineering. Thus, a reduction of the layer thickness of the winding as well as a reduction of costs are conceivable. The ﬁrst sample of windings for electrical energy converters printed at the Institute for Print and Media Technology at Chemnitz University of Technology (pmTUC) and tested in our laboratory was a threephase stator winding for a permanent magnetic synchronous motor described in [16] and [17]. Its excitation is done by a rotor made of NdFeB magnets. The layout for this machine consists of two conductive layers printed with silver paste and two insulating layers printed with a dielectric paste. Both conductive silver layers are connected through an interlayer connection. In Fig. 4 an example of a screen printed winding is shown. Currently, the windings were printed with a semi automatic screen printing machine with an optical positioning system. A PET plastic ﬁlm with a thickness of 50 μm is used as substrate but also PEN and PEEK or ceramic foils and smaller thicknesses are conceivable. The drying of the printed layers has taken place at a temperature of 150 °C. The screen used has a cloth of 120 threads per centimetre with a denier of 34 μm. Its mesh size is 45 μm and the rake angle 22.5 °. As substrate DuPont Teijin Melinex 401, a 50 μm PET plastic ﬁlm, was used. All
104  M. Lindner et al.
layers were printed with a print speed of 100 mm/s and a squeegee pressure of 1 bar. For the conductive layers the silver paste DuPont 5029 was used and dried for 5 min at 120 °C with a belt drier. All dielectric layers were printed with DuPont 5018G and dried with UV light at approximately 6.5 J/cm2 .
Fig. 4. Samples of screen printed airgapwindings of different three phase synchronous smallpower electrical machine.
3.3 Effects on the machine design Used instead of classic windings screen printed windings have a much smaller geometry to reach similar ampere turns. The reason for this is the high current density of up to 100 A/mm2 the windings can be used with. The reason which makes this possible is the proportion of crosssectional area to surface that is responsible for a much better heat dissipation. Another advantage of screen printed windings is the reduction of construction volume for the end winding. It is possible to reduce its volume nearly completely. In Fig. 5 the step of assembling a screen printed winding to a conventional stator back iron can be seen. It is usable with linear and rotary machines as well. After assembling, the end windings can be turned down. Therewith nearly no space is necessary for the end windings in axial machine direction. Figure 6 is showing the construction volume of the end windings for different winding technologies. Especially compared with distributed windings much space for the end windings can be economized. Another beneﬁt is the better utilization of space in the slot without the use of special preformed windings.
Innovative Materials and Winding Technologies  105
tooth
back iron
screenprinted winding
Fig. 5. Assembling process of a screen printed winding to a conventional stator back iron. It can be used for linear and rotary machines.
ω
ω
(a)
(b)
Fig. 6. Reduction of necessary space for the winding head by using screen printed windings (b) in comparison to classic windings (a) without preformed windings.
Beyond the mentioned beneﬁt the screen printed windings are mechanically as robust as known from conventional copper windings. Once installed and ﬁxed with glue the winding is not able to move caused by electromotive force in any direction. In this paragraph a calculation should show the advantages of screen printed windings in comparison to classic distinct phase windings and tooth coils. In order to have an objective comparison the number of windings ws and the nominal current IN are kept constant in all variants. Starting point for the calculations should be the
106  M. Lindner et al.
power dissipation and thermal behaviour of the screen printed winding. Table 2 shows all known parameters. The resistivity of the screen printed winding is an experimental value and depends on the screen printing process as well as the precise geometry of the printed conductors.
Tab. 2. Characteristic Winding Parameters. parameter
Symbol
value
speciﬁc electric resistivity speciﬁc electric resistivity speciﬁc electric resistivity
ρAg ρCu ρAg(sp)
1, 587 · 10−2 Ωmm /m 2 1, 678 · 10−2 Ωmm /m −2 Ωmm2 72, 6 · 10 /m
rated temperature rise rated temperature rise
ϑEndCu ϑEndAg(sp)
155 K 300 K
thickness of silver paste thickness of galvanic copper thickness of dielectric widths of conductor distance between conductors
hAgsp) hCusp) hdielectricsp wsp asp
0.015 mm 0.010 mm 0.010 mm 2.000 mm 0.200 mm
space factor of copper
φCu
2
0.6
Because of the higher permissible rated temperature rise the losses of the screen printed winding can be up to 1.94 times higher than this of classic copper windings under condition of an equal surface and speciﬁc heat dissipation. To handle this high temperatures a ceramic foil is used as a substrate. The surface responsible for heat dissipation is nearly the same with screen printed windings in comparison to classic copper windings. Beyond this, screen printed windings have an up to 10 times better speciﬁc heat dissipation because of their ﬁxing with heat transfer paste in the slot as well as the end windings folding of the back iron. Therewith, good heat dissipation and a reduction of construction volume of the winding head in comparison to classic windings especially distinct phase windings are possible. With help of Fig. 7, equation 5 gives the crosssectional area of a screen printed conductor in comparison to a classic copper wire. After the printing process the structures were electroplated with copper. Therewith it is possible to reduce the resistance appreciable. hAg(sp) + hCu(sp) · wsp Asp = = 0.2 (5) ACu 0.25 mm2 The crosssectional area of the copper depends on the diameter of the conductor that is varied between 0.25 mm2 and 0.75 mm2 [12]. For all following calculations the
Innovative Materials and Winding Technologies  107
Height: hAgsp + hCusp
Silverpaste
Ceramicfoil
Distance: a Width: wsp Fig. 7. Schematic of the dimensions of the screen printed windings.
worst case was used. Therewith, the relation between the resistance of screen printed windings and classic copper wire windings can be calculated with equal conductor length. Equation 6 shows that the resistance of the screen printed windings is about 12 times higher than this of classic windings. ⎛ ⎜ ⎜ ⎝ Rsp = RCu
⎞−1 1 ρAgsp ·
lsp AAgsp
1
+
ρCu ·
ρCusp · lCu ACu
lsp ACusp
⎟ ⎟ ⎠ = 12.08
(6)
Against a background of the same phase current the relation of the ohmic losses is the same than this of the resistance. Now it is possible to calculate the necessary construction volume of the winding in the slot and furthermore the change in machine out diameter 7. Therefore the same geometric relations as in section 2.1 were used and are given in Tab. 3.
Tab. 3. Variation of Reference Machine Parameters. relation
value
hy/dy
0.03 . . . 0.075 0.5 . . . 1.5
hy/hs
Dsp = 1+2· DCu
−1 hy hssp Dr −1 · · 1+ hsCu hs hs
with: hssp Asp = hsCu ACu
(7)
108  M. Lindner et al.
Under the condition of a constant inner diameter Di , induction in the airgap Bg and nominal current Is , the torque T would also be constant. Therewith the torque density is calculated like in equation 8. C∼
1 Dsp DCu
2
(8)
Design parameters for the calculation of other screen printed windings with special geometries for smallpower electrical machines with airgap windings and magnetic bearings can be found in [18] and [19].
4 Results The increase in utilization factor of machines adopting the three examined materials FeCo, FeNi and SMC differ highly depending on the mentioned quantities and ratios. Extensive variation calculations were performed to ﬁnd the maximum realistic values: FeCo: Cmax = 165 % · CFeSi FeNi: Cmax = 110 % · CFeSi SMC: Cmax = 97 % · CFeSi . As stated, FeCo enables to utilize machines much higher then FeSi does. FeNi can also be lucrative. Only SMC does not allow any higher torque densities in the considered parameter ranges. This might be different at frequencies well above 1 kHz or with other machine types. However, all three materials can also lead to signiﬁcantly decreased torque densities or even infeasible designs with inappropriate parameters. FeCo is only little prone to this effect but SMC highly suffers from too high temperatures because of high core loss at low frequencies. Tab. 4 outlines parameter tendencies within the ranges given in Tab. 1 of a reference FeSi machine. They describe the requirements to facilitate highest torque density improvements when utilizing different core materials. Another method of increasing the torque density is the use of screen printed windings. According to the calculations in this article a reduction of volume of the electrical machine between 10 and 40 % is possible. The correct value is depending
Innovative Materials and Winding Technologies  109
Tab. 4. Necessary parameter tendencies of a FeSi machine to achieve a maximum utilization factor with different core materials.
Parameter tendencies of a FeSi machine
f Ht , Hy wt, FeSi /ws, FeSi hy, FeSi /hs, FeSi hs, FeSi /ws, FeSi dy, FeSi /hy, FeSi Pc t, FeSi /PI2R, FeSi Pc y, FeSi /PI2R, FeSi
Maximum utilization factor when FeSi exchanged with
low low high high low — low high
low low high high high — high high
high high low low — — high —
FeCo
FeNi
SMC
on the exact machine geometry as well as the layout of the screen printed windings. The resulting increase of torque density is Cmaxsp = 110...160 % · CCu . At least it could be declared that the use of screen printed windings has its maximum beneﬁt compared to machines with comparatively long teeth. Because of the higher possible working temperature, the ﬂat structure as well as a better heat dissipation the screen printed windings are thermal stable. Beyond this it has to be kept in mind that the efficiency of screen printed windings is worse to this of classic ones. But there are many interesting approaches for a solution to get equal resistances like in conventional windings. Combining both technologies can lead to signiﬁcant higher utilization factors. Both derivations change different parameters. Varying the core material calculates new crossareas of the soft magnetic sections leading to bigger slots and higher current with constant machine volume. Utilizing screen printed windings changes the slot height with constant current and soft magnetic crossareas resulting in smaller machine volume. Thus, both calculations can be superimposed resulting in utilization factors of up to Cmaxsp,FeCo = 160 % · 165 % · CCu,FeSi = 260 % · CCu,FeSi . An example of the geometry optimization by utilizing the mentioned technologies is given in Fig. 8.
110  M. Lindner et al.
(a) FeSi reference machine
(c) Screen printed windings
(b) FeCo as core material
(d) Screen printed windings and FeCo core
Fig. 8. Reduction of construction volume by using screen printed windings instead of classic windings and increase of slot dimensions by substituting FeSi with FeCo.
5 Conclusion This article shows that it is possible to increase the torque density and therewith the utilization factor of electrical machines in a wide range. Therefore the inﬂuence of new materials like FeCo, FeNi and SMC for the construction of the core is discussed. The biggest beneﬁt in utilization factor of up to 160 % could be achieved with the use of FeCo due to the higher possible inductions in the core sections. Therewith, more construction volume for the winding is possible and the torque can be increased based on a higher magnetomotive force. Beyond this, the inﬂuence of screen printed windings on the torque density is discussed. The biggest advantage of this technology is the small construction volume and the good thermal behaviour. Thus, the overall volume of the machine can be
Innovative Materials and Winding Technologies  111
reduced increasing the utilization factor up to 260 %. This is possible because the screen printed windings use a ceramic foil as a substrate and therewith allows a factor 1.94 higher possible winding temperature. Combined with a 10 times better heat dissipation the higher loss caused by the higher electrical resistance and higher current density is controllable. Combining both approaches can lead to utilization factors that are hardly realizable with traditional surfacecooled machines. This leads to new freedoms of design, smaller machines, higher torques and better cooling compared to conventional technologies. Acknowledgment: The authors would like to thank M.A. Maxi Bellmann from pmTUC. Only with her help we were able to use the screen printing technology for producing electrical windings. This research is funded within the European Social Fund (ESF) by the Free State of Saxony and the European Union.
Bibliography [1]
M. Lindner, R. Werner, Optimale Antriebsdimensionierung im mechatronischen Gesamtsystem ElektromotorGetriebeLast (in German), VDIBerichte 2138, Antriebssysteme 2011  Elektrik, Mechanik, Hydraulik in der Anwendung, 2011. [2] L.G. Cravero, Entwurf, Auslegung und Betriebsverhalten von dauermagneterregten bürstenlosen Motoren kleiner Leistung (in German), Dissertation, Ilmenau University of Technology, 2005. [3] M. Beier, Der Ausnutzungsfaktor moderner elektrischer Maschinen in Abhängigkeit charakteristischer Maschinenparameter (in German), Chemnitz University of Technology, 2011. [4] R. Tzscheutschler, H. Olbrich, W. Jordan Technologie des Elektromaschinenbaus (in German), Berlin, Germany: Verlag Technik, 1990. [5] L.O. Hultman, A.G. Jack, Soft magnetic composites  materials and applications, Proceedings on Electric Machines and Drives Conference IEMDC’03, 2003. [6] M. Lindner, Untersuchung von modernen Magnetkreismaterialien und Wicklungstechnologien für energetisch hocheffiziente Antriebsmotoren (in German), Chemnitz University of Technology, 2009. [7] VAC Vacuumschmelze GmbH & Co KG, Internal test certiﬁcates: Trafoperm N4, Vacoﬂux 50, Permenorm 5000V5, 2009. [8] Höganäs AB, SMC Datasheet Somaloy 700 3P, 2009. [9] G. Müller, K. Vogt and B. Ponick, Berechnung elektrischer Maschinen (in German), 6. Auﬂage Weinheim, Germany: WILEYVCH Verlag GmbH & Co KGaA, 2008. [10] W. Schuisky, Berechnung elektrischer Maschinen (in German), Wien, Austria: Springer Verlag, 1960. [11] G. Gotter, Erwärmung und Kühlung elektrischer Maschinen (in German), Berlin, Germany: Springer Verlag, 1954. [12] DIN (Hrsg.), IEC 6031701  Speciﬁcations for particular types of winding wires  Part 01: General requirements  Enamelled round copper wire, Beuth Verlag GmbH, 2009.
112  M. Lindner et al.
[13] D. Hanselman, Brushless Permanent Magnet Motor Design, 2. Edition Cranston, Rhode Island, United States of America: The Writers’ Collective, 2003. [14] H.J. Hanke (Hrsg.), Baugruppentechnologie der Elektronik (in German), Berlin, Germany: Verlag Technik, 1994. [15] M. Fairley (Hrsg.), RFID Smart Labels  A “How to” guide to manufacturing and performance for the label converter, Düsseldorf, Germany: Tarsus Publishing Ltd., 2005. [16] P. Bräuer, Th. Schuhmann, R. Werner, K. Weigelt, M. Hambsch Wicklungen für elektrische Energiewandler (in German), Deutsches Patent und Markenamt, Deutsche Patentanmeldung, 2010. [17] P. Bräuer, T. Schuhmann and R. Werner, Screen Printed Windings for SmallPower Electrical Machines, The 8th FranceJapan and 6th EuropeAsia Congress on Mechatronics, Yokohama 2010. [18] P. Bräuer, R. Werner, Herstellung von Wicklungen für rotierende Kleinantriebe mittels Siebdruckverfahren (in German), VDIBerichte 2138, Antriebssysteme 2011  Elektrik, Mechanik, Hydraulik in der Anwendung, 2011. [19] P. Bräuer, M. Bartscht, R. Werner, Reduzierung des Bauraums bei einem aktiven dreipoligen Radialmagnetlager unter Verwendung von siebgedruckten Wicklungen (in German), Proceedings on 8. Workshop Magnetlagertechnik ZittauChemnitz, 2011.
Biographies Mathias Lindner has studied electrical engineering concentrating on electrical energy at Chemnitz University of Technology and received his Diploma with distinction in 2009. After working as Project Engineer at Devotek AS in Kongsberg/Norway he returned to Chemnitz to take up an employment as research assistant at the Chair of Electrical Energy Conversion Systems and Drives of Prof. Dr.Ing. Ralf Werner at Chemnitz University of Technology. He has specialized in design and analysis of electrical machines, properties of conventional and innovative materials as well as system optimization.
Patrick Bräuer has studied mechatronics and micro technologies at Chemnitz University of Technology. He wrote his thesis with distinction and received the Diplom in 2008. Since then he has been a research assistant at the Chair of Electrical Energy Conversion Systems and Drives of Prof. Dr.Ing. Ralf Werner at Chemnitz University of Technology. He has specialized in the design and control of electrical drives, new technologies and materials, especially the usage of 2D/3D screen printing as novel production technology.
Innovative Materials and Winding Technologies  113
Ralf Werner has studied electrical engineering at the University of Technology in KarlMarxStadt and received his doctor’s degree in 1989. Afterwards, he was employed in several positions at Chemnitz University of Technology and at EAAT GmbH, specializing in magnetic bearings and electric drives to the point of physical limits. In 2005 he was appointed to the professorship of Electric Drives at the University of Applied Sciences Mittweida. Finally, since 2009 he has headed the Chair of Electrical Energy Conversion Systems and Drives at Chemnitz University of Technology.