These DPetrichenko VF PDF

Calculation and Simulation of Turbogenerators Using
Permeance Network. Optimization Application.

Dmitry Petrichenko
To cite this version:

Dmitry Petrichenko. Calculation and Simulation of Turbogenerators Using Permeance Network. Op-
timization Application.. Electric power. Ecole Centrale de Lille, 2007. English. <tel-00346766>
HAL Id: tel-00346766

https://tel.archives-ouvertes.fr/tel-00346766
Submitted on 12 Dec 2008
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
N o Ordre : 53
ÉCOLE CENTRALE DE LILLE
THESE
Présenté en vue d’obtenir le grade de
DOCTEUR
en
Génie Électrique
par
Dmitry PETRICHENKO
Doctorat délivré par l’École Centrale de Lille.
Titre de la thèse :
Contribution à la modélisation et à la conception

optimale des turbo-alternateurs de faible puissance
Soutenu le 10 juillet 2007 devant le jury d’examen :
Rapporteurs : L.GERBAUD Professeur au LEG (Grenoble)

B.VANDEPUT Professeur à l’Université d’Eindhoven
Examinateurs : P.BROCHET Professeur à l’Ecole Centrale de Lille
M.HECQUET MdC-HDR à l’IUT à Lille
V.KUZNETSOV Professeur au MPEI (Moscou)
D.LALOY Resp. R&D Jeumont Framatome
H.LOMONOVA Professeur à l’Université d’Eindhoven
Directeur de recherche : P.BROCHET Professeur à l’Ecole Centrale de Lille
Co-encadrant : M.HECQUET MdC-HDR à l’IUT A Lille
Thèse préparée dans le laboratoire, L2EP, à l’École Centrale de Lille.

Calculation and Simulation of Turbogenerators Using
Permeance Network.
Optimization Application.
Dmitry Petrichenko
22 May 2007
Preface
The presented work was realized in Ecole Centrale de Lille in the laboratory of Electrotech-
nics and Power Electronics (Laboratoire d’Electrotechnique et d’Electronique de Puisqnce,
L2EP) of Lille, team COME (Conception et Optimisation de Machines Electriques) with
participation of CNRT (Centre National de Recherche Technologique de la France) in col-
laboration with Jeumont Framatome Areva Nuclear Power, region Nord – Pas de Calais and
Moscow Power Engineering Institute.
Acknowledgments
I wish to express my grateful acknowledgments to professor Pascal BROCHET, the professor
of Ecole Centrale de Lille and the responsible of the COME team who hearty welcomed me
in his team, for his belief in me. His human qualities created a perfect team and working
among such people helped me a lot in my research.
I wish to express my gratitude to doctor Michel HECQUET, the associate professor
of l’IUT de Lille, for his valuable advices, extraordinary energy, powerful support and very
nice personality. Without his help and encouragement I wouldn’t have finished my work
and definitely wouldn’t have spoken french.
I wish to acknowledge professor Vyacheslav KUZNETSOV, the professor of the de-
partment of Electromechanics, Moscow Power Engineering Institute. His passion of science,
amicability, his personality made me choosing the way of researcher in electromechanics
after the studentship and led me through hardship to the desired result while working on
thesis.
I want to thank Mr. Patrick Brutsaert, engineer of the turbogenerator department
of Jeumont-Framatome for his valuable remarks, advices and sharing his experience during
my work.
I want to thank Mr. David BONNET (and Mr. Rith LY, the student-interns who made
valuable contribution to the development and validation of our approach and software.
I would like to express my hearted acknowledgments to Mr. Andre Vandenput, the
Professor of the Eindhoven University of Technology, Electrical Engineering Department,
and Mr. Laurent GERBAUD, the Professor of the National Polytechnic Institute of Greno-
ble, who do us the honor of participating in the jury as reporters.
I want to express my sincere acknowledgments to Mr. Daniel LALOY, the coordinator
of R&D department of electrotechnics and electronics laboratory of Jeumont-Framatome in
the equipments business unit of AREVA, Mr. Jean-Louis COUDERT and Mr. Damien
VERBEKE, the engineers of Jeumont-Framatome, for their valuable help.
I also thank all the administrative, technical and professor staff of Ecole Centrale de
Lille and my colleagues for perfect working conditions and support. Among them I would
like to thank personally Darius VIZIREANU, Frédéric GILLON, Bellemaine NAPAME,
Xavier CIMETIERE, Olivier BARRE for their help and support.
And finally I want to thank my wife, my mother, my relatives for the tonnes of patience
and firm belief in me.
To my mother Vera
To my wife Julie
To my aunt Kate
To my wife’s mother Elena

Abstract
Huge recent growth of electric energy consumption all over the world applies higher demands
to turbogenerator performances and characteristics. The activities of different turbogener-
ator manufactures are being invested in order to ameliorate the products and increase the
speed of project development.
An electrical machine is one of the most complicated parts of any electromechanical
converter and can be itself considered as a complex system characterized by magnetic, elec-
trical, wave, thermal, mechanical air- and hydrodynamic phenomena. For these reasons
the main attention of a scientific researcher has been focused on the development of reli-
able multiphysical and coupled models and methods of electrical machines simulation and
optimization.
For the purpose of turbogenerator optimization its model using parameterized electro-
magnetic coupled permeance network approach was created. Different phenomena like axial
heterogeinity of the machine were taken into account. The model and developed software
were tested in different cases to perform rapid calculation of the turbogenerator charac-
teristics and operating modes. The calculation core of the software was coupled with the
optimization tool to perform the machine’s parameter influence study.
The developed approach and the flexibility of the software give the possibilities to
further development and extension of the model by multi-physical phenomena, coupling
realized model with vibrational, acoustic and especially thermal ones.
Keywords
Multi-physical model
Turbogenerator
Electro-magnetic coupling
Permeance network
Optimization model
Contents
Introduction 1
1 Problem definition and overview 5

1.1 Existing methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 Conventional methods of calculations . . . . . . . . . . . . . . . . . . 8
1.1.2 Field calculation methods . . . . . . . . . . . . . . . . . . . . . . . . 15
1.1.3 Permeance network methods . . . . . . . . . . . . . . . . . . . . . . . 19
1.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Used Approach 25
2.1 Theoretical basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.1.1 Calculating magnetic fields using vector or scalar magnetic potentails 26
2.1.2 Magnetic charges approach . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.3 Coils in ferromagnetic systems . . . . . . . . . . . . . . . . . . . . . . 31
2.1.4 Idea of a tooth contour and special boundary conditions . . . . . . . 33
2.1.5 Magnetic charges and permeance network . . . . . . . . . . . . . . . 37
2.2 Machine’s permeance network . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2.1 General approach for zones’ division of the machine . . . . . . . . . . 40
2.2.2 Stator network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2.3 Rotor network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.2.4 Airgap characteristics calculation . . . . . . . . . . . . . . . . . . . . 49
2.3 Taking into account axial heterogeinity . . . . . . . . . . . . . . . . . . . . . 55
2.3.1 Steel zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.3.2 Air-gap zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.3.3 Slots zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.4 Permeance Network Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.4.1 PN equations in matrix form . . . . . . . . . . . . . . . . . . . . . . . 62
2.4.2 Tableau formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.4.3 Taking into account nonlinearities . . . . . . . . . . . . . . . . . . . . 66
2.5 Steady-state fixed-rotor model . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.5.1 Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.5.2 Characteristics calculation . . . . . . . . . . . . . . . . . . . . . . . . 69
2.6 Model taking into account rotation . . . . . . . . . . . . . . . . . . . . . . . 71
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
v
3 Software implementation 73
3.1 General Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2 Software structure and approach . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.2.1 Circuit Description Tools overview . . . . . . . . . . . . . . . . . . . 76
3.2.2 Matlab Solver general approach . . . . . . . . . . . . . . . . . . . . . 78
3.2.3 Matlab Solver structure . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.3 Graphical User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3.1 Program Main Window . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3.2 Setting up project data . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3.3 Model generation and its characteristic view . . . . . . . . . . . . . . 88
3.3.4 Calculating the modes and characteristics of turbogenerator . . . . . 89
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4 Application 99
4.1 Synchronous machine of 3kW . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.1.1 Machine description . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.1.2 Machine’s permeance network . . . . . . . . . . . . . . . . . . . . . . 101
4.1.3 Calculation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2 Two pole turbogenerator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.2.1 Machine description and its permeance network . . . . . . . . . . . . 107
4.3 Multi-pole turbogenerator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.3.1 Jeumont 4-pole turbogenerator . . . . . . . . . . . . . . . . . . . . . 114
4.4 Optimization application example . . . . . . . . . . . . . . . . . . . . . . . . 117
4.4.1 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Conclusion and perspectives 121
A Matlab Solver functions 125

A.1 General functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
A.2 Calculation functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
A.2.1 Mode calculation functions . . . . . . . . . . . . . . . . . . . . . . . . 126
A.2.2 Characteristic calculation functions . . . . . . . . . . . . . . . . . . . 126
A.3 Low-level calculation functions . . . . . . . . . . . . . . . . . . . . . . . . . . 127
A.4 Post-processing functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
B 2-pole turbogenerator result comparison 129

List of Figures
1 4-poles rotor and stator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 2-poles rotor of turbogenerator . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Synchronous model equivalent circuit . . . . . . . . . . . . . . . . . . . . . . 8

1.2 Synchronous model phasor diagram . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Equivalent circuit for Pottier diagram . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Potier diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Frame conversion diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Phasor diagram of a synchronous generator working on an active-inductive
load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7 Flux density in the air-gap of a big generator . . . . . . . . . . . . . . . . . . 14
1.8 Variation of xd and xq parameters . . . . . . . . . . . . . . . . . . . . . . . . 14
1.9 A 2D mesh of a synchronous machine and a 3D mesh of a claw-pole machine 20
1.10 Flux tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.11 Permeance network of a synchronous generator . . . . . . . . . . . . . . . . . 22
1.12 Different calculation approaches groups . . . . . . . . . . . . . . . . . . . . . 24
2.1 A part of magnetic system’s equivalent circuit in terms of SMP . . . . . . . 27

2.2 A part of magnetic system’s equivalent circuit in terms of VMP . . . . . . . 28
2.3 Magnetic shell and its potential. . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Magnetic charges field in air . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Forms of magnetic leaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6 Distribution of SMP and magnetic charges density for a coil with current . . 33
2.7 Stator and rotor tooth contours . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.8 Tooth contour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.9 A magnetic shell stretched on tooth . . . . . . . . . . . . . . . . . . . . . . . 38
2.10 Other configuration of magnetic shells. . . . . . . . . . . . . . . . . . . . . . 38
2.11 Magnetic shells in case of two layers of PN. . . . . . . . . . . . . . . . . . . . 39
2.12 Main zones of a machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.13 Characteristic zones of machine’s magnetic core . . . . . . . . . . . . . . . . 41
2.14 Zone definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.15 Full stator scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.16 Tooth scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.17 Tooth division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.18 The stator yoke element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.19 Concentric rotor PN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.20 2-pole rotor PN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
vii
2.21 Adaptive rotor PN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.22 TC field under SBC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.23 Tooth contour unipolar permeance calculation . . . . . . . . . . . . . . . . . 52
2.24 Finite element solution for air-gap unipolar permeance calculation . . . . . . 54
2.25 TC permeance calculation comparison . . . . . . . . . . . . . . . . . . . . . 54
2.26 Stator magnetic system cross-section in axial direction . . . . . . . . . . . . 55
2.27 Flux buging in the air of machine . . . . . . . . . . . . . . . . . . . . . . . . 56
2.28 BH curve modification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.29 Flux displacement to the ventilation duct area in case of high saturation . . 58
2.30 Field in the airgap of a machine . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.31 Field in the slot zone of the machine. . . . . . . . . . . . . . . . . . . . . . . 61
2.32 Synchronous machine vector diagram . . . . . . . . . . . . . . . . . . . . . . 68
2.33 Mutual chracteristics between stator and rotor teeth. . . . . . . . . . . . . . 72
3.1 General structure of TurboTCM . . . . . . . . . . . . . . . . . . . . . . . . . . 75

3.2 Interaction between TCMLib and Matlab . . . . . . . . . . . . . . . . . . . . 77
3.3 Collaboration between different parts of the software . . . . . . . . . . . . . 79
3.4 Input data specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5 Program main window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.6 Setting up rated data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.7 Stator parameters setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.8 Setting up stator slot parameters . . . . . . . . . . . . . . . . . . . . . . . . 84
3.9 Defining slot parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.10 Slot distribution window for the rotor . . . . . . . . . . . . . . . . . . . . . . 86
3.11 Rotor slot distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.12 Simulation parameters window . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.13 Calculation of excitation current . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.14 Air-gap characteristic viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.15 All characteristics viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.16 Operating mode calculation window . . . . . . . . . . . . . . . . . . . . . . . 92
3.17 Ampere-turns of PN elements (top) and air-gap flux density (down) . . . . . 93
3.18 Input data for no-load characteristic . . . . . . . . . . . . . . . . . . . . . . 94
3.19 Input data for load characteristic . . . . . . . . . . . . . . . . . . . . . . . . 95
3.20 Input data for regulation characteristic . . . . . . . . . . . . . . . . . . . . . 96
3.21 Input data for V-shaped characteristic calculations . . . . . . . . . . . . . . 96
3.22 Input data for angular characteristic . . . . . . . . . . . . . . . . . . . . . . 98
3.23 Results of angular characteristic calculation . . . . . . . . . . . . . . . . . . 98
4.1 Rotor dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.2 Steel BH-curves: rotor (left), stator(right) and shaft (bottom). . . . . . . . . 101
4.3 Rotor simplified scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4 OPERA-RM model (fragment). . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.5 Fluxes comparison: TurboTCM – OPERA RM . . . . . . . . . . . . . . . . 103
4.6 Noload (left) and load (right) characteristics: TurboTCM – OPERA RM . . 104
4.7 Dynamic calculation: TurboTCM – OPERA RM . . . . . . . . . . . . . . . 104
4.8 Experimental workbench . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.9 Static characteristics: comparison with experiment . . . . . . . . . . . . . . 106
4.10 Rotation effect: experimental comparison . . . . . . . . . . . . . . . . . . . . 106
4.11 Regulation characteristic of machine 1. . . . . . . . . . . . . . . . . . . . . . 111
4.12 V-shaped characteristic of machine 1. . . . . . . . . . . . . . . . . . . . . . . 111
4.13 Regulation characteristic of machine 1 and variation of its xd and xq parameters.112
4.14 No-load air-gap flux density distribution. . . . . . . . . . . . . . . . . . . . . 112
4.15 Rated mode air-gap flux density distribution. . . . . . . . . . . . . . . . . . 113
4.16 Short-circuit mode air-gap flux density distribution. . . . . . . . . . . . . . . 113
4.17 Rotor of the 4-pole turbogenerator. . . . . . . . . . . . . . . . . . . . . . . . 114
4.18 Taking into account rotor arms . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.19 4-pole rotor permeance network. . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.20 No-load characteristic comparison for 4-pole machine. . . . . . . . . . . . . . 116
4.21 KhB3 for Tp1 min (left) and KhB3 for Tp1 max (right) . . . . . . . . . . . . . 118
4.22 KhE3 (left) and KhE1 (right) for different Tp1 . . . . . . . . . . . . . . . . . . 119
4.23 If for Di1 min (left) and Di1 max for different Tp1 . . . . . . . . . . . . . . . . 119
B.1 No-load characteristic comparison: Machine 1. . . . . . . . . . . . . . . . . . 132

B.9 Short circuit characteristic comparison: Machine 1. . . . . . . . . . . . . . . 138
Introduction
The development of new methods of investigation of electromechanical systems, consisting

of different electromechanical converters, electrical, mechanical elements and semiconduc-
tor devices, and methods of representing magnetic fields and processes in electromechanical
converters are among the most important problems of contemporary electromechanics. His-
torically, these two problems have been rather separated from one another and different
methods have been used to solve them. This can be explained both by the diversity of the
phenomena that occur in these systems and the modest capability of the computer equip-
ment which has been used over many decades. The situation has changed. During the
last 20-25 years a lot of new mathematical methods for analyzing phenomena and processes
in electromechanical devices and systems have been developed and have become available.
The growth of computer technologies and calculation speed has become an important event
which made it possible the optimization problems to be shown for any objective function
and the behavior of electromechanical systems and their elements to be analyzed both under
nominal and extreme operating conditions.
An electrical machine is one of the most complicated parts of any electromechanical
converter and can be itself considered as a complex system characterized by magnetic, elec-
trical, wave, thermal, mechanical air- and hydrodynamic phenomena. For these reasons
the main attention of a scientific researcher has been focused on the development of reli-
able multiphysical and coupled models and methods of electrical machines simulation and
optimization.
The production and maintenance of turbogenerators currently constitute one of the
main activities of Jeumont S.A.1 . This activity is being invested in order to ameliorate
the products and increase the speed of project development. The following figures present
examples of 4-poles and stator (Figure 1) and 2-poles (Figure 2) rotors of turbogenerators
produced in Jeumont.
The main objective of that study was to create a rapid model (and a software) of a
turbogenerator which allows to calculate its characteristics and can be used together with
optimization software.
This work consists of several chapters:
Chapter 1 presents the problem definition and overview of existing approaches used
to calculate electrical machine characteristics. Describing conventional, numerical and com-
bined methods we define our solution based on permeance network approach in order to
obtain a rapid model which anyway can represent internal processes in an electrical ma-
chine in details.
In chapter 2 we thoroughly describe theoretical basis and realization of the approach
1
Jeumont S.A. is a part of Framatome ANP - AREVA group. ANP - Advanced Nuclear Power
1
2
Figure 1: 4-poles rotor and stator
Figure 2: 2-poles rotor of turbogenerator
that we use. The description of the permeance network, the method of its construction
and calculation of its parameters are given. The use of adaptive permeance network con-
3
struction give wide flexibility in implementation of different slot and yoke configurations.
The proposed method of taking into account axial heterogeinity of turbogenerators can be
spread on any type of machines having irregular axial structure; it can be used with any
other calculation technique besides permeance networks. The fixed rotor algorithm shows
high performance while calculating static characteristics of synchronous machines.
Chapter 3 gives the explanation of the way how our approach was realized in software,
called TurboTCM. The tool is written in Matlab with some core objects written in C++ using
efficient object-oriented approach and component object model technique. Splitting the
software on two levels – calculation core and graphical interface – gives the flexibility in use
both in project definition with general calculations and in using specific features. The last
allow extracting any available information about the internal networks from within Matlab.
That last feature helps coupling TurboTCM with any calculation software and user-defined
scripts in Matlab giving possibility to use the software with external optimization tools.
The description of the internal software structure is followed by graphical user interface
presentation.
Chapter 4 shows the results of model application to different machines. First the
method was tested on a small 3 kW synchronous machine and compared with finite elements
calculation. Then we present a 2-pole turbogenerator case with experimental validation.
The 4-pole turbogenerator, its particularities and model realization are shown later. And in
the end of the chapter we present the use of TurboTCM coupled with optimization supervisor
Sophemis performing some results of screening and influence factor studies.
Finally, the main conclusions are presented with the perspective future use, develop-
ment and application of our model and software.
4
Chapter 1
Problem definition and overview
This chapter gives a view on the aim of the thesis, the overview of several methods of
electrical machine calculation, the choice of the method of simulation or so-called “virtual
prototyping” applied to the case of turbogenerators.
The most part of the electrical machines are well-known and everything seems to be
already done and studied in this area.
The scientific and technological progress in such areas like electrotechnics, permanent
magnets, steels, composite and powder ferromagnetic materials, power electronics, convert-
ers, controllers delivers exceptional opportunities for further development and innovations
in electromechanics. The same situation can be found also in further development of the
structure of electrical machines and simplification and quality of their production process.
The objective of the project is to develop a turbogenerator calculation tool which can
be used as a part of an optimal design software.
This tool must be able to perform all possible steady-state analysis of turbogenerators.
For Jeumont it is necessary to obtain rapidly steady-state no-load and load characteristics
of a machine, calculate its parameters in any operating mode:
• Any steady-state operating modes given by U, I and Power Factor (PF) angle.
• Regulation, V-shape and Power-load angle characteristics.
• P-Q diagrams for different restrictions.
• Parameters and time constants, masses of active materials, losses and efficiency.
• Flux density distribution in the air-gap, stator and pole teeth, and stator yoke.
• Xd, Xq parameters for every calculated mode.
A modern turbogenerator is a complex, high-technological product which is described

by hundreds of parameters starting from the basic geometry and structure and finishing
by the material properties. Jeumont produces several series of turbogenerators of basically
2-poles and 4-poles construction for the power of 10-100MW. But also the production of
machines up to 10 poles is under consideration. This fact demands rather flexible tool and
mathematical model, which allows taking into account the special features of the turbogen-
erators, like:
5
6 CHAPTER 1. PROBLEM DEFINITION AND OVERVIEW
• complex geometry of the turbogenerator’s rotor, its tooth zone;

• complex geometry of the slots;
• complex distribution of the wires both in stator and rotor winding;
• winding cooling system;
• axial heterogeinity of a machine caused by stator and/or rotor ventilation ducts.
A lot of different methods which allow calculation of the characteristics listed above.
Some of them can give only the estimations, some of them provide the redundant informa-
tion. While creating a software for an engineer we should balance between them and use all
the possible knowledge to obtain a model not too general to be fault and not too precise to
be useless.
But first of all let’s take a glance on the existing methods of calculations.
1.1 Existing methods

The equations of electrical and magnetic circuits, which describe the processes in electrical
machines, are obtained from the electromagnetic field equations according to the several
assumptions.
The second Maxwell equation
→
−
→
− ∂ B
rot E = − , (1.1)
∂t
written in an integral form links together the electrical field in a machine’s winding and the
changing speed of the magnetic field linked with the winding. This equation can be simply
converged to the differential equations of voltages for separate electrical circuit contours
uk + ek = Rk ik , (1.2)
which link together the parameters of the electrical circuit’s contour k (ik – winding current,
uk – external voltage, Rk – winding’s active resistance, ek = −dΨk /dt – an EMF induced in
the winding) and magnetic circuit parameters – the winding flux linkage Ψk .
Contrary, the first Maxwell’s equation, which determines the relation between mag-
→
−
netic field (with the strength H ) and electrical field (with the winding’s current density
→
−
J ),
→ −
− →
rot H = J , (1.3)
written in an integral form, is much more difficult to be transformed into the magnetic
circuit equations
Φi = Λi Fi (1.4)
where Φi – magnetic flux; Fi – magneto-motive force; Λi – magnetic permeance of the i-th

branch.
The difficulties of that transform have the following reasons:
1.1. EXISTING METHODS 7
• the borders of the cores have rather complex form, for example, teeth zones of stator
and rotor;
• the geometry of the air-gap between stator and rotor is variable according to their
mutual movement;
• the magnetization characteristic of the stator and rotor cores are nonlinear.
Thus, in order to be able to calculate analytically the magnetic field in such a complex
object, a set of assumptions has been defined. The main of them are as follows:
• the processes of the electromagnetic energy transform are determined mostly using
the main harmonic of the air-gap flux density;
• the salient surfaces of the magnetic cores can be replaced with the non-salient air-gap,
which is equivalent for the unipolar field;
• the main field of a saturated machine (the field of the first harmonic of flux density)
can be considered as a sum of two independent fields – directed along longitudinal and
transverse axes;
• the field factors for the basic harmonics of the flux density along longitudinal and
transverse axes, found in linear approximation, can be applied to the saturated ma-
chine;
• the main field and the leakage fields of a saturated machine can be considered inde-
pendently;
• the fields of the higher and lower spacial harmonics of flux density can be considered
independently both on each other and on the main flux density harmonic.
The assumptions listed above, were formulated more than 70 years ago during the
period of the basic creation of the electrical machine theory and were well presented in the
fundamental monograph of R. Richter.
These assumptions allowed to divide the area of the field of an electrical machine into
a set of zones with rather simple borders’ geometry which does not depend on the rotor’s
movement (for example, an air-gap zone on a half-period of the first harmonic of flux density;
zones corresponding to the slots, teeth and half-period of the field in the stator and rotor
yokes). Using the magneto-motive force (MMF) makes it possible to calculate the magnetic
field in such zones analytically and to find analytical equations determining the magnetic
permeances of the specified branches Λi .
Thus, these assumptions helps to transform the description of the magnetic field in
the machine in the form of Maxwell’s equations (1.1), (1.3), fulfilled with the boundary
conditions and complementary equations, into the description of the processes in the machine
by means of its electrical and magnetic circuits (1.2), (1.4). Based on these equations and
these assumptions a lot of different analytical methods of electrical machines calculations
were developed.
1.1.1 Conventional methods of calculations

The methods developed during the period of basic development of the electrical machine
theory, as it has been shown above, were based on rather strong assumptions, but even
nowadays they are widely used in the design practice. The behavior of electrical machines
both in steady-state and transient regimes is estimated using these methods. All of them are
based on a determination of special parameters of the machines, which are used to calculate
the characteristics of the last ones later on. For the moment there is a big variety of
analytical methods of electrical machine calculations but here only some models concerning
synchronous electrical machines will be discussed.
General methods
The synchronous reactance model is the simplest one. This model can be applied only for the
first estimation of the behavior of machines with rather low level of saturation. Additional
assumptions are used in this model:
• magnetic circuit saturation is neglected;
• hysteresis and eddy currents effects are neglected;
• electromotive forces (EMF) are supposed to be proportional to the excitation current;
• the machine is supposed to have non-salient poles.
Figure 1.1 shows the equivalent electrical circuit of the machine according to this
model. The circuit is presented for one stator phase.
Figure 1.1: Synchronous model equivalent circuit
This model allows using the following equation in a complex form:
Ė1 − R1 I˙1 − jx1 I˙1 − V˙1 = 0 (1.5)
R1 – active resistance of a stator phase;
x1 = xa + x1σ – full synchronous reactance of a stator phase, where Xa – stator winding

reactance, X1σ – stator leakage reactance;
Ė1 – EMF of a stator phase;

I˙1 – stator current;

V˙1 – external phase voltage.
This scheme corresponds to the following phasor diagram (Figure 1.2):
Figure 1.2: Synchronous model phasor diagram
The EMF Ė1 is determined from the linear no-load characteristic. The model rep-
resents separate MMFs induced by stator and rotor windings (F˙f and F˙1 respectively),
while in reality there is only a full magnetic field in the machine produced by two windings
simultaneously.
To take into account the saturation a real no-load curve must be used. Anyway, in this
case the influence of slot, differential and end-winding leakages is neglected. That means
that the value of x1σ is considered linear and not depending on the magnetic state of the
machine. This can be easily assumed for differential and end winding leakages but not so
evident for the slot one.
x1σ = xσslot + xσdif f + xσend (1.6)

Figures 1.3 and 1.4 show the equivalent circuit and the phasor diagram respectively,
representing that one approach. In this case the EMF source Ėδ represents the EMF caused
by the interaction of both stator and rotor MMF-s: F˙f and F˙1 . Ėδ is so-called “EMF beyond
the leakage”.
The diagram 1.4 is also called Pottier diagram and is widely used to calculate the
excitation MMF (thus, field winding current) for the particular operating mode. Knowing
the value of x1σ (1.6) and the values of the mode: V˙1 , I˙1 and cosφ, one can calculate the
EMF Ėδ and the value of corresponding Ḟδ using the precalculated no-load characteristic.
Using the vector sum of Ḟδ and the referred value of stator winding MMF F˙10 , the value of
excitation MMF Ff can be found.
The reduction of stator MMF F1 to F10 is usually realized by armature reactance factor
ka :
F1
F10 = ka F1 = (1.7)
kf
Figure 1.3: Equivalent circuit for Pottier diagram
Figure 1.4: Potier diagram
where kf – excitation field form factor. It’s value for non-salient pole machines can be found
in general [22]:
Bδ1m Ff 1m 4 8sin ρπ
2
kf = = = kpf = (1.8)
Bδ Ff m π π2p
where: Ff m = If wf – field winding MMF (on the pole axis); Ff 1m = π4 kpf Ff m – the
amplitude of the field winding MMF first harmonic; kpf ≈ sin ρπ 2
/ ρπ
2
– distribution factor
for the first harmonic of MMF. This factor realizes the reduction of the distributed stator
winding to the rotor winding according to their first harmonics.
Sometimes the Pottier reactance x1σ is multiplied by a factor of 1.1 to take into
account the increase of the EMF of slot leakage due to saturation.
DQ-frame methods
The method is based on the determination of the inductances of a machine in linear mode
for the transient or steady-state operational conditions and though this method is mostly
used for salient pole machines, we should mention it here, since its “philosophy” affects other
methods of calculations.
The dq-frame is connected with the rotor axes: d -axis is aligned with the one of
excitation winding, while q-axis leads the d -axis on π/2 electrical radians. After that all the
currents, fluxes, flux densities, voltages, EMFs, MMFs both of stator and rotor are projected
to these axes and converted from a three-phase frame to dq-frame (see Figure. 1.5).
Figure 1.5: Frame conversion diagram
The coordinate transform of the currents can be performed using the Park-transform:
id = 23 (ia cos γ + ib cos(γ − 2π/3) + ic cos(γ + 2π/3))

(1.9)
iq = − 23 (ia sin γ + ib sin(γ − 2π/3) + ic sin(γ + 2π/3))
ia , ib , ic – stator phase currents;
id , iq – stator currents in dq-frame.
The reverse transform can be performed using the following equations:

ia = id cos γ − iq sin γ 
ib = id cos(γ − 2π/3) − iq sin(γ − 2π/3) (1.10)
ic = id cos(γ + 2π/3) − iq sin(γ + 2π/3)

If the stator winding of a synchronous machine is connected in star with neutral or

triangular, it is possible that ia + ib + ic 6= 0. In that case it is necessary to add an equation
for the zero sequence currents:
1
i0 = (ia + ib + ic ) . (1.11)
3
The equations (1.9) and (1.10) set up the direct and reverse correspondence between
the real currents in stator phases and new variables in the axes d, q. The equations for the
flux linkages and voltages of the stator winding are obtained in the same manner:

Ψd = 32 (Ψa cos γ + Ψb cos(γ − 2π/3) + Ψc cos(γ + 2π/3)) 
Ψq = − 32 (Ψa sin γ + Ψb sin(γ − 2π/3) + Ψc sin(γ + 2π/3)) (1.12)
1
Ψ0 = 3 (Ψa + Ψb + Ψc )

To determine the correspondence between flux linkages and currents, including the
rotor excitation winding, the following equations are used:

Ψd = Ld id + Laf d if = xd id + xaf d if  
Ψq = Lq iq = xq iq

(1.13)
Ψ0 = L0 i0 = x0 i0 

Ψf = 32 Laf d id + Lf if = xaf d id + xf if

Ψf , if – Rotor excitation winding flux linkage and current respectively.
Ld – An inductance of a stator phase with symmetric sinusoidal current for a synchronous

rotation speed, while the stator field axis is aligned with the longitudinal axis of rotor.
The corresponding reactance xd is called synchronous longitudinal reactance.
Lq – An inductance of a stator phase with symmetric sinusoidal current for a synchronous

rotation speed, while the stator field axis is aligned with the transverse axis of rotor.
The corresponding reactance xd is called synchronous transverse reactance.
L0 – An inductance of a stator phase with zero-sequence current. The corresponding

reactance x0 is called zero-sequence reactance.
Laf d – A mutual inductance between a stator phase with symmetric sinusoidal current for a
synchronous rotation speed, while the stator field axis is aligned with the longitudinal
axis of rotor, and an excitation winding; xaf d – its corresponding reactance.
The voltages can be described by the following system of equations:


ud = dΨd /dt − ωΨq + r1 id 
uq = dΨq /dt + ωΨd + r1 iq

(1.14)
uq = dΨ0 /dt + r1 i0 

uf = dΨf /dt + rf if

The equations (1.14) are the well-known Park-Gorev ones. If the saturation of a
magnetic circuit and the rotation speed are constant, then these equations become the
linear differential ones. Figure 1.6 shows the phasor diagram corresponding to synchronous
machine working as a generator on an active-inductive load.
The inductive parameters of the machines, which can be used in this model, are usually
obtained analytically based on geometrical configuration and normal operational conditions
of a machine. But this model doesn’t take into account the variation of xd and xq according
to the particular operational conditions, different loads and different load types, thus it’s
not really suitable for turbogenerators, which are known to work in wide variety of loads.
Figure 1.6: Phasor diagram of a synchronous generator working on an active-inductive load
The appropriate precision of the electromagnetic calculations is achieved using special

coefficients which are obtain from comparison of experimental and calculated data. Applied
to the special modes and electrical machines with non-typical dimensions, these methods
give too big errors of calculations.
The set of assumptions, which the existing methods are based on, cannot be fully
applied to such spread electrical machines like salient and non-salient poles synchronous
machines with non-integer number of slots per pole and per phase. The field in the air-
gap for both of them has very non-sinusoidal form and contains both high and low spacial
harmonics. This effect is significantly expressed in big turbo- and hydrogenerators of extra
power with high electromagnetic loads (Figure 1.7).
A lot of parameters and characteristics of electrical machines, which are significantly

influenced by double-sided saliency of stator and rotor, by changes of geometrical config-
uration of the magnetic field caused by the rotation and saturation, cannot be absolutely
estimated, applying the assumptions above. For instance, it concerns the harmonic factor
of the synchronous generators’ line voltage. Its value is limited by the standards of the
network but cannot be precalculated for a machine taking into account the influence of
saliency, saturation and damper winding currents in no-load mode.
Such well-known parameters as xd and xq , being used in many contemporary models of

electrical machines, are in fact nonlinear, contrary to the assumptions. Even for non-salient
poles synchronous generators xd and xq vary a lot according to the operating mode of a
machine (Figure 1.8).
Finally, a big disadvantage of the existing conventional methods is their “non-universality”,

that leads to the development of the methods for each significantly new implementation of
an electrical machine.
Figure 1.7: Flux density in the air-gap of a big generator
Figure 1.8: Variation of xd and xq parameters

1.1.2 Field calculation methods
The development of electrical machinery, growth of their power, especially that of turbo-
and hydrogenerators, highlighted the disadvantages of existing methods of electromagnetic
calculations. From the other hand, as it has been mentioned above, the evolution of the
information technologies and computing facilities allows the creation of the methods which
numerically solve Maxwell’s equations in two-dimensional and even three-dimensional rep-
resentation.
Many different techniques are realized to calculate the electromagnetic field in ma-
chines. The main of them are Finite Difference Method (FDM), Finite Element Method
(FEM), Finite Integration Technique (FIT). All of them are based on a principle of applica-
tion of a mesh on an object of calculation and further determination of the solution variable
in the nodes of the mesh. While FDM has a rectangular mesh (grid), the others represent
the geometry using the elements of different shapes starting from triangular.
These methods allow calculation of the equations of any complexity, taking into non-
linearity of the cores, account movement of rotor, eddy currents, performing resolution of
the coupled thermal and mechanical problems simultaneously. Since we always have to
solve the “direct task”, the solution involves many mutually linked iterative processes and
the bigger the complexity of the problem, the greater the time of calculation.
For general time-varying fields, the original first-order full Maxwell equations can be
written as follows [25]:
→
−
−
→ ∂B
∇× E + =0 (Faraday’s Law) (1.15a)
∂t
→
−
→ ∂D −
− → − →
∇×H − − J = J imp (Maxwell-Ampere’s Law) (1.15b)
∂t
→
−
∇ · D = ρimp (Gauss’s Law – electric) (1.15c)
→
−
∇· B =0 (Gauss’s Law – magnetic) (1.15d)
→
− →
−
D = εE (1.15e)
→
− →
−
B = µH (1.15f)
→
− →
−
J = σE (1.15g)
→
− →
−
where: E and H are the electric and magnetic field intensities, respectively; ρimp is the
→
−
imposed source of electric charge density; J imp is the imposed source of electric current
→
− →
− →
−
density; D and B are the electric and magnetic flux densities respectively; J is the electric
(eddy) current density; ε, µ, σ are the constitutive parameters: permittivity, permeability
and conductivity of the medium, respectively.
All the imposed sources in the equations (1.15) are fiven functions of the space and
time coordinates. Generally, the medium constitutive parameters can be nonlinear and
anisotropic, as well.
At the interface between two media, for example, medium (+) and medium (-), the
boundary conditions can be expressed electric fields as

−
→ →
−
n × E + − E − = 0, (1.16a)
−
→ →
−
n · D + − D − = ρimps (1.16b)
And similarly for magnetic fields as

−→+ − →− − →
n× H −H = J imp
s , (1.17a)
−→ →
−
n · B+ − B− = 0 (1.17b)
where n is the unit vector normal to the interface, pointing from medium (-) to medium
→
−
(+), ρimp
s is the imposed surface charge density and J imp
s is the imposed surface electric
current density.
The boundary conditions can be reduced to a spacial case when one of the media
becomes a perfect conductor wall:
→
−
n × E = 0, (1.18a)
→
−
n· B =0 (1.18b)
or a perfect magnetic wall:

→
−
n × H = 0, (1.19a)
→
−
n· D =0 (1.19b)
To look through the finite element solution let’s simplify the equation set (1.15), ne-
glecting the eddy currents and electric charge density. The system of equations (1.15)
becomes as follows:
→
−
−
→ ∂B
∇× E =− (1.20a)
∂t
→ −
− →
∇ × H = J imp (1.20b)
−
→
∇· B =0 (1.20c)
−
→ →
−
B = µH (1.20d)
In order to simplify the solution, a special dummy variable is introduced to the system
- a vector magnetic potential :
→
− →
−
B =∇× A (1.21)
After putting (1.17) to (1.16) we obtain:

→
− →
− →
− →
− →
−
∇ × B = ∇ × ∇ × A = ∇ ∇ · A − ∇2 A = µ J imp (1.22)
→
−
or, considering that ∇ · A = 0:
→
− →
−
∇2 A = −µ J (1.23)
For 2D sinusoidally time varying magnetic field can be described with the aid of the
magnetic vector potential in scalar form [15]:
→!
− →!
−
∂ 1 ∂A ∂ 1 ∂A →
− →
−
+ = − J + jωσ A . (1.24)
∂x µ ∂x ∂y µ ∂y
where ω is the angular frequency of the magnetic field.

Similar equation can be written for the 2D magnetostatic field:
→!
− →!
−
∂ 1 ∂A ∂ 1 ∂A →
−
+ = −J . (1.25)
∂x µ ∂x ∂y µ ∂y
The further solution of the equations can be found using one of the field calculation
methods, mentioned above, or any of their modifications. Let’s discuss a little bit one of
the most spread method in this area.
Finite Element Method

As it has been mentioned above, the in FEM we obtain the values of variables in the nodes
of a mesh, which is usually triangular, but generally can have any polygonal form. The
mesh is applied to the geometry of a real object.
Minimization of the magnetic energy functional over a set of elements (mesh) leads
→
−
to a matrix equation that has to solve for the magnetic vector potential A . It is in fact
→
−
the minimization of A throughout the mesh. A planar mesh is represented by a set of
elements: triangles or quadrilaterals, each of them having at least three vertices called
nodes or grid points. The number of nodes corresponding to each element depends on the
shape of the element and also on the type of function used to model the potential within the
element. This function is called the shape function and can be of any order depending on
the desired complexity. Usually linear or second-order shape functions are used. Assuming
→
−
a linear shape function of A within an element, the vector potential inside each element is
described as:
A = α1 + α2 x + α3 y (1.26)
If considering a triangular element with nodes l, m and n, the values of A at each

node are:
Al = α1 + α2 xl + α3 yl
Am = α1 + α2 xm + α3 ym (1.27)
An = α1 + α2 xn + α3 yn
where Al = A(xl , yl ), Am = A(xm , ym ) and An = A(xn , yn ).

In matrix form the equation set (1.26) is:

Al 1 xl yl αl

Am 1 xm ym = αm (1.28)

An 1 xn yn αn
The solution of the equation (1.27) yields:

A xl yl 1 Al y l
1 l

1
α1 = Am xm ym α2 = 1 Am y m
2∆ 2∆
An xn yn 1 An y n

1 x l Al 1 xl yl

1
α3 = Am x m Am 2∆ = 1 xm ym (1.29)
2∆
An x n An 1 xn yn
where ∆ is the surface area of a triangle with nodes l, m, n. Putting the equation (1.29)
into equation (1.26) the linear interpolation polynomial function is:
 
Al
1 X
A= [ak + bk x + ck y] Ak = [Nl Nm Nn ]  Am  (1.30)
2∆ k=l,m,n
An
where Nl = (al + bl x + cl y)/(2∆), Nm = (am + bm x + cm y)/(2∆) and Nl = (am + bm x +

cm y)/(2∆)
The flux density is constant within a particular triangular finite element:
−
→ 1 X − → →
−
B (x, y) = ( 1 x ck − 1 y bk )Ak (1.31)
2∆ k=l,m,n
→
− →
−
where 1 x and 1 y are the unit vectors.
The node point potentials Ak can be calculated by minimizing the energy functional
that is the difference between the stored energy and the input (applied) energy in the system
volume. For magnetic systems and 2D problem the minimization functional is:
Z 2
B →−
− → 1 −→2
F = − J A + jω σ A dS. (1.32)
S 2µ 2
The functional is minimized when ∂F

∂A
= 0. Considering a single triangular finite
element, we can write:
2
→−
− → 1 −
→2
Z
∂ B
− J A + jω σ A dS = 0 (1.33)
S ∂Ak 2µ 2
or
Z ( " 2 2 # )
1 ∂ ∂A ∂A ∂ →
− −→
+ − J ·A dxdy
S 2µ ∂Ak ∂x ∂y ∂Ak
Z
1 ∂ − →2
+j ω σ A dxdy = 0 (1.34)
S 2 ∂Ak
where dS = dxdy.
The minimization of the functional with respect to the magnetic vector potential can
be approximated by the following set of equations:
[G][A] = [I] (1.35)

where [G] is the global coefficient matrix, [A] is the matrix of nodal magnetic vector poten-
tials and [I] matrix represents nodal currents (forcing functions).
The elements of [G] and [I] matrices are expressed as follows:
 
bl bl + c l c l bl bm + c l c m bl bn + c l c m
1 
[S] = bm bl + c m c l bm bm + c m c m bm bn + c m c m 
4µ∆
bn bl + c n c l bn bm + c n c m bn bn + c n c m
 
2 1 1
ωσ∆ 
+j 1 2 1  (1.36)
12
1 1 2
 
1
∆
[I] = J  1  (1.37)
3
1
→
−
The equations (1.34)-(1.36) are solved for the potential A in a region containing the
triangle with nodes l, m and n. For the real problems with many elements this solution is
repeated for each one.
In order to obtain the EMF using these methods, a direct task must be solved: de-
termination of the phase EMF-s from the defined current distributions in stator and rotor
windings. That means, that the steady-state operating mode must be defined by three vari-
ables: excitation current If , stator current Is and the angle between them Ψ. The full period
must be calculated to take into account double saliency of the magnetic cores. Practically,
one has to solve the problem defined by stator voltage Us , stator current Is and angle φ
between them. That leads to the iterative process of the solution and in most cases the
steady-state mode must be determined from solving the transient one.
A FEM 2D mesh is presented on Figure 1.9, left. One can see a huge amount of
elements of the mesh, especially in the air-gap zone of the machine. While modelizing
machines with rather complex geometry, the number of elements can exceed even 100000.
It means that a lot of redundant information is calculated for each step of each iteration,
while practically one needs only windings’ flux linkages to be obtained.
Besides, the most of calculations are performed on 2D models without proper taking
into account axial heterogeinity of the cores, like ventilation ducts and end effects. Generally,
only the real lengths of stator and rotor cores are taken into account, while the influence of
that axial structure of the machine on the zones of air-gap and slots are neglected. When
these neglections significantly decrease the accuracy of the calculations, usually 3D FEM
solutions are proposed (Figure 1.9, right), thus more exceeding information is calculated.
All these facts leads to tremendous time of calculations and show the useless of these
techniques in case of general optimization of electrical machines and entire electromechanical
systems (drives). Thus these methods can find their best application in the problems of
investigation of particular phenomena in the electrical machines, like thorough end-windings’
forces calculations, edge losses and etc.
1.1.3 Permeance network methods

To decrease the waste of time while calculating each time a lot of redundant information,
a lot of methods were proposed. They can be combined together as the methods, which
Figure 1.9: A 2D mesh of a synchronous machine and a 3D mesh of a claw-pole machine
use the particularities of electrical machines, such as their peridiocity, a lot of areas of their
magnetic system with well determined flux orientation. These methods exploit the idea of
calculating the parameters of the equivalent lumped elements which represent the particular
parts of the magnetic system. This group of methods is called Permeance Network Methods
(PNM).
The main advantage of the PNM approach is that comparing with the field-based
methods one obtains much less amount of elements describing the field inside the electrical
machine with the sufficient accuracy.
The PNM approach has been in use since the works of Roters [48], Carpenter [8]
and later of Ostovic [35], Latreche [30], Slemon [53], Ivanov-Smolenskii [21], Khvostov [26],
Kuznetsov [23] and others.
In CEGELY several works are devoted to the development of dynamic flux tubes
taking into account saturation and hysteresis applied to loss estimation problem [33, 14,
52]. The works of LEEI [12] and LEF [24] use PNM approach in diagnostic domain. The
works of LEG [45, 2] present the PNM coupled with different optimization methods; some
works are oriented to design of electromechanical devices and optimization of their static
characteristics [9, 13].
Some international references also are of of the PNM approach [56, 36, 37, 11, 34, 54].
In L2EP such works have been conducted for many years using coupling of that method
with finite element analysis. Michel Hecquet applied it to a claw-pole machine [17]; C.
Delmotte-Delforge [10] realized a model of AC motor associated with its control system; H.
Roisse [47] worked on a synchronous permanent magnet machine. In order to enforce the
idea of using PNM in optimal design of electrical machines several works were devoted to
couple the approach with response surface method. From the other hand, L2EP is trying to
implement a multi-physical approach, taking into account mechanical (vibro-acoustic) and
thermal processes along with electro-magnetic one, applied to optimal design of electrical
machines. The work of M. Hecquet [18] comprises all the efforts done in L2EP in the domain
of PNM.
Flux tubes method

Method of flux tubes is one of the most popular implementations of PNM method. It was
proposed by V. Ostovic [35].
Figure 1.10: Flux tube
The magnetic circuit of an electrical machine is decomposed on several parts. Each of

the parts is represented by the permeance limited by the contours of a flux tube (Figure 1.10).
For these small regions the following conditions are applied:
• the flux φ is considered to be constant inside the tube;
• the surfaces Sa and Sb are the magnetic equipotentials.
In this case, the permeance:

φ
P = (1.38)
A − B
where A and B – the scalar magnetic potentials of the surfaces A and B respectively.
The equation (1.38) and the condition above lead to another assumption that the
magnetic fluxes in the cores and even in the air regions of the magnetic system “are frozen”,
i.e. they do not change their shape in all operating modes.
The method was realized in many different models. For example, it was realized in
the software Sirepce developed at L2EP (Herve Roisse) [47], and validated by the model of
a salient-pole synchronous generator (Michel Hecquet, [17]). A simplified equivalent circuit
(permeance network) of that generator is shown on Figure 1.11.
Nonlinear permeances Psd and Psy represent stator tooth and yoke parts, nonlinear
permeances Prp and Pry represent rotor pole and yoke parts. Stator slot leakage is rep-
resented by linear permeances Psl while rotor linear permeance Prl represents inter-pole
leakage. MMF sources Fsd and Frp are placed in stator teeth and rotor poles, they imple-
ment stator and rotor currents, realizing the coupling of the electric and magnetic parts of
the model.
The air-gap is implemented by means of permeances Pagp from each rotor pole to each
stator tooth. These permeances are obtained by FEM analysis of the machine as functions
of the rotor angular position [19, 20]. It allows to take into account the rotation, performing
transient analysis.
Figure 1.11: Permeance network of a synchronous generator
Tooth Contour Method
In order to generally take into account the influence of such factors in electrical machines
as double sided saliency of the stator and rotor cores, the change of the form of air-gap
due to rotor rotation and the saturation of the elements of the cores, the Tooth Contour
Method (TCM) was designed, theoretically proved and developed at the department of
Electromechanics of Moscow Power Engineering Institute (MPEI).
The basis of TCM was generalized in a monograph by Doctor of Technical Sciences,
Professor A. Ivanov-Smolenskii in [21] and generalized in [23]. Further development of
TCM was made in the Technical Sciences Doctoral thesis of Professor V. Kuznetsov [28].
The TCM was applied to different types of machine including turbogenerators [26].
The TCM is based on theoretically validated representation of the total magnetic field
in the air-gap as the sum of local magnetic fields of the particular elements of the so-called
tooth contours. Usually the tooth contour is considered to be an element formed by the
conductors located in adjacent slots, but it may be also represented as any part of the core
surface e.g., the pole-shoe may be divided into several “tooth contours”.
The local magnetic fields of tooth contours are to be calculated under the artificial
boundary conditions, called by the authors of the method the Special Boundary Conditions
(SBC), supposing that the permeability both of stator and rotor iron is infinitely large.
This special field is localized in a small region, which usually does not exceed three or
four tooth pitches and so it could be easily calculated. As a result, a mutual permeance
between excited contour of a stator and one of a non-excited rotor contours are obtained
as a function of their relative position. The total air gap field may be represented with
non-planar permeance network with relatively small number of branches.
The calculation of the air gap magnetic field in the absence of iron saturation is carried
out on the basis of this PN without any assumptions. We have to calculate the field only
1.2. CONCLUSION 23
inside the small area of air gap, limited by 3 − 4 tooth contours. Thus the calculation time
is extremely reduced.
More detailed description of the method and approach is given in Chapter 2.
1.2 Conclusion
So, one can easily find that the conventional methods listed in the beginning of the chap-
ter with all the assumptions applied to them, cannot now serve to satisfy the needs of
contemporary electrical machinery. The following reasons limit the use of these methods
nowadays:
• non-sinusoidal distributions of both stator and rotor MMF;
• weak taking into account the influence of double sided saliency;
• impossibility to completely take into account the variation of the machines’ parameters
as a function of magnetic saturation and mutual stator-rotor angular position;
• huge amount of “adjusting” coefficients, developed for each new kind of a machine
leading to non-universality.
The growth of the part of numerical modeling, numerical methods of direct solution of
Maxwell problem seems to significantly increase the accuracy of calculations. But tremen-
dous amount of redundant information about the magnetic field in each moment of the
simulation and in each point of the machine’s volume, while using these methods, leads to
tremendous time of calculation. It makes it almost impossible to use the methods in rapid
tasks of electric machine’s optimization. In most cases, the field calculations are used in
2D, neglecting the biggest part of information on the axial structure of the machines.
Some works are devoted to coupling of FEM calculations with conventional models
in order to obtain steady-state characteristics of a machine, as it was done in [46] for a
turbogenerator case. Parameters Xd and Xq were obtained from FEM calculations and
then regular dq-model was used. It helps to win calculation time comparing to pure FE-
models and gives a big advantage of taking into account induced currents comparing to
conventional approach.
Permeance network methods seem to show the good compromise between these two
approaches (Figure 1.12). Using the particularities of electrical machines’ magnetic fields,
these methods represent the magnetic fields with accuracy enough for the most of contem-
porary practical calculations.
The proposed approach is based on a theoretically validated tooth contour method.
The method was developed having the following aims:
• implement the complex geometrical structure of modern turbogenerators;
• take into account their particularities;
• generalize and automatize the approach of PN generation;
• create the model which can be used as a part of optimal design software.
Figure 1.12: Different calculation approaches groups
The basis of the calculation of magnetic field in an electrical machine using this method
was a little bit discussed above and will be detailed in the following chapter. Here it is
necessary to remind, that using this method, one solves the direct problem, defined by
stator and rotor currents and the spacial angle between them. So, the iterative process is
needed to calculate the particular operating mode.
For each calculated point in steady state mode, a real distribution of the windings’
FMM-s and fluxes is represented. And finally the flux linkages of the windings are obtained.
These flux linkages already include the influence of differential and slot leakages.
Besides, the influence of axial heterogeinity of the machine on the magnetic field in
different regions is taken into account. It is shown, that the radial ducts distributed axially
in the stator and/or rotor cores differently influence the following different areas of the
machine:
• stator tooth and yoke zones;
• stator slot air zone;
• air gap zone between stator and rotor;
• rotor slot air zone;

• rotor tooth and yoke zones.
The calculation of permeance network model can be rather fast and performed both
in steady-state and transient mode.
The following chapter presents thoroughly the chosen calculation approach.
Chapter 2
Used Approach
As it has been mentioned in Chapter 1, the growth and development of the computer technics
nowadays permits to numerically modelize any phenomenon in an electrical machine. For
instance, the FEM models of electromagnetic, mechanical, thermal and other phenomena
in electrical machine might contain dozens and hundreds thousand elements. Anyway, the
increase of the detailing of EM is limited since now it is not enough just to investigate some
processes in EM but what is more important – to find the solutions close to optimal ones.
Moreover, modern optimization strategy has a multiphysical orientation, i.e. it concerns
thermal, aerodynamic, vibroacoustic problems along with electromagnetic one.
In this case the electromagnetic model must be modified in order to have enough
accuracy with as less number of elements as possible. It can be achieved exploiting and
taking into account the particularities which can be found in electrical machines:
1. linear zone of air-gap;
2. high permeability of stator and rotor cores (µr ≥ 10 even if saturated);
3. well-defined direction of the flux in magnetic cores.
The proposed approach uses these particularities to construct a general permeance

network (PN) of a turbogenerator and calculate its parameters. The approach is based on
a further development of the Tooth Contour Method (TCM). The last one was designed,
theoretically proved and developed at the department of Electromechanics of Moscow Power
Engineering Institute (MPEI).
The basis of TCM was generalized in a monograph by Doctor of Technical Sciences,
Professor A. Ivanov-Smolenskii and his co-authors [23]. Further development of TCM was
made in the Technical Sciences Doctoral thesis of Professor V. Kuznetsov [28].
The idea of TCM is to divide a two-dimensional field of an electrical machine into a set
of small sub-areas where the field can be easily calculated. So-called “tooth contours” can
be chosen as such sub-areas. A “tooth contour” is a contour which embraces one tooth of a
machine’s core. Its field is being calculated under the special boundary conditions, which
diminish the size of a tooth contour down to only several tooth pitches. Since the zone of
the air-gap is linear, the sum of the magnetic fields of these contours gives a full magnetic
field in the active zone of a machine (air-gap), including both main field and the fields of
slot and differential leakage. The field of the coil end leakage must be taken into account
on the stage of the electrical calculations by means of corresponding inductances.
25
26 CHAPTER 2. USED APPROACH
The calculation of the electromagnetic processes in a machine can be generally divided

into several stages. On the first stage the calculation of the tooth contours’ fields is being
performed while exciting only one contour of all. That calculation is repeated for different
possible mutual positions of the stator and rotor contours. As a result one obtains the
dependency of the permeance between an excited tooth and one of non-excited teeth.
On the second stage a full magnetic equivalent circuit is constructed including non-
linear parts of the machine (in the cores) and the permeances of the air-gap mentioned
above. By means of this PN the branch fluxes and coils flux linkages can be calculated for
the defined distribution of the stator and rotor currents and their mutual position.
The calculation of the electromagnetic processes in the coupled electrical and magnetic
circuits is performed on the third stage.
Some works use a 3D permeance networks in order to take into account the config-
uration of magnetic field in all dimensions like it was performed for claw-pole machine in
[17]. In the case of more conventional machines (like turbogenerators, large synchronous
machines) and considering the use of the model in optimal design software, the obtained PN
is the one in-plane. Thus, in order to take into account axial heterogeneity of a machine, its
three-dimensional peculiarities and edge-effects, special techniques must be involved. Ac-
tually, all these effects have to be considered on the stage of the PN creation, i.e. during
the calculation of the parameters of the PN. By means of obtained PN both dynamic and
steady-state calculations can be performed.
This chapter presents theoretical basis of the approach, strategy of machine’s PN con-
struction and calculation of its parameters. Section 2.1 discusses magnetic charges approach
in application to PN construction and gives an idea of a “tooth contour”. The PN of an
electric machine is presented in section 2.2, showing the approach to its construction. The
particularities of taking into account axial heterogeinity of a machine are shown in sec-
tion 2.3. Section 2.4 highlights the equations of the PN and the methods used to solve
them. The method of steady-state operating mode and characteristic calculation is given in
section 2.5 while the approach of taking into account rotation is observed in section 2.6.
2.1 Theoretical basis

This section describes basic hypotheses and theoretical approach to creation of a mathe-
matical model of an electric machine using PN.
2.1.1 Calculating magnetic fields using vector or scalar magnetic

potentails
Generally, electrical machines can be decomposed on three characterized zones: ferromag-
netic cores, nonmagnetic areas filled with conductors having electrical current and nonmag-
netic areas without conductors. In the linear approximation the permeability of the cores
can be assumed infinite, thus the calculation can be performed only in the nonmagnetic
areas solving the Maxwell equations for electromagnetic field. While using a vector mag-
netic potential (VMP), the field is found solving the Poisson’s equation (1.25) with Neuman
boundary conditions. Nowadays VMP is widely used in field calculation methods (section
1.1.2) since it easily permits to take into account the areas conducting current.
2.1. THEORETICAL BASIS 27
The better convergence of the solution gives Dirichlet problem, for which a harmonic
function must be found while its boundary values are known. In this case the problem is
solved using scalar magnetic potential (SMP). Its use naturally corresponds to the ideas
of electrical potential (analogy with SMP) and current (analogy with magnetic flux) in
electrical circuits. Very well developped theoretical methods of electrical circuit calculations
are easily transposed to the magnetic area. Under several conditions a magnetic system
equivalent circuit with lumped parameters can be built, taking into account the saturation
of the cores, using SMP.
A fragment of a regular simple magnetic scheme representing the part of a stator using
SMP is given on Figure 2.1. There are three slots presented with the windings holding the
full currents ik−1 , ik and ik+1 . The yoke permeances Λyk−1 , Λyk , Λyk+1 and tooth permeances
Λtk , Λtk+1 are nonlinear. The slot leakages are represented by Λsk−1 , Λsk and Λsk+1 . The
sources of MMF Fk = ik − ik−1 and Fk+1 = ik+1 − ik are placed in tooth branches. The
air-gap permeances Λgkm ,Λgkm+1 ,Λgk+1m and Λgk+1m+1 connect stator teeth k and k +1 with
rotor teeth m and m + 1, which are not shown on the picture.
Figure 2.1: A part of magnetic system’s equivalent circuit in terms of SMP
In the circuit on Figure 2.1 the node potentials correspond to SMP and currents –
to magnetic fluxes. The MMF sources are represented by their electrical analogies – EMF
sources.
Actually, the magnetic equivalent circuit can be built using a VMP as well (Figure 2.2).
In this case we have to “rotate” the previous scheme on 90◦ . Now the flux, transversing the
branch, will be determined by potential difference between its nodes. If in case of SMP the
excitation was created by the scalar potential difference – MMF –, here its created by the
corresponding “external” currents imposed to the nodes in the area of the conductors ik−1 ,
ik and ik+1 , taking into account a scaling factor. On the external border of the yoke (blue
→
−
line on the Figure 2.2) the radial component vector B y = ∂A ∂x
= 0, i.e. Ax = const. That
leads this line to be equipotential in the scheme thus implementing Neumann boundary
condition.
Figure 2.2: A part of magnetic system’s equivalent circuit in terms of VMP
One should pay attention on the representation of the air-gap permeances. Like all the
other branches of the circuit they must be oriented transversely to the flux. But the flux in
the area of air-gap has no evident limits – the last ones depend on the mutual positions of
rotor and stator as well as on the operating mode of the entire machine. This is an important
obstacle which prevented the use of VMP in computations according to the models with
lumped parameters, like permeance networks.
In our model we use the PN based on SMP representation of magnetic field. The
magnetic charges approach permits to theoretically prove the process of PN construction.
2.1.2 Magnetic charges approach

The magnetic systems with current-carrying areas can also be handled by means of SMP
applying the magnetic charges approach even including zones inside the conductors with
current [23].
→
−
The magnetic field H can be presented as a result of overlapping of two fields
→ −
− → −→
H = H1 + H2 . (2.1)
The resultant field complies to the well-known equations (1.20) rewritten in a differ-
ential form:
→
−
→
− ∂B
rot E = − (2.2a)
∂t
→ −
− →
rot H = J (2.2b)
→
−
div B = 0 (2.2c)
→
− →
−
B = µµ0 H (2.2d)
−
→
The field intensity H2 is chosen to comply two equations:
−
→ − →
rotH2 = J (2.3a)
−
→ −
→
B2 = µµ0 H2 (2.3b)
where magnetic permeability must be the same as in the real field. The divergence of that
field is now not equal to zero since the magnetic charges were introduced:
−
→
div B2 = −ρ, (2.4)
where ρ – magnetic mass density, W b/m3 .

−
→
The curl of the component vector H1 turns out to be zero:
−→ −
→ − →
rotH1 = rot H − H2 = 0. (2.5)
−
→
Thus, the field H1 is potential and defined by SMP ϕ, which depends on the introduced
magnetic charges:
−
→
H1 = −gradϕ. (2.6)
−
→ −
→
The divergence of the field H1 along with the field H2 depends on the sources of the
field
−
→ −
→ − → −
→
div B1 = div B − B2 = −div B2 = ρ. (2.7)
Considering that H = −gradϕ and (2.7) it’s possible to obtain the equation for SMP:
div (µµ0 gradϕ) = −ρ. (2.8)

−→
The equation rotH2 = J is complied automatically if:
−
→ → −
− →l
Z
H2 = J × dl l0 , (2.9)
→
−
where dl – arbitrary directed vector.
→
−
Thus, on the first stage vector dl must be chosen and field H2 is calculated in each
point. Further, the volumetric (or surface) magnetic charge density can be found and solving
−
→ →
− →
−
equation (2.6), we can find H1 . The sought field H is equal to H in the areas without current
−
→ −
→
and equal to the sum of H1 and H2 in those carrying current.
Let’s consider a contour l carrying current I (Figure 2.3. The magnetic potential in
any point M of ambient space is created by a double magnetic layer, situated on surface
S stretched on the contour l. Moreover, the potential of the point M depends only on the
product of the current I and the space angle Ω – the angle of observation of the contour l
from the point M ,
IΩ
ϕ= (2.10)
4π
Figure 2.3: Magnetic shell and its potential.
and doesn’t depend on the form of the surface stretched on the contour.
The fact that the form of the surface stretched on the contour with current can be
chosen arbitrary is very important and gives wide possibilities and flexibility in simplifying
the boundaries and achieving better convergence of a solution.
The equation (2.10) is similar to that of electric field potential, created by a double
layer of electric charges. Generally, all the equations describing electrical charges are similar
to those of magnetic charges. The application of Gauss theorem to magnetic charges leads
to the following:
σ
HBn − HAn = , (2.11)
µ0
−→ −→
where HAn , HBn – are the projections of HA and HB onto a normal line to surface S. The
magnetic charge density σ is evident to be both positive and negative.
The equation (2.11) gives the most correct interpretation of the nature of magnetic
charges. The surfaces with distributed magnetic charges are neither equipotentail surface
nor the ones with equal potential difference between them. The imposed condition (2.11)
determines only a sudden change, a discontinuity of the magnetic field strength while cross-
ing the surface. The value of that change for the medium with permeability of µ0 concures
with the magnetization. Thus the magnetic charge density can be found like
µ0 I
σ= = µ0 Ja, (2.12)
b
where a and b – are the dimensions of the cross-section of a rectangular conductor with
current density J (Figure 2.4).
The sign of σ is determined by the sign of magnetic mass, which forms the surface
layer, i.e. it is determined by the fact wether that one magnetic mass is a “source” or a
“sink” of the field. The magnetic mass density inside the magnetic layer
σ µ0 Ja
ρ= = . (2.13)
b b
Figure 2.4: Magnetic charges field in air
The magnetic charges distributed on the surface of magnetic layer create the magnetic
field not only outside the layer but also inside. It is evident that since the property of
field line continuity is broken on the double magnetic layer surface, then the field inside
the layer is not reproduced neither by sign nor by value. For example, for a contour in the
air the magnetic charges on the surfaces of the layer stretched on that contour create field
which has an opposite sign and is much stronger inside the contour that the one outside
(Figure 2.4). The property of potentiality of the field is broken while crossing the layer. It
is not physically justified. Thus, it is necessary to consider the layer impermeable. The field
−→
in any point k can be found also by adding two fields: H1 – the field of initial magnetic
−→ → −
− → − →
charges distribution and H2 – the field of the charges, formed by a “cut” AA: H = H1 + H2 .
The surface magnetic charge density in the zone of the cut doesn’t depend on the
position of that surface (everywhere except the transverse section of the contour) and is
determined by (2.12), i.e.:
I σ
H2 = = Ja = . (2.14)
b b
In the transverse section zone the surface magnetic charge density can be considered
linearly decreasing in the line of external part of the contour if the current density is equal
in all the points of the contour.
For the chosen form of the surface stretched on the contour (Figure 2.4), the field
−→ −
→
vector H1 inside the coil has opposite sign relative to the real field, while vector H2 has the
same direction.
2.1.3 Coils in ferromagnetic systems

The assumption that the steel permeability of EM magnetic cores µF e = ∞ leads to the
assumption that tangential component vector of magnetic field on the edge between air and
steel is almost zero:
∂ϕm
Ht = − = 0, (2.15)
∂t
i.e. ϕm |t = const.
In other words, since the magnetic field lines “come” to the air-steel edge at angle of
◦
90 , the SMP of the cores on these edges can be assumed constant, i.e. these edges are the
equipotential lines of scalar magnetic potential. Actually, the equation (2.15) is true even if
the relative permeability of the core µr ≥ 10 [23].
The principle of magnetic charges (leaves, layers, shells) applied to a signle contour
can be spread on multicoil windings. The windings can be divided into the contours, each of
them being replaced by a magnetic leaf. If the leaf surface coinsides with the ferromagnetic
core’s surface, then the SMP of the core obtains a sudden change in the place of contact
equal to the magnetic charge density.
Figure 2.5: Forms of magnetic leaves
The contour embracing the tooth (Figure 2.5) is divided into 3 parts (A, B and C).
Two of them (A and C) are placed completely in air and are formed by real sections of the
contour and infinitively thin current layers carrying the full current of the contour. The part
of a contour B crosses the ferromagnetic surfaces and creates the sudden change of SMP in
the points of contacts:
ϕ2 − ϕ1 = i (2.16)
Actually, the surface stretched on the contour can have any form, not necessarily
rectangular one. It is possible create a surface, which doesn’t cross ferromagnetic core, like
shown in Figure 2.5 as dash line. In this case there is no necessity to assign the potential
on ferromagnetic cores, the last one is always zero.
As an example, let’s consider a coil with number of turns wk , current ik , which is
placed on a tooth with infinite permeability (Figure 2.6). Dividing the coil onto three parts
A, B, C as it was discribed above, we can calculate the field as a sum of fields of different
parts. The distribution diagrams of sufrace magnetic charges on the end surfaces of the coil
is shown on Figure 2.6; the maximum value of surface magnetic charge density is:
µ0 ik wk
σmax = . (2.17)
b
Figure 2.6: Distribution of SMP and magnetic charges density for a coil with current
If the density of turns of the coil is constant, then in the zone of its cross-section σ
increases linearly starting from the external border of the coil. Similarly to the magnetic
charge density changes the field strength H2 inside the coil. Its maximum value inside the
coil is:
ik wk
H2max = . (2.18)
b
In the shown version of the allocation of the surface stretched on the contour vector
−→
H2 is directed strictly vertically, i.e. its horizontal component is zero in all the points of
magnetic layer.
On a ferromagnetic surface in zones abc and hlm magnetic potential can be considered
as zero. It changes linearly along the lines cd and gh. Its maximum value takes place on the
line def g and is equal to ik wk . According to (2.15) the ferromagnetic surface is equipotential
−
→
what is complied only taking into account H2 or current layers on the ferromagnetic surface.
2.1.4 Idea of a tooth contour and special boundary conditions

To obtain the magnetic field of an electrical machine and its parameters, a method of
scalar magnetic potential can be used. To do this let’s replace the machine winding with
a magnetic leaf which is partially aligned with the surface of the core, that the winding is
placed on. Real conductors of the windings will be considered to be formed by an infinitively
big amount of thin current fibers. Further, let’s divide the leaf surface on the parts of equal
magnetic charges densities. Since the task is linear (in the air-gap) the full magnetic field
of a machine can be considered to be formed by different magnetic leafs’ fields of equal
densities. Thus, each equal density part of the magnetic leaf can be considered to be formed
by its current. So, the complex winding of an electrical machine can be divided into the
contours. The coil ends form additional contours. Now let’s consider the magnetic field in
the air-gap of a machine without coil ends, which are replaced by the boundaries of the
magnetic leafs formed by thin current fibers. The air-gap field will not change much in this
case, if the length of a machine is much bigger than the value of the air-gap. The equal
density parts of the magnetic leaf are aligned with the teeth surfaces and are limited by
two adjacent slot currents. The change of the magnetic potential while moving from the
surface of one tooth to the surface of another one is equal to the change of a magnetic charge
density, i.e. the slot current. Each equal density part of a magnetic leaf is stretched on the
contour formed by two adjacent slot currents and a current fiber replacing the coil-end part.
The current of the contour is equal to the magnetic charges density of the leaf’s part.
Let’s call such contour a Tooth Contour (TC). The currents of tooth contours certainly
must follow the condition:
i0k+1 − i0k = isk (2.19)
The currents can be determined up to a constant value. For instance, if we assume

the current of the first TC to be zero, then for the k th TC we obtain according to (2.19):
k−1
X
i0k = isj (2.20)
j=1
The magnetic potentails of teeth are the sum of the core’s potential and that of the
magnetic charges density of the corresponding TC.
Let’s consider two tooth contours: on stator and rotor (Figure 2.7). The permeability
of stator µS and rotor µR cores is infinite. The stator TC k is excited, i.e. its current i0k 6= 0.
The magnetic leaf, stretched on the tooth, produces the potential of the tooth’s surface
ϕ = i0k . All the other surfaces have zero potentials. Such conditions of potential distribution
are called Special Boundary Conditions (SBC). The field calculated under SBC permits to
obtain the flux of a non-excited TC produced by the excited one. The flux of a rotor TC m
is obtained on the surface stretched on that contour – surface abcdef .
In order to determine the permeance Λkm between stator TC k and rotor TC m one
should excite the TC k with a unit current i0k = 1 and calculate the flux Φm on the surface
abcdef . The permeance then will be
Φm Φm
Λkm = = 0 = Φm . (2.21)
ϕk − ϕm ik
Tooth Contour deep in slots

Along with the assumptions which are accepted in general theory of electrical machines and
which allow to consider the task of the calculation of a magnetic field as a quasistationary
one, let’s assume the following:
1. the field in the air-gap of the machine is an in-plane one and the coil end field can be
considered independently;
2. the tooth contours of the machine are wide and deep in slots;
3. the saturation of the cores stays in limits when the air-gap surfaces of the teeth can
be considered as equipotentials for the scalar magnetic potential (this requirement is
usually satisfied for existing electrotechnical steels since they usually have µr > 10
even in high saturation).
Figure 2.7: Stator and rotor tooth contours
The first assumption is the conventional one and the axial heterogeneity of the machine
caused by coil ends and ventilation ducts, is taken into account separately.
The idea of “wide and deep in slots” contour is given in [23]. One of its properties
making it deep in slots lies in the fact that the field’s force lines, which cross the air-gap,
are completely linked with the slot’s conductor, i.e. they do not cross the latter. According
to [23, pp. 62-63] a contour can be considered as deep in slots if the relative immersion of a
conductor into a slot h/bs is equal or bigger than the value given in table below, depending
on the relative slot opening bs /δ 1 . Most of turbogenerators do satisfy these conditions.
bs /δ 0.5 1.0 2.0 4.0 6.0 ≥8.0

h/bs 0.00 0.15 0.35 0.50 0.57 0.6
The third assumption is the most important one, since it allows applying the TCM to
the saturated machines. The influence of the saturation on the air-gap permeance has been
investigated in several works ([23], [26], [31]). In [26] it is shown that for the typical tooth
zone proportions of turbogenerators, the unipolar air-gap permeance can be considered not
depending on the saturation.
According to the second and third assumptions the air-gap field of an electric machine,
being completely linked with the contours, is formed by a system of equipotential teeth sur-
faces. At the same time, the slot’s field located deeper in the slot than a conductor, doesn’t
depend on the field of the air gap. Thus, the air-gap field can be considered independently
of the slot’s one.
1
Here: h – immersion of a conductor in slot, bs – slot opening, δ – air-gap value.
SBC and limits of the zone

The calculation of the mutual permeances between the teeth (tooth contours) of stator
and rotor is one of the most important stages of the PN parameters’ calculation, since its
accuracy is the one that the representation of a total magnetic field of an electrical machine
depends on.
To determine these permeances a magnetic field created by the current of a tooth
contour must be considered. Let’s consider a tooth contour with index s presented on
Figure 2.82 . It’s field is created by the wires with current placed in the slots between the
tooth s and adjacent teeth s − 1 and s + 1.
Figure 2.8: Tooth contour
The magnetic field of the current is is calculated applying the following assumptions:
• all other currents in the magnetic system are equal to zero;
• the current is distributed regularly inside the wires;
• the permeability of the steel of both cores (stator and rotor) is infinite (µc = ∞);
2
In order to show more spectacular slot effects an interaction of a hydro-generator pole and stator tooth
was chosen [27].
• the magnetic field is two-dimensional.
The mutual position of the stator (1) and rotor (2) cores is set by the distance bsr
between the axis of an excited tooth placed on the core 1 and the axis of a tooth placed on
the core 2.
The boundary conditions are chosen in the manner that the magnetic field lines have
only one direction in the air-gap and the consumption of the magnetic potential in the airgap
is equal to the tooth contour current is . This is reachable if the tooth s has the magnetic
potential φ = is , while all the other teeth of both cores have potential φ = 0.
The results of a field calculation of the tooth contour are presented on Figure 2.8 by a
distribution of normal part of flux density Bn on a middle line of an air gap. As one can see,
the magnetic flux density of a tooth contour in the air gap, being calculated with special
boundary conditions, fades out rather not too far away from the excited tooth contour’s
axis.
The most part of the contour’s mutual flux
Z ∞
Φs = lδ Bn dx, (2.22)
−∞
where lδ is an effective length of a machine, is concentrated in the zone of [23, p.69]
δ 2δ ∼
b1 = tz1 + 2 ln = tz1 + bs + 3δ, (2.23)
π πgtz1
where tz1 is a tooth pitch of an excited contour; δ is an air gap; g = (∆Φ0s + ∆Φ00s ) /Φs =
0.01 ÷ 0.001 is a relative tolerance of flux determination.
In order to form the limits of the zone of calculation a special, special “magnetic shunts”
can be used. These shunts are shown on Figure 2.8 as lines 10 − 20 and 100 − 200 . They have
infinitely big permeability and thus connect two magnetic cores (stator and rotor) into one
zone of a tooth contour.
2.1.5 Magnetic charges and permeance network

A permeance network is a representation of a continious magnetic system by a discrete one
having lumped parameters. As it was shown in Section 2.1.2, the magnetic charge surface
can have any form. For a permeance network based on SMP it is necessary to have well
determined boundaries with determined values on them. The best way to provide that
determinancy is to make magnetic shell to cross ferromagnetic surface.
Let’s consider a ferromagnetic tooth (Figure 2.9) surrounded by conductuctors carrying
currents ik and ik + 1 with a very simple magnetic circuit. In the magnetic circuit the tooth
k is represented by one branch, containing MMF source Fk and nonlinear permeance Λtk ;
the yoke parts under the adjacent slots are represented by permeances Λyk and Λyk+1 .
The magnetic shell stretched on the tooth will produce a change of SMP on its surface
at it was described in section 2.1.3. Since the whole tooth is represented by only one branch
that branch obtains the full change of SMP Fk = ik − ik+1 .
Generalizing, we can say that the branch of a permeance network being crossed by a
magnetic shell obtains an MMF equal to the change of SMP produced by that shell.
Figure 2.9: A magnetic shell stretched on tooth
Figure 2.10: Other configuration of magnetic shells.
A magnetic shell can be stretched also in the direction crossing the yoke part. Let’s
consider now the same magnetic system but with the configuration of magnetic shells, shown
on Figure 2.10. In this case we can introduce fake currents outside the magnetic system
which correspond to the currents in slots: i0k = −ik and i0k+1 = −ik+1 . Now the magnetic
shells stretched between these currents cross the yoke branches producing in them potential
differences Fk = ik and Fk+1 = ik+1 . This corresponds to a well-known electrotechnical
principle of transference of EMF sources over a circuit node.
The scheme on Figure 2.10 is easier to be realized since each source of MMF cor-
responds to a full slot current. Let’s introduce then another layer of PN between teeth
(Figure 2.11). The PN now obtains additional tooth permeance Λ0tk . Branch Fsk –Λsk corre-
spond to the slot leakage of the left slot and branch Fsk+1 –Λsk+1 – of right slot. The branch
k will still be crossed by the shell stretched on a full winding carrying slot current ik while
the new branch sk will be crossed by the shell stretched only on the top half of the winding
thus having the value of MMF slot Fsk = ik /2. The same effect is observed on the right
slot.
Thus, according the chosen configuration of magnetic shells (like on figures 2.10 and
2.11) one can formulate the following principles:
2.2. MACHINE’S PERMEANCE NETWORK 39
Figure 2.11: Magnetic shells in case of two layers of PN.
1. MMF sources are placed in azimuthal branches of the PN.
2. The value of the source is equal to the part of the slot current which embraces the
branch.
3. The higher the number of azimuthal branches the better the precision of the PN.
2.2 Machine’s permeance network

The space of an electrical machine can be divided into several main zones (Figure 2.12):
• stator tooth and yoke zones;
• stator slot zone;
• air gap;
• rotor tooth and yoke (central part) zones;
• rotor yoke zone (central part of the rotor).
Generally, any PN describing the magnetic system of an electrical machine consists of

the elements of following types:
• linear passive elements (permeances or reluctances);
• nonlinear passive elements (permeances or reluctances) which depend on a solution

variable (flux or ampere-turns);
• sources of MMF;
• special linear elements which depend on an independent parameter (mutual position

between stator and rotor).
Figure 2.12: Main zones of a machine
The linear elements represent the fluxes in the air-zones of stator or rotor - the slot
leakages. Their values are calculated and fixed during the calculation process.
The nonlinear elements represent the parts of the cores of stator or rotor, i.e. stator
and rotor teeth and yokes. The values of these elements are given as weber-ampere char-
acteristics Φ(Um ), where Φ – the element’s flux and Um – the element’s magnetic potential
difference corresponding to that flux.
The dependent linear elements are the ones that represent the mutual permeances
between stator and rotor teeth. They are given by characteristics Λ(Θ), where Λ – the
element’s permeance and Θ – the mutual position between stator and rotor (angle). These
characteristics are calculated on the preparation stage. The values of these elements are
fixed during the calculation process for the particular rotor position.
2.2.1 General approach for zones’ division of the machine

In order to divide the volume of an electrical machine onto discrete elements and calculate
their characteristics, we must determine the geometrical limits of the zones represented by
the elements and assume the direction of the flux in these zones. Figure 2.13 shows a part
of a magnetic system (stator or rotor) of a machine and we shall explain the process of zone
determination on that example.
Figure 2.13: Characteristic zones of machine’s magnetic core
The characteristic zones of any machine’s core are the following:

Tooth zone – the one representing a tooth of a magnetic system. The tooth is made
of magnetic material and according to the assumption that the permeability even in
saturation is relatively high, we can say that generally the flux in that zone has radial
direction only. Let’s call such a zone radial zone.
Yoke zone – the one representing a yoke of a magnetic system. The material of that zone
is also magnetic and the flux goes in the direction from one tooth to another. That
zone will be called tangential zone.
Slot zone – the one representing the air between teeth. In these zones the flux is also
directed from one tooth to another, thus it is also the tangential zone.
Zone definition
Each zone of a machine is defined with two pairs of vectors: [hl , wl ] and [hr , wr ]. These
vectors define the points of the left and right borders of the zone relative to its flux axis
(Figure 2.14). Vectors hl and hr represent the height from the top of the zone, while vectors
wl and wr represent the half-width, i.e. the distance between the point and the flux axis.
According to the specification of the machine, slot zones are defined. Tooth zones
are obtained as a geometrical differences between adjacent slots. For these radial zones the
vectors wr and wl represent actually the arc lengths of the corresponding segments.
According to the chosen number of layers, slot zones and tooth zones are divided into
elementary zones corresponding to the elements of the network. Each elementary zone is
also represented by vectors [hl , wl ], [hr , wr ].
Figure 2.14: Zone definition
This principle of zone definition is used to represent all the zones of an electrical
machine corresponding to the elements of PN, except air-gap permeances, since the latter
have no well-defined boundaries.
Automatic PN generation – Variable Topology Network

The method of general zone description of a machine discussed above gives a possibility to
implement automatic network generation according to demanded precision abstracting from
the structure of the machine.
The network generation is realized separately for three different parts: air-gap, stator
and rotor. The topology of air-gap is always fixed and determined by the number of stator
and rotor teeth. Stator and rotor magnetic systems networks are generated using the same
method and the same algorithm. The latter can be described as following:
1. Distinguishing tooth and yoke zones.
2. Tooth zone: describing the dimensions of each tooth using zone definition shown above.
Used zones are radial.
3. Tooth zone: splitting the zone of each tooth into layers one obtains a set of radial
(teeth) and tangential (slots) zones.
4. Tooth zone: forming the network of the tooth zone according to that set of zones –
each element of the network has a corresponding zone.
5. Yoke zone (central part in case of rotor): splitting the zone of yoke into sub-zones
placed between the teeth.
6. Yoke zone: forming the yoke network according to the set of zones obtained on step 5.
7. Linking tooth and yoke networks together.
The obtained networks of stator and rotor are linked together by means of air-gap
network. The resulting PN has a variable topology which depends on particular geometrical
structure of the components (rotor and stator)3 .
2.2.2 Stator permeance network

The total stator PN is presented on Figure 2.15. It consists of nonlinear radial tooth
Figure 2.15: Full stator scheme
permeances Λt 1-Λt10 , azimutal yoke permeances Λy and linear slot leakage permeances Λs2
- Λs6 . The number of elements and, thus, the precision of the network depends on the
number of layers in tooth and yoke zones. For Ny greater than 1 the layers of azimuthal
yoke permeances are added connected by additional radial yoke permeances. Tooth top slot
permeance Λstt represents inter tooth top flux.
The zoom of the tooth top on Figure 2.15 presents the permeance network of that
one part of the tooth. It has additional azimuthal nonlinear branches which represent the
saturation of that part of a tooth.
Let’s further consider the calculation of the parameters of the stator elements.
Tooth permeances
Each stator tooth is presented by several permeances radially directed. A simplified scheme
of a stator tooth is shown on Figure 2.16.
3
Topology variation is discussed in rotor network section 2.2.3.
Figure 2.16: Tooth scheme
The scheme consists of radial passive nonlinear elements connected in series. The
number of branches is determined by the number of layers which the tooth zone of stator
is divided in. For rectangular slots the teeth have a trapezoidal form. The magnetization
curve Φ(Um ) of each zone is being calculated from the BH-curve of the steel. In order to
take into account the variable cross-section of the tooth, the last one is divided into several
elementary parts (Figure 2.17). For each point of the BH-curve the flux Φ is calculated as
following:
Φ = B · lef f · bmin (2.24)
where lef f – effective core length; bmin – minimal part’s width. The magnetic potential
difference of the part is obtained from that flux as a sum of the ampere-turns of elementary
subdivisions:
H1 + 4H3 + H2
Um.el. = hel · (2.25)
6
where hel – the elementary part’s height; H1 , H2 , H3 – the magnetic strength in the top,
bottom and middle sections of the elementary part respectively. These magnetic strengths
are determined of the corresponding values of flux densities taking into account the flux
displacement into the air part: both in the slot areas between the teeth and the ventilation
duct areas between the steel packs in axial direction. The approach is described later in
Section 2.3.1. That approach allows a precise calculation of weber-ampere characteristic of
any part of a tooth even with rather complex geometrical configuration.
Stator yoke permeances

The stator yoke can be also presented by several layers of permeances. But it is sufficient to
use only one layer for that part. The magnetic scheme parts are limited in azimutal direction
by the axes of adjacent teeth. The magnetization curve of a yoke part is calculated the same
Figure 2.17: Tooth division
manner as for the tooth zone while the length of each zone is obtained by the middle line of
the flux in that one. This means that we represent the curved elements as rectangular ones
(Figure 2.18).
Figure 2.18: The stator yoke element
Slot slot leakage permeances

As it has been mentioned above the tooth zone of a stator is divided into several layers.
The slot leakage fluxes are calculated with a definite precision. The bigger the number
of the layers, the hight the precision of that calculation. But even if the winding layer is
represented by only one zone, the calculation error is less than 4% [23].
The slot leakage permeances for the zones with the winding in case of the rectangular
slot can be found as follows:
µ0 lx hi
Λσi = (2.26)
bs
where hi – zone height; lx – the effective length taking into account radial ventilation ducts
(see further); bs – the slit height.
The slit zone permeance can be found as follows:
µ0 lx hs
Λσs = + Λh (2.27)
bs
where Λh – the permeance between top surfaces of adjacent teeth in special boundary
conditions [23]:
" p #
µ0 lδ 2δ bs 1 + (2δ/bs )2
Λh = arctg + ln − ln2 (2.28)
π bs 2δ 2
where δ – air-gap, bs – slot opening.

Other formulas for slot leakage permeances can be implemented in future, like, for
instance, presented in [16, 32].
2.2.3 Rotor permeance network

As it has been mentioned in the beginning of this section the permeance networks of tooth
zones of stator and rotor are the same. So, here we shall discuss the representation of the
central part of a rotor.
General case
First, we have constructed a concentric rotor network, in which the central part layers are
formed by circular segments. Figure 2.19 shows a simplified example of that PN, outlining
the position of its elements. Each radial elements of the yoke is formed by a trapezoidal
segment like the one of radial tooth zone. Azimuthal elements have curved rectangular form
like the ones of azimuthal stator yoke zones. The methods of their calculation are the same.
Remark. For multipole machines PN is constructed only for 2 adjacent poles, since all
the processes are same. But in cases of non-integer number of tooth per pole per phase q,
the PN cannot be reduced to 2 poles: a whole machine must be calculated leading to higher
discretization and greater time of calculation.
2-pole specific case

The calculations for a 2-pole machines showed that the model presented above is not ade-
quate. Indeed, the flux in multipole machines flows usually in the segments close to tooth
zone and angular distance between poles is much less than 180◦. The 2-pole machine’s flux
have direct flow from radially opposite poles. Thus another PN of the 2-pole rotor was
proposed.
Figure 2.20 presents the PN of the central part of the rotor. Permeances Λyh1 - Λyh8
represent d-axis flux, while permeances Λyv1 - Λyv4 . Tooth zone of the machine and MMF
sources are not shown in details on the figure. D-axis directed elements correspond to trape-
zoidal zones, q-axis – rectangular zones. The methods of their characteristics calculation is
the same as presented above.
Figure 2.19: Concentric rotor PN
Figure 2.20: 2-pole rotor PN

Figure 2.21: Adaptive rotor PN.
Adaptive junction between tooth and yoke zones
In general, rotor can have different slot characteristics, like height, width, different venti-
lation ducts, etc. It leads to different PN construction for different slots. From the other
hand, all of them must be joint to the yoke PN. Thus, an adaptive variable topology PN
principle was proposed.
Figure 2.21 presents the permeance network of the tooth zone (MMF sources are
skipped). One can see that left and central slots have different heights. The general idea
is to add special transitional permeances between adjacent slots in order to pass from one
dimensioned PN to another.
Regular assumptions discussed in Section 2.1 allow to think that the border between
core and slot is an equipotential line, thus the zone of tooth is independent regarding
adjacent tooth. It gives us the possibility to apply different (modified) permeance networks
to different teeth. On the figure 2.21 “Bottom zone 1” is different of “Bottom zone 2”. The
nonlinear bottom element of the zone 1 is replaced in zone 2 by another one having different
characteristic (shown in red). Transition slot linear permeance (red) represents additional
slot flux from the right tooth to the left.
If adjacent slots have same heights (like central and right slots on Figure 2.21), no
transition layers are needed.
2.2.4 Airgap characteristics calculation

According to the principle of TCM, the air-gap permeances should be calculated between
the tooth contours of stator and rotor under SBC (section 2.1.4). The calculation of the
permeance can be realized using any available methods: FEM, FDM, FIT, MCM, analytical,
approximative, etc.
The field of an excited TC is shown on Figure 2.22. It was calculated using FEM
under SBC: the zone of calculation was limited, only one tooth was excited by means of the
current isk . The distribution of the flux density on the middle line of the air-gap is given on
the same figure. Actually, it seems to be rather easy to obtain the permeance using FEM.
In order to do this, the flux must be calculated on the line abcdef . All the calculations must
be performed for several mutual positions between rotor and stator.
Calculating the permeances using analytical approach

To calculate the air-gap permeances one have to calculate the magnetic field in the air-
gap region. The field calculations can be performed by many different methods. Here an
analytical method of calculations is given [23, pp.72-75].
Usually the calculations in the air-gap zone are performed neglecting the curvature of
the air-gap, i.e. supposing that the radiuses of the sufraces divided by air-gap are infinite.
Neglecting the curvature of the air-gap is approved for rather small ratios δ/D (D – diameter
of stator bore). But turbogenerators have rather big relative air-gap and this fact can lead
to increase of calculation errors.
In order to take into account the curvature of the air-gap a logarithmic conformal
mapping can be used. It maps a real annular air-gap δ in z-plane into a flat air-gap with
infinite curvature in t-plane:
t = K · ln z, (2.29)
where K – scale factor.

In polar coordinates of basic plane z = r · ejΘ , where r and Θ – polar radius and angle
respectively. Using t = u + jv from (2.29), one can obtain:

u = K · ln r
(2.30)
v =K ·Θ
It permits to pass from the polar coordinates of the basic plane in the area of annular
air-gap of z-plane to rectangular coordinates of flat air-gap in t-plane. The conformal
mapping is performed for the first kind boundary conditions and the mapping invariant is
the magnetic flux or the permeance. Thus it doesn’t matter which constant K is used for
the mapping.
Let’s perform the mapping for air-gap value δ = const:
r1
δ = r1 − r2 = K ln . (2.31)
r2
Then the scaling factor K:
δ
K= (2.32)
ln rr21
Figure 2.22: TC field under SBC

All the dimensions including slot optnings, tooth pitches and etc., on the surface of
the first core having radius r1 will be:
δ
x01 = x1 . (2.33)
r1 ln r1 /r2
Similarly, the dimensions of the surface of the second core having radius r2 will be:
δ
x02 = x2 . (2.34)
r2 ln r1 /r2
As soon as the dimensions of the stator and rotor slots and teeth are transformed using
conformal mapping (2.33)-(2.34), the calculation of the permeances can be performed. The
assumptions that claim the general assumptions of TCM are applied:
• the slots’ form is not taken into account, they are considered to be rectangular with
the width of slot opening equal to the width of the slits;
• the field is considered to be 2D;
• the field in the slot region is considered not to influence the neighbouring slots of the
same core (stator or rotor).
The calculation process is divided into two parts. First only the salience of core
opposite to the contour (stator) is taken into account (Figure 2.23). That core (rotor) of
the contour is assumed to be smooth; the gap has the size δ. The field in the air-gap can be
calculated using a coefficient of a magnetic permeance of the air-gap, which is built taking
into account the slots of the stator. The homopolar contour permeance of the rotor per
length unit is as follows:
Z x2
Λc = λc (x)dx (2.35)
x1
where: x1 , x2 – coordinates of the centres of the wires which form the contour.
After that the decrease of the permeance caused by the rotor salience must be taken
into account, assuming now the stator being smooth. Let’s set kγl and kγr as the decrease
coefficients of the permeance caused by the influence of the left and right contour slots
respectively. The value of kγ depends on the position of this particular slot under the
opposite tooth or the opposite slot. Thus, for each slot (left and right) the equivalent air-
gap δeq is calculated. Equivalent gap is derived based on the mean value of λc above the
part of the slot which corresponds to the contour:
Z x3 −1
1 1
δeql = = λc (x)dx (2.36)
λml x 3 − x 4 x4
Z x2 −1
1 1
δeqr = = λc (x)dx (2.37)
λmr x 2 − x 1 x1
Figure 2.23: Tooth contour unipolar permeance calculation

The value of kγ is derived from the curve of the air-gap permeance λc , taking into
account the corresponding rotor slot and the corresponding air-gap δeq . For a symmetrically
placed wire in the slot:
 s 
2
1 2 0.5b 0.5b 0.5b 
kγ = γ =  arctan − ln 1 + (2.38)
2 π δeq δeq δeq
where b – a contour slot width.

The resulting contour permeance is obtained as:
Λ = Λc − kγl − kγr . (2.39)
In order to calculate the function of the permeance an approximation function can be

used:
c + d cosh xp , |x| ≤ p

λ̃(x) = (2.40)
λmax = 1δ , |x| > p
The approximation coefficients are obtained as following:
x = 0, λ̃ = λmin = q 1 δ 2


b 1+ b

x = p, λ̃ 1 (2.41)
R p= λmax = δ 1


0
(λ max − λ̃)dx = 2
γ
Solving equations (2.40) and (2.40) together gives:

λmin cosh 1−λmax
c = cosh 1−1
λmax −λmin
d = cosh 1−1 (2.42)
γ
p = 2(λmax −c−d sinh 1)
That method of calculation of tooth contour permeances was compared with finite
element solution (using OPERA 2D). The form of the teeth is presented on Figure 2.24.
The solution was performed using TCM assumptions and SBC for µr = 100000. The
comparison of the results for different mutual positions is given on Figure. 2.25.
Figure 2.24: Finite element solution for air-gap unipolar permeance calculation
Figure 2.25: TC permeance calculation comparison

2.3. TAKING INTO ACCOUNT AXIAL HETEROGEINITY 55
2.3 Taking into account axial heterogeinity

In modern turbogenerator of high power stator and/or rotor have radial ventilation ducts
which should affect a lot on the magnetic permeance network of the model. Besides, stator
usually is made of laminated steel with its own lamination factor. All these facts show that
pure 2D representation of the fields in an electrical machine is not correct; pure cross-section
without taking into account the axial heterogeneity of the machine is impossible. To take
into account these effects in the permeance network the permeances must be modified in
order to reflect axial effects in the network [42].
Tor example, the lamination factor Ks depends on the thickness of steel sheets; for
example, for non-isolated or oxidized sheets the Ks can descend down to 0.86 (Table 2.3)
for the sheet thickness 0.15 mm.
Sheet thickness, mm 1 0.5 0.35 0.25 0.15

Packet lamination factor 0.98 0.95 0.93 0.91 0.86
For varnished sheets the factor Ks for the thickness of 0.15 can be even 0.81 (that
means the reduction of the steel’s length almost on 20%). Definitely, such a reduction of the
cross-section of steel severely affects the saturation of the core. Consequently this brings
to the increase of the reluctances of steel parts and changes the distribution of magnetic
field. In case of high saturation of several zones this circumstance influences even the field
of air-gap, thus, influencing the calculated air-gap permeances.
Let’s consider the magnetic system of a stator of turbogenerator shown on Figure 2.26.
It consists of the support, yoke part, tooth part and a winding. The ventilation ducts of
yoke part and tooth part can generally be different.
Figure 2.26: Stator magnetic system cross-section in axial direction
Several zones can be distinguished in the machine which are differently influenced by
axial heterogeneity:
1. yoke layers of stator;
2. tooth layers of stator;

3. slot areas of stator;
4. air-gap area;
5. slot areas of rotor;
6. tooth areas of rotor;
7. central core areas of rotor.
In order to take into account the axial effects in a PN generated “in-plane”, the physical
properties of the materials in these zones must be changed. For the nonlinear areas of steel
1 − 2 and 6 − 7 it should be BH-curve and it seems to be easy, since the flux is limited by the
physical borders of magnetic packs. But the zones 3 − 5 are linear and the flux there is not
limited. In the air-gap and slots the flux bulges from the magnetic system into surrounding
air (Figure 2.27). For this case the permeability of these zones should be modified: µr 6= 1.
Figure 2.27: Flux buging in the air of machine
In the PN of a machine all the elements’ parameters are calculated and presented
either as linear permeances Λk = Φk /Uk = const or as nonlinear permeances defined by
a function Uk = f (Φk ). The physical properties like BH-curves and µr are recalculated
according to the geometry of the machine. In this case it is easier to modify the actual
length of a machine. Moreover, this modification should be done depending on the layer and
the place of a permeance in a PN.
2.3.1 Steel zones

Generally, the nonlinear permeance charactersics calculation in two-dimensional represen-
tation should be performed not for a real magnetization curve B(H) but for the one with
the reduced component B in the following proportion: the axial steel length by full axial
length of the cores including the length of the ventilation ducts:
Lef f = (L − nd · wd ) · kl (2.43)
where Lef f – effective length of the core; L – real physical length of the core; nd – number
of ventilation ducts; wd – width of a duct; kl – lamination factor.
As an example two curves – basic 1 and modified 2 – are presented on Figure 2.28.
The curve was modified for the core of the length of 990 mm, which consists of 20 packets
having width of 40 mm with radial ducts of 10 mm each. For the lamination factor of 0.91
the values of flux densities of the basic curve must be reduced in 1.36 times.
Figure 2.28: BH curve modification
For the flux densities exceeding 1.8 − 2.0 T another modification should be added
to the curve 2, caused by the flux displacement into the inter-sheet non-magnetic space in
the steel packets. The gap between the sheets being determined by the thickness of the
insulation layer usually doesn’t exceed 10-15% of the thickness of the sheet itself. Thus,
the flux displacement plays a noticeable role only in the zone of high saturation. It can be
calculated using the same method as to obtain a new magnetizing characteristic of teeth
with flux displacement into slots supposing that the magnetic flux lines are parallel inside
the tooth and the slot. An approximate shape of the magnetizing curve taking into account
the flux displacement is presented by curve 3 on Figure 2.28.
Let’s consider two adjacent packs of steel and a ventilation duct between them, as it is
shown on Figure 2.29. In case of low saturation (left case on the figure), all the radial flux
passes only inside the pack, what means that the total flux density Btot passing through
the core on a duct pitch td contains only the tooth pack component Bz ; no flux is found in
the area of a duct wd .
Figure 2.29: Flux displacement to the ventilation duct area in case of high saturation
But this picture changes as soon as the saturation level raises up (right case on the
Figure 2.29. Total flux density on a pack pitch now contains also the part of the flux in the
air of the duct:
wd
Btot = Bz + Bair = Bz + µ0 Hp kl , (2.44)
td
where Hp – the field of the duct on a line AA; kl – lamination factor; wd – duct width; td –
duct pitch.
The same formula (2.44) is used to modify the BH-curve in order to taking into account
the flux displacement from teeth to slots. Combining the effects of field displacement from
teeth to slots and ducts, we can find that:
wd bs
Btot = Bz + Bslot + Bduct = Bz + µ0 Hz kl , (2.45)
td tz
where: Btot – total flux density of the element; Bz – part of flux density in the steel of tooth
and pack; Bslot – part of the flux displaced into slot area; Bduct – part of the flux displaced
into duct area; Hz – magnetic field; kl – lamination factor; bs – slot width; tz – tooth pitch.
The modification of the BH-curve is performed for each layer of the permeance network
according to the layer’s dimensions like tz , td , bs , wd . Thus, the characteristic must be
modified differently for different layers of PN.
The formula (2.44) is used to iteratively find the modified function H(B) for each
layer. The values of H(B) are used then in (2.25) in order to calculate the characteristic of
the element.
2.3.2 Air-gap zone

The field in air zones in axial direction is not limited by ferromagnetic cores and its distri-
bution depends on the duct configuration both of stator and rotor and on the dimensions
of air-gap and slots.
The flux distribution in the air-gap zone of a machine in axial direction is sketchly
shown on Figure 2.30.
Figure 2.30: Field in the airgap of a machine
Obviously, the zone of existence of the field in the air-gap is assumed to be limited by
Lδmax = max (Ls , Lr ) + 2 · δ, (2.46)
where Ls , Lr – stator and rotor physical lenghts respectively; δ – air-gap. But the flux
density Bδ in the air-gap has dips caused by the presence of ducts on stator and rotor and
the edge effects of the cores.
As it was discussed earlier in that section, in order to take into account the influence
of the ducts on the characteristics and values of the permeances in our PN, the property of
the zone of air-gap must be modified. The easiest way to do it is to modify the effective
length of the elements of air-gap:
Le f f = Kδa · Lδmax (2.47)
where Kδa – a factor taking into account axial effects; Lδmax – obtained by (2.46).
Generally, Kδa represents the ratio between the mean value of air-gap flux density in
presence of the ducts and that of smooth cores of the same length:
hBδ i
Kδa = . (2.48)
Bδmax
The value of Bδmax can be easily analytically calculated supposing that both cores are
smooth and have the length Lδmax .
In order to obtain the value of hBδ i many different methods can be used: from finite
element method and to analytical solutions. The easiest and rather effective way is to use
Carter formula for the calculation of (2.48). For different duct distribution on stator and
rotor we have:
Kδa = Kδs Kδr (2.49)
The values Kδs and Kδr :

!
δ (bdr /δ)2
Kdeltar = tdr · tdr − (2.50)
5 + bdr /δ
!−1
δ (bds /δ)2
Kdeltas = tds · tds − (2.51)
5 + bds /δ
where: tdr , tds – the step of stator and rotor steel packets respectively; bds , bdr – width of
stator and rotor packets respectively; δ – air-gap.
2.3.3 Slots zone

The principle of taking into account the influence of the ducts on slots is the same as for
air-gap. Figure 2.31 shows the field distribution in the slot zone of the machine and the
influence of the ducts.
The calculation of that influence can be performed also by Carter formulation:
!−1
δ (bd /b0s )2
Kslot = td · td − (2.52)
5 + bd /b0s
where: td – steel packet step; bd – duct width; b0s = 0.5bs – half of the slot width.
Figure 2.31: Field in the slot zone of the machine.

2.4 Permeance Network Equations

After the creation of the structure of PN according to the parameters of the turbogenerator
and calculation of the characteristics and values of PN elements, the network is ready for
magnetic state calculations. As a result of calculation we obtain the branch fluxes in the
PN setting up the values of MMF sources in the network according to the currents of stator
and rotor. Let’s consider the equations of magnetic system.
2.4.1 PN equations in matrix form

All the information about the topology of PN, the connections between its elements is
presented by the incidence matrix having the size T × P , where T – number of lines in
matrix corresponding to the number of nodes, P – number of columns corresponding to the
number of branches in the PN:
 
a1,1 ... a1,P
A=  ... ai,k ...  (2.53)
aT,1 ... aT,P
where ai,k = 1, if branch k belongs to node i and has direction from node i; ai,k = −1, if
branch k belongs to node i and has direction to node i; ai,k = 0 if branch k doesn’t belong
to node i.
The information about the permeances is stored in the diagonal matrix having size
P × P:
 
Λ1 ... ...
Λ=  ... Λk ...  (2.54)
... ... ΛP
where Λk – a permeance of branch k of PN. The values of this matrix are calculated according
to the methods given in Section 2.2 and contains the following permeances:
1. linear constant permeances with Λk = const, for example, representing slot leakages;
2. linear permeances of air-gap which depend on the angle of between rotor and stator
Λk = f (α);
3. nonlinear permeances, which generally are defined by Λk = ΦUkk = f (Φk ) = g(Uk )4 ,

where Φk and Uk are the flux and the potential drop of branch k.
The MMF fk in branches of PN, produced by currents ij in the branches of electric

circuit, are defined as a vector:
f = |f1 , f2 , ..., fk , ...fP |t , (2.55)
where:
4
The method of taking into account nonlinear permeances will be described later in this section.
2.4. PERMEANCE NETWORK EQUATIONS 63
PQ
fk = j=1 fk,j – MMF of branch k of PN;
fk,j = Wk,j · ij – MMF of branch k of PN produced by current ij in branch j of electric
circuit;
Wk,j – number of turns (bars) with current ij , embracing branch k of PN;
Q – number of branches in electric circuit.
The values Wk,j have zero values for the most part of the branches of PN except those
of having MMF sources, i.e. tangential branches in the areas of slots and yokes.
The values of Wk,j are calculated according to the method described in Section 2.1.5.
For the PN branches having MMF sources, Wk,j is the number of wires with current ij ,
placed in the zone between branch k and the air-gap. Actually, this corresponds to the
number of elementary fibers whose magnetic shells, being stretched in the yoke direction,
cross the particular branch k of the PN. The sign of the value Wk,j corresponds to that of
the current ij .
According to (2.55) the MMF vector of the PN can be presented as follows:
f =W·i (2.56)
where i – vector of currents in electric circuit.

The matrix W is called the transformation matrix of electric currents of electric circuit
into MMF sources of magnetic circuit:
 
W1,1 ... W1,Q
W =  ... Wi,k ...  (2.57)
WP,1 ... WP,Q
The matrix (2.57) has the number of columns equal to the number of electric circuit
branches Q and number of lines equal to the number of PN branches P .
In order to obtain the vector of fluxes in the branches of PN
Φ = kΦ1 , Φ2 , ..., Φk , ...ΦP kt , (2.58)
which are produced by MMF sources f (2.56), a node-potential method can be used. The
vector of magnetic potential drops
U = kU1 , U2 , ..., Uk , ..., UP kt (2.59)
can be expressed in SMP
ϕ = kϕ1 , ϕ2 , ..., ϕk , ..., ϕT kt . (2.60)
by means of incidence matrix (2.53):
U = At · ϕ. (2.61)
The potential drops can be expressed for each permeance in the network:
U Λ = kUΛ1 , UΛ2 , ..., UΛk , ..., UΛP kt = U + f . (2.62)

The vector of branch fluxes can be expressed from the vector of potential drops as
follows:
Φ = Λ · U Λ, (2.63)
where Λ – matrix of permeances (2.54). Taking into account also (2.61) the second Kirch-
hoff’s equation can be written for a PN:
Φ = Λ · At · ϕ + f .

(2.64)
In order to unambiguously determine the unknown fluxes Φ and node potentials ϕ for
given branches MMF-s f , the first Kirchhoff’s equation must be added in the matrix form:
A · Φ = 0. (2.65)
The equations (2.64) and (2.65) completed with (2.56) describe the magnetic state
of a PN for given currents in electric circuit. The solution of these nonlinear equations is
realized using the methods described in sections 2.4.2 and 2.4.3.
The flux linkages of stator phases can be expressed through the branch fluxes found
from the solution of the equations.
The flux linkages of branch j of electric circuit can be presented as a sum of flux
linkages of all the magnetic branches with that electric branch:
X
Ψj = Ψjk Pk=1 , (2.66)
where Ψjk = Wjk · Φk – the flux linkage of the flux of branch k of magnetic circuit with the
wires of branch j of electric circuit; Wjk = W kj – number of wires of branch j which are
linked with the flux Φk , i.e. that of magnetic branch k.
The matrix W has the same structure as matrix (2.57). The vector of flux linkages of
electric branches can be expressed:
p
Ψ = kΨ1 , Ψ2 , ..., Ψj , ...ΨQ kt = Wt · Φ, (2.67)
a
where p – number of pole paires; a – number of parallel branches of stator winding.
2.4.2 Tableau formulation

Let’s discuss the general method to solve circuit equations, both electrical and magnetic.
The equations will be considered in electrical notation. And in order to pass to the descrip-
tion of magnetic circuit, simple and well-known analogies can be used: electric current –
magnetic flux, electric potential – SMP, voltage – magnetic potential difference.
We will use the tableau formulation [58] to describe the circuit equations. In this for-
mulation, all equations describing the network are collected into one large matrix equation,
involving the Kirchhoff voltage law (KVL), Kirchhoff current law (KCL) and the constitutive
equations, CE.
For initial consideration let the network have b branches; n + 1 nodes; R, G, L, C
elements; and sources. The KCL equation is expressed by incidence matrix means A
A · I b = 0, (2.68)
2.4. PERMEANCE NETWORK EQUATIONS 65
while KVL is given by
U b − At V n = 0; (2.69)
where: I b – branch currents; U b – branch voltages; V n – node potentials. The subscript n

stands for nodes, the subscript b for branches, and in applications b will be replaced by the
element number.
The general representation describing all possible constitutive equations has the fol-
lowing form:

currents Y1 K1 S b1
Ub + Ib = (2.70)
voltages K2 Z2 S b2
where: Y1 and Z2 represent admittances and impedances, respectively; K1 and K2 contain

dimensionless constants; and S b1 , and S b2 include the independent current and voltage
sources, as well as the influence of initial conditions on capacitors and inductors. For
notational compactness, we will use the following form:
Yb U b + Zb I b = S b . (2.71)
In all subsequent formulations, capacitors will be entered in addmittance form and

inductors in impedance form, to keep the variable s in the numerator. Since the Laplace
transform variable s is equivalent to the differentiation operator, we obtain a set of algebraic-
differential equations when performing time domain analysis. Table (2.1) indicates the
choices of Yb , Zb and Sb for various twoterminal elements including nonlinear elements.5
Element Consitutive Equation Value of Yb Value of Zb Value of Wb

Resistor Ub − Rb Ib = 0 1 −Rb 0
Conductor Gb Ub − Ib = 0 Gb −1 0
Capacitor sCb Ub − Ib = Cb V0 sCb −1 Cb V0
Inductor Ub − sLb Ib = −Lb I0 1 −sLb −Lb I0
Voltage source Ub = Eb 1 0 Eb
Current source Ib = Jb 0 1 Jb
Nonlinear element I(U ) Ib = I(Ub ) 0 1 I(Ub )
Nonlinear element U (I) Ub = U (Ib ) 1 0 U (Ib )
Table 2.1: Choices of matrices’ elements
Equations (2.68)-(2.71) can be collected in the following sequence:
U b − At V n = 0 (2.72a)
Yb U b + Zb I b = S b (2.72b)
AI b = 0 (2.72c)
5
The method of taking into account nonlinearities will be described below.
and put into the matrix equation
b b n
     
b 1 0 At Ub 0
b Yb
 Zb 0   Ib  =  Sb  (2.73)
n 0 A 0 Vn 0
The matrix in (2.73) has the size of (b + b + n) × (b + b + n). Its size can be reduced
by substituting (2.72a) into (2.72b):
Yb At V n + Zb I b = S b (2.74a)
AI b = 0 (2.74b)
and in matrix form:

Yb At Zb Vn Sb
= (2.75)
0 A Ib 0
The size of this matrix is now (b + n) × (b + n). The matrix (2.75) can be written in a
simplified form:
Tx = S (2.76)
where: T – network matrix; x – network variable vector; S – network sources vector.

The matrix form (2.75) can be used to calculate both magnetic and electric circuits.
The matrices Yb and Zb are actually diagonal matrices and represent the permeances and
reluctances in magnetic network respectively.
2.4.3 Taking into account nonlinearities

To solve nonlinear equations a well-known Newton-Raphson method can be used. It gives
rather good quadratic convergence and is very attractive to be used in systems with high
nonlinearity. The solution of the system of equations in the following form can be found
iteratively:
F(x) = 0 (2.77)
While solving this equation the iterations of Newton-Raphson are defined as follows:
xk+1 = xk + ∆xk (2.78a)

J∆xk = −F(xk ) (2.78b)
where: F(xk ) – the set of equations (2.77) calculated for x = xk ; J – Jacobi matrix, for
∂Fi
which Ji,j = ∂xj
.
Thus, the solution of (2.78b) linearized for the step k and obtain ∆xk . Then we obtain
the new values of variables vector xk+1 using equation
(2.78a). The convergence criterion
k+1 k
is the reduction of the norm of errors: F(x ) ≤ F(x ) . That method converges in

approximately 3-5 iterations.
2.5. STEADY-STATE FIXED-ROTOR MODEL 67
In order to solve nonlinear circuit equations (2.75) we must transform them to fit the
form (2.77):
U b − At V n
 
F(x) =  Yb U b + Zb I b − S b (I b , U b )  = 0 (2.79)
AI b
Nonlinear elements are included in S b (I b , U b ) component of equation system (2.79) as

it was shown in Table 2.1. The Jacobi-matrix for the system of equations (2.79) generally
has the following form:
Y0 b (I, U )At Z0 b (I, U )

J= (2.80)
0 A
The components of matrices Y0 b and Z0 b differ according to the type of the element in
the particular branch. For linear elements:
0
Yi,i (Ii , Ui ) = Yi
0 (2.81)
Zi,i (Ii , Ui ) = Zi
For nonlinear elements the components of matrices will be:

dIi
Yi,i0 (Ii , Ui ) =

dUi
0 dUi (2.82)
Zi,i (Ii , Ui ) = dIi
The structure of Jacobi-matrix (2.80) is the same as (2.75), thus it is very easy to
obtain that matrix as soon as matrix (2.75) is defined.
2.5 Steady-state fixed-rotor model

To calculate the steady-state mode of the machine a mathematical model with a fixed rotor
can be used considering the following assumptions:
• the stator windings voltages and currents are considered to be sinusoidal;
• the calculation of the magnetic circuit is being produced for an arbitrary rotor position,
considering that the calculation results for the first harmonics don’t depend on that
position.
2.5.1 Basic equations

The steady-state symmetric mode of the electrical machine is defined by an effective value
of the stator current Irms , excitation current If and an angle β which defines the position
of the stator current vector on phasor
√ diagram in d − q-axes (Figure 2.32). As a result we
obtain phase voltage vector U = 2Urms and angle ϕ between stator voltage and current.
Figure 2.32: Synchronous machine vector diagram
The MMF sources in PN branches will be defined by stator phase currents and rotor
current:
i1 = I · cos β + π2 + α

i2 = I · cos β + π2 + α − 2π

3 (2.83)
i3 = I · cos β + π2 + α + 2π 3
if = If
√
where: I = 2Irms – an amplitude of the stator phase current.
After magnetic circuit calculation all the branch fluxes are obtained from which one
can calculate phase flux linkages using (2.84). Actually, these flux linkages contain both the
fundamental and the high-order harmonics. In a real machine with real windings the most
of harmonics have low amplitudes, except the 3rd one. To exclude it one can convert phase
flux linkages into linear ones. The system of phase flux linkages can be easily transformed
into d − q frame:
Ψd = 23 (Ψa cos γ + Ψb cos(γ − 2π/3) + Ψc cos(γ + 2π/3))

(2.84)
Ψq = − 32 (Ψa sin γ + Ψb sin(γ − 2π/3) + Ψc sin(γ + 2π/3))
The following equations result to the EMF:

q
Ψδσ = Ψ2d + Ψ2q (2.85)
2.5. STEADY-STATE FIXED-ROTOR MODEL 69
Ψd
γ = arctan (2.86)
Ψq
The flux linkage found by (2.85) takes into account not only the mutual inductance
field but also slot leakage field and differential leakage field produced by higher harmonics
of stator field without taking into account damping effect.
The vector of flux linkage can be presented in a complex plane, which real axis is
aligned with d-axis and imaginary one – with q-axis:
Ψ̇δσ = Ψδσ ejγ . (2.87)
There is a vector of phase EMF corresponding to the vector of Ψδσ . This vector is
produced by the same magnetic fields:
Ėδσ = −jω Ψ̇δσ , (2.88)
where ω = 2πf – angular frequency of stator current.
Now the stator phase voltage U can be calculated. For symmetrical steady-state
operating mode and assuming that all the values change sinusoidally, the voltage U can be
obtained from the voltage equation written in complex form:
U̇ + RI˙ + jxσe I˙ − Ėδσ = 0. (2.89)
where: U̇ , Ėδσ , I˙ – complex amplitudes of stator voltage, EMF and current, respectively; R
– active resistance of stator phase; xσe = ωLσe – end-winding reactance.
Many different approaches are known to determine the value of xσe . Comparing dif-
ferent formulations, we have chosen the one giving best results for turbogenerators [7]:
2
2 Zn · q · p 0.7Ds
xσe = 2πf µ0 20 + ξ (2.90)
p a p
where: f – electric frequency, Hz; µ0 – air permeability; p – pole pairs; Zn – conductors per
slot; q – slots per phase per pole; a – number of parallel circuits; Ds – stator inner diameter;
ξ – winding factor.
By means of (2.89) a complex vector of phase voltage U̇ can be determined along with
its RMS value
√
Urms = |U̇ |/ 2 (2.91)
and power factor angle between U and I
π π
ϕ = arg U̇ − arg I˙ = − − Θ − − − β = β − Θ. (2.92)
2 2
2.5.2 Characteristics calculation

All the steady-state characteristics of the turbogenerator can be calculated using the model
with fixed rotor presented above. As it was shown above the direct solution can be obtained
as function of three defined variables: excitation current If , stator current Is and electrical
angle between them β. Thus, only no-load characteristic of turbogenerator can be obtained
using direct solution. Any other operating mode is usually defined by other variables, like,
for instance, stator voltage Us , stator current Is , power factor (PF) or angle ϕ. In this case
the operating mode and characteristics can be obtained using iterative process.
Calculating operating mode defined by stator voltage, stator current and power
factor
√
The calculation
√ of operating mode defined by stator voltage U 0 = 2Urms , stator current
I 0 = 2Irms and angle between them ϕ0 , is usually performed in order to obtain the
excitation current If and stator current angel β.
The calculation algorithm is as following:
1. set up initial values of If , β and I;
2. calculate complex value of stator voltage U̇ using (2.89) and |U | = fu (If , β) using
(2.91); angle ϕ = fϕ (If , β) is obtained from (2.92);
3. comparing obtained values with the defined U 0 and ϕ0 , the following discrepancies
can be found:

∆U = U 0 − fu (If , β);
(2.93)
∆ϕ = ϕ0 − fϕ (If , β);
4. finally, the system of nonlinear equations must be solved:

∆U = U 0 − fu (If , β) = 0;
(2.94)
∆ϕ = ϕ0 − fϕ (If , β) = 0.
The system (2.94) can be solved using any of the numerical methods, for instance, the
secant method. According to the method, the following coeffients of linear correspondence
must be determined on iteration j:

∆U ∆U
K11j = ∆Ifjj ; K12j = ∆βjj ; 

β=const If =const (2.95)
∆ϕ ∆ϕ
K21j = ∆Ifjj ; K22j = ∆βjj ; 

β=const If =const
Then the increments ∆If j and ∆βj which make vanish the discrepancies are found
solving the following system:
K11j ∆If j + K12j ∆βj = ∆Uj
(2.96)
K21j ∆If j + K22j ∆βj = ∆ϕj
The new values of variables are obtained:
If j = If,j−1 + ∆If j ; βj = βj−1 + ∆βj . (2.97)
The convergence criterion can be taken as the relative RMS value:
s
∆If j 2 ∆βj 2
+ ≤ , (2.98)
If j βj
where – relative error of the solution.
Described algorithm of operating mode calculation includes two iterative processes:
inner process of magnetic PN calculation and outer process of excitation current calculation.
It gives very good convergence of up to 4 iterations. The algorithm is used to calculate rated
mode excitation current along with the regulation and V-shaped characteristics.
The same algorithms are used to obtain other characteristics of turbogenerator which
depend on other variables. The examples of calculated characteristics are shown in Chap-
ter 4.
2.6. MODEL TAKING INTO ACCOUNT ROTATION 71
Calculating xd and xq parameters

Having built a vector diagram for a particular operating mode, we can easily calculate
machine’s xd and xq parameters, using following relationship:
E0 − Uq − Rs · Iq
xd =
Id
(2.99)
Ud + Rs · Id
xq =
Iq
where: Ud , Uq – stator voltage projections on d- and q-axis respectively; Id , Iq – stator

current projections on d- and q-axis respectively; Rs – stator active resistance; E0 – rectified
no-load stator EMF corresponding to selected excitation current If .
2.6 Model taking into account rotation

The model with fixed-rotor discussed in Section 2.5 having many advantages still has its
limits of use. Some applications strongly demand taking into account the real rotation of
the machine’s rotor. Having in mind that this thesis is devoted to big turbogenerators and
the calculation of their steady-state characteristics, we should say that implementation of
the rotation, especially with transient analysis, is out of scope of that study.
However, the used approach gives flexible possibility to take into account the mutual
movement of the stator and rotor cores. The characteristics of air-gap elements which are
precalculated on the network construction stage and determined for the whole range of
possible cores’ mutual positions.
As an example, Figure 2.33 presents mutual permeances between one stator tooth
and all rotor teeth of a turbogenerator. The general approach of taking into account the
rotation is to implement a variable which determines the mutual position between stator
and rotor. That variable is then used to calculate the particular values of the permeances
for the current mutual position. The algorithm was realized in the software (see Chapter 3)
and some results can be found in Chapter 4.
2.7 Conclusion
This chapter presented the used approach, its theoretical basis and methodical implementa-
tion. The first section leads step by step to the choice of the PN configuration, the approach
of its parameters’ calculation. General theoretical ideas are presented.
Te second section described the particularities of the PN construction and calculation
of the parameters of its parts: nonlinear and linear tooth elements, air-gap elements, yoke
elements. An idea of variable topology model was proposed in order to take into account
geometrical peculiarities of some machines.
The method of taking into account the axial heterogeinity of magnetic cores was dis-
cussed in the third section. We showed that special techniques must be implemented to
properly reflect axial effects while using a model “in-plane”. Different parts of the machine
are differently influenced by axial heterogeinity, characteristic zones were detected and the
implementation on the level of PN was given. However, the proposed method is rather
Figure 2.33: Mutual chracteristics between stator and rotor teeth.
general and can be easily used with other methods of machine’s characteristic calculations,
even FEM (in plane).
Steady-state fixed rotor algorithm was presented in the fourth section. The method
has very high speed of calculation since all the characteristics can be obtained without
“moving” the rotor from its primary position.
Finally, we said a couple of words about the model which takes into account rotation.
Chapter 3
Software implementation
3.1 General Information

The software TurboTCM is designed for calculation of different operating modes of turbo-
generators 10 − 100 M W with air-cooling. All the calculations are performed based on
modified Tooth Contour Method, replacing the magnetic circuit of an electrical machine
with rather precise permeance network, taking into account the particularities of magnetic
field distribution in a machine.
The calculation is performed in two stages. On the first stage – the generation – the
geometry of a machine is analyzed based on its parameters and the generation of a PN with
calculation of its parameters is performed. As a result, a PN-model of electrical machine is
created.
On the second stage – calculation – the generated PN-model is used for the calculation
of the static operating modes or movement analysis:
• modes, defined by stator current, rotor current and power factor;
• modes, defined by stator current, stator voltage and power factor;
• machine’s characteristics: no-load, load, regulation, V-shaped, angular, etc...;
• for each point of characteristic and for each calculated operating mode the complete in-
formation about magnetic and electric state of a machine can be obtained: spacial and
time distribution of flux density in the air-gap, teeth, yokes, visual field distribution
map in the cross-section of the machine, xd , xq parameters, etc...;
• evolution of stator and rotor fluxes, flux linkages and EMFs in time (in rotation).
TurboTCM is created in Matlab and can be compiled in two ways:
1. the Matlab program;
2. separate software for Microsoft Windows with Matlab Component Runtime(MCR)

module.
The software was designed using modular principle: calculation core – graphical user
interface (GUI). The GUI allows to set up the input project data, generate PN-model,
73
74 CHAPTER 3. SOFTWARE IMPLEMENTATION
verify the calculated parameters and characteristics of the model, perform the calculation
of different modes and charactersitics of turbogenerator.
Direct use of the calculation core gives a user more flexible capabilities1 . The function
of the core can be called directly from Matlab command line or from the user-defined scripts
and programs. The functions give the following capabilities:
• create the project or read and modify the existing one, created by GUI;
• generate the PN-model based on the project;
• calculate any characteristics and modes listed above or user-defined;
• access the complete information about PN generated, all its elements, characteristics,
etc.
• based on the full information about each element of the PN any additional calculations
can be performed in order to meet the particular demands of the user.
Easy direct access to the model permits its use as a part of turbogenerator optimiza-
tion software. The example of combined use of TurboTCM and Sophemis2 is shown in the
Section 4.4.
3.2 Software structure and approach

The software implementation of the method described earlier consists of several tasks. As
an input data we have a full specification of our turbogenerator: stator and rotor sizes,
number of stator and rotor slots, their dimensions, insulation, full description of the stator
and excitation windings, damper winding, characteristics of the materials, specification of
the electrical part, etc. Based on that information we have to:
• form the equivalent circuit of the magnetic system;
• form the circuit of the electric system;
• form the equations describing magnetic and electric circuits and mechanical system.
As soon as the system of equations is obtained, it can be solved to calculate a set

of characteristics in different operating mode. The approximate structure of the software
system is presented on Figure 3.1.
The design of the software was divided in two layers:
• Circuit Description Tools, developed in C++, which manage a network (electric or

magnetic or any other) and produce its equations as well as relations between electric
and magnetic parts;
1
The direct use of the core is not available in the separately compiled version of the software and can be
realized only from inside Matlab
2
Sophemis is an optimisation software package developped by Stephan VIVIER in L2EP of Ecole Centrale
de Lille
3.2. SOFTWARE STRUCTURE AND APPROACH 75
Figure 3.1: General structure of TurboTCM

• Mathematical Solver and Graphical User Interface, developed in Matlab, which inter-
acts with the user and solves all the equations to obtain the results.
Further, circuit description tools will be presented followed by the general approach
to construct the solver and main program application.
3.2.1 Circuit Description Tools overview

The circuit description tools were designed to obtain the matrices of the systems of equations
to be solved. These tools have been designed in Visual C++ .NET [55, 5] using Component
Object Model (COM) technology [50, 6]. COM is based on interface and object oriented
programming. In a couple of words, this technology allows to combine together different
software tools (which follow COM standard) in order to obtain the best of their features.
Using this technique one can write the software in any preferable for the particular task
language and then use it as “bricks” (COM object) from another software written on any
other language. The only demand – COM support in the compiler. As it was previously
said, in this work the part of the software which was dedicated to building of circuit matrices
was written in C++ language using this COM technique. This part implemented the graph
approach of building circuit equation matrices and special indexing technique to access as
fast as possible any of the desired elements of the circuit and its parameters in any desired
time during the simulation.
All these tools were compiled as a COM type library TCMLib. It provides all the
components which generate and manage in real time the circuit equation matrices. Those
components can be used from any solver written, for instance, in Matlab or Visual Basic
or any other programming language.
Two main objects of the library – CircuitBuilder and CircuitConnector – provide
the solver with the possibility to add and manage the elements and characteristics of the
networks as well as the interaction between networks (Figure 3.2).
As soon as we define a network (electric or magnetic) by means of CircuitBuilder, it
analyses the circuit and provides all the necessary matrices for the equations to the Solver.
It is done both for electric and magnetic circuits. Solver, in its turn, solves the systems of
equations. During the solution the Solver affects the CircuitBuilder object to maintain
the current state of the circuit, depending on parameters specified (as, for instance, rotor
angle or elements’ characteristics, etc...).
Two CircuitBuilder objects are linked to each other by means of CircuitConnector
which provides the solver with correspondence matrix.
The interaction between the components of TCMLib and Matlab is shown on Figure 3.2.
CircuitBuilder
This component provides the functionality to build and manage any circuit (network) based
on graph theory. This circuit can represent either electric or magnetic or thermal one – this
is just a question of analogies.
CircuitBuilder is the main component of the library. It has the following features:
• Adding and maintaining the elements of the following element types to the circuit
description:
Figure 3.2: Interaction between TCMLib and Matlab
linear permeances, resistances, capacitors and inductances (last two for electrical
circuits);
current and voltage sources;
all the elements above with the dependency of a particular parameter;
nonlinear elements specified by characteristics of I(U) or U(I).
• Adding and maintaining the parameters of the circuit (for example, rotor angle), which
the dependent elements’ values are based on. In linear case the parameter’s value is
just applied to the element’s value. In nonlinear case, the parameter’s value is the
argument for I(U ) (or U (I)) characteristic, associated with the element.
• Adding and maintaining the characteristics for the elements. One characteristic can
be associated with many elements. This is necessary to maintain the group of equal
nonlinear elements.
• Generation and providing circuit matrices, including Jacobi matrix components.
• “Tuning-up” features which allows the user to control the names of the types of ele-
ments.
CircuitConnector
This component was designed in order to manage the connections between two circuits
described by CircuitBuilder object. It provides the following features:
• Two connections with two CircuitBuilder objects, which represent, for instance, the
electrical and magnetic circuits.
• Adding and maintaining the relations between the first and second circuits’ branches.
• Providing the information to the solver about the correspondence matrix.
3.2.2 Matlab Solver general approach

This section presents the structure of the entire software realized in Matlab v.7.0. It con-
tains the description of the software creation approach without any details of its particular
realization.
TurboTCM being a rather complicated software was designed using an Object Oriented
Programming (OOP) approach. But we decided to void using native class and object imple-
mentation realized in Matlab, since its realization is rather “heavy” and not so transparent.
Thus a so-called “lite OOP” approach was used.
Most of the structures have so-called “TurboTCM object form”, which means that they
behave like real objects with the help of associated member-functions. Anyway, there is a
limit of this behaviour since there is no variable scope checking and inheritance mechanism
implemented.
To realize OOP behavior most of the structures (with few exceptions) have special
obliged fields:
.Name – defines the name of the structure (analygoy in OOP: class name).
.Version – defines the version of the structure as a string in the format: TurboTCM.1
(1 – version number).
The .Name field allows checking the type of the structure in the functions’ calls, while
by means of the .Version field a backward compatibility mechanism is realized.
The functions linked to the specified structures and performing operations on its mem-
bers are called “member functions” (like in OOP). The member functions have special name
decoration: StructureName_FunctionName, where: StructureName – the name of an asso-
ciated structure, FunctionName – the name of a member function.
Moreover, each member function must have a maintained structure as the first input
and first output arguments, declared with the same name. The type and version (for back-
ward compatibility) of the structure given as the first argument is usually checked using the
function CheckTypeVersion.
For each structure there is a constructor member function which has only output ar-
gument of the structure generated. This function is usually called StructureName_Create.
It creates all the fields used in the structure with default values (usually []) and assigns
the .Name field to a structure name and .Version field to the version number.
Sometimes, the constructor supports the second output argument which has the same
fields as the structure, except that all of them are the strings with the fields’ descriptions.
Note. Usually, the constructor is the function which provides all the information of its
fields to the user in its help. So, if one wants to know the fields and their meaning for a
structure, he should just type in the Matlab command line:
help StructureName_Create
Also, a structure may contain a destructor – a member function with the following
name:
StructureName_Delete.
It is not required but may be usefull when the sturcture contains ActiveX and/or Java
objects as members.
The following section presents global view of the basic structures and their interaction.
3.2.3 Matlab Solver structure

TurboTCM is divided into several logical parts which have their own particular purposes:
Structure – definition of all the structures used in TurboTCM. It aslo contains the most
part of their member functions.
Generation – generation of the Permeance Network Model (PNM) based on the input
parameters.
Core – calculation of the output modes and characteristics based on PNM generated.
Services – service functions which are widely used in all other parts of the software.
According to Matlab habits, the software is placed in a special folder – its root folder.
The parts described above correspond to its subfolders. The GUI part of TurboTCM is placed
in root folder.
Figure 3.3 shows the global view of the collaboration and interaction between different
parts of TurboTCM. The following subsections will describe that interaction.
Figure 3.3: Collaboration between different parts of the software

Input data specification

The input data of a turbogenerator are stored in the TCM object which is called TCMProject.
The project of TurboTCM contains the parameters of the turbogenerator along with the
simulation parameters. Besides, it stores the generated PNM and possible results. The
structure of the input parameters of the project is shown on figure 3.4.
Figure 3.4: Input data specification
Input data contain Nominal Data of the machine, the parameters of its Stator and Ro-
tor, Simulation parameters. The Symbols substructure contains special internal information
concerning the PNM generation, like elements’ and characteristics’ symbol masks.
The description of stator and rotor of a machine is realized by means of the TCM object
MagneticSystemData. It generalizes the description of any magnetic system of rotating
machine. It describes the dimensions and distribution of slots, general dimensions of the
magnetic part, winding parameters and distribution along with the material specifications.
Model Data structure

According to the Input Data Specification which defines geometrical and other parameters
of the turbogenerator, the main TCM object of calculation is generated – ModelData. That
object reflects the current state of the PNM and is used to perform all the possible calcu-
lations of the model. The member functions of that object start with the name “MD_” or
“ModelData_”.
The ModelData structure contains the following main information about the model:
.cbm – CircuitBuilder object containing magnetic circuit description, i.e. magnetic circuit
elements, their relations, characteristics, dependencies, etc.
3.3. GRAPHICAL USER INTERFACE 81
.cbe – CircuitBuilder object containing electric circuit description.
.cc – CircuitConnector object containing the magnetic-electric connection specification.
.Base – The structure which contains the reference valules for per-unit calculations.
.Mode – The structure which describes the values of mode calculations (stator and rotor
currents and voltages in complex form along with other necessary values).
Besides, though all necessary matrices can be obtained from CircuitBuilder objects,
ModelData structure cashes frequently used matrices (like electric and magnetic incidence
matrices, etc) and the state matrices: voltages, currents, magnetic potentials, fluxes. Those
last reflect the state of the model after each calculation. They are used to calculate the
values for post processing after each state calculation. Additional information like number
of iterations and progress callback functions is also provided.
3.3 Graphical User Interface

The graphical user interface (GUI) is designed for the creation of the project, PN-model gen-
eration according to the input data of turbogenerator, verification of the generated model,
calculation of the operating modes and characteristics.
3.3.1 Program Main Window

Typying the command TurboTCM in the command line of Matlab will open the main window
of the program, shown on Figure 3.5.
The main window represents the project structure as a tree. It allows to set up the
parameters of each part of the turbogenerator and parameters of simulation, analyze the
parameters, generate the PN-model, analyze the model elements’ characteristics, calculate
the operating modes and characteristics. The project, generated model and the results can
be stored in corresponding files.
Project files
Each saved project is represented at least by one file. This project file has a binary format
and the extention .tcmprj.
If the project is saved after the model generation, another file is created. It contains
the model description with extention .tcmmdl. Next time the project is loaded that file is
loaded as well, thus the regeneration of the model is not necessary.
If the project has calculated results, the last are saved in separate file with extention
.tcmres. Similarly to the model file, the results file is loaded along with the project next
time.
3.3.2 Setting up project data

The following data of turbogenerator must be defined in order to generate the model:
• common geometrical data of stator and rotor;

Figure 3.5: Program main window
• distribution and geometry of stator and rotor slots;
• winding parameters of stator and rotor;
• magnetic parameters of the cores;
• axial configuration of the turbogenerator;
• simulation parameters.
The parameters windows can be open using any following way:
• double-clicking the necessary item of a project tree;
• choosing the menu item in the menu Project.
Machine’s rated data

The window of rated data setup is shown on Figure 3.6. It allows to setup the following
parameters:
• pole pair number;

Figure 3.6: Setting up rated data
• full power of turbogenerator;
• rated linear voltage;
• rated power factor;
• supposed excitation current (it can be recalculated and replaced as soon as model is
generated);
• rated rotation speed.
Stator parameters
The stator parameters window is presented on Figure 3.7. The top-left fields allow to set up
the basic dimensions of the stator magnetic system: external and internal (bore) diameters,
full length of the magnetic syste, number of slots (teeth). The Ventilation section allows to
set the number and width of radial ventilation ducts. The Material section allows to set
the lamination factor and load the BH-curve from the external file by pressing the button
Load BH-curve. As soon as it is loaded it is shown in the right graph in the window.
TurboTCM supports two formats of BH-curve files:
• simple text format files, where the information is stored in two columns: the first one
contains the values of field strength H, (A/m), the second one – values of flux density
B, (T ); each line forms one point of the curve;
• .bh-files, saved by VectorFields Opera 2D.

Figure 3.7: Stator parameters setup
The slot parameters and winding parameters can be defined by pressing the corre-
sponding buttons.
Slot distribution
The choice of slot type and their distribution in a magnetic system are defined in the window
shown on Figure 3.8.
Figure 3.8: Setting up stator slot parameters
TurboTCM supports slot distribution symmetrically-defined relatively to one pole, i.e.

the slots can have different parameters and even types between a pole and q−axis. This
feature is mostly used for the rotor definition, since stators usually have even and uniform
slot distribution.
The buttons Add and Remove can be used to add or remove slots from the defintion.
The Distribution section allows to define the angular distribution of the slots in the
magnetic system. This feature will be discussed later in Section 3.3.2.
In order to set the slot parameters, at least one slot must be defined in the list. Thus
the last slot cannot be deleted. Its parameters can be set using the button Edit....
Slot parameters
The window shown on Figure 3.9 allows to define the slot parameters.
Figure 3.9: Defining slot parameters
The fields (1) − (7) serve to define general slot dimensions and number of coils in it.
Parameters (8) − (12) define the wedge sizes along with the form of the top wedge. The coil
parameters, including the number and dimensions of elementary bars, isolation and etc, can
be defined in fields (13) − (23).
All the changes in geometrical parameters are immidiately reflected on the view of the
slot on the right side of the window.
Notes:
• Parameter (7) defining the number of wires per slot mustn’t be less than number of
coils in slot and divible by them. That parameter determines the real number of turns
in the winding.
• The product of the number of rows (15) and columns (16) of bars can differ from the
entire number of bars in coil (17), because some coils may have some rows not filled
completely by the bars. For example, for two-column coil the top and the bottom
rows usually have only one bar.
Rotor parameters
The parameters of rotor basically have the same structure as those of stator, thus here only
the particularities of the rotor will be discussed.
The window for the parameters of the rotor is the same as that of stator, shown on
Figure 3.7. Generally, the inner diameter of the rotor is zero. But main differences concern
the parameters of slot distribution.
The peculiarity of the parameters of rotor slot distribution of two-poles turbogenerator
is shown on Figure 3.10.
Figure 3.10: Slot distribution window for the rotor
The left part of the window allows to set different slot configurations; the right part –
Figure 3.11: Rotor slot distribution
angular intervals between them (tooth pitches). For the rotor at least one interval must be
defined – it is used to calculate the rotor pole pitch.
If more than one interval is specified along with more than one slot, their distribution
is calculated symmetrically realively to the pole axis and between the pole and q-axis.
Figure 3.11 presents the distribution of three different slots (Slot 1, Slot 2, Slot 3) with
four different tooth pitches (α1 , α2 , α3 , α4 ).
The number of possible variants of slots and tooth pitches is limited:
Zr
Nlim = (3.1)
4·p
where Zr – number of rotor slots, p – number of pole pairs.

The definition of the parameters of each slot is similar to the that of stator.
Simulation parameters
Simulation parameters define the division of the magnetic systems of stator and rotor on
layers. Window presented on Figure 3.12 allows to set the following parameters:
• number of layers in tooth zone of stator;
• number of layers in yoke zone of stator;
• number of layers in tooth zone of rotor;
• number of layers in yoke (core) zone of rotor (for the machines with p ≥ 2).
Figure 3.12: Simulation parameters window
The number of layers in tooth zones actually defines the number of slot leakage per-
meances per slot height without the slit. Besides, tooth zone always contains branches,
representing tooth top parts (slit zone) and tooth top leakages.
The PN of a central zone of rotor is fixed and its discretization is defined by number
of rotor slots (see Section 2.2.3 for details).
The Use FEM checkbox allows to chose the method of air-gap permeances calculation:
finite element method with special boundary conditions according to TCM3 or analytical
method presented in section 2.2.4.
The Take into account axial effects checkbox allows to switch on/off taking into account
the influence of ventilation ducts on the field of slots and air-gap of the machine according
to the method described in Section 2.3
3.3.3 Model generation and its characteristic view

As soon as all input data describing turbogenerator are filled down, a PN-model must be
generated. It can be realized by chosing the menu item Calculations→Generate in the main
3
To use this feature a PDE Toolbox for Matlab must be installed.
window of the program (Figure 3.5) or by pressing corresponding button in the toolbar.
After the model generation is finished cross-section of a machine will be shown on right side
of window. This view provides generated zones of stator and rotor (the central part of the
rotor is not shown on the view).
Upon the generation of the model the user is asked to recalculate excitation current
for rated mode (Figure 3.13)
Figure 3.13: Calculation of excitation current
If the generation of the model was successfull the menu items of characteristic visual-
ization and calculation become available.
Viewing air-gap characteristics

A window of the air-gap characteristic viewer is open by chosing menu item Calculations-
>View airgap characteristics (Figur 3.14).
The teeth of stator and rotor between which the user wants to see the characteristics
can be chosen by mouse using the list boxes on the left side of the window. The user can
chose either one stator tooth and one rotor tooth or several teeth from each core. The
range of teeth can be chosen by mouse holding Shift-key and different teeth can be chosen
holding Ctrl-key. Note that the first and middle teeth of rotor are considered to be poles.
The control panel on the top of the window allows to zoom in/out a part of the graph,
shoft it, switch on/off the legend, etc... A prepared graph can be shown in a separate
standard figure Matlab window for its further printing or modification. It can be done
using a button Open in a new figure.
All model’s characteristics viewer

Actually, all the model elements’ characterics (including nonlinear) can be viewed by chosing
the menu item Calculations→View all characteristics (Figure 3.15).
The characteristics of elements (including air-gap) can be chosen according to their
name from the list on the left of the window. In the example presented on Figure 3.15, only
radial rotor pole characteristics are chosen.
The checkbox Invert axes allows to rotate the graphic view on 90◦, i.e. swap x and y
axes.
3.3.4 Calculating the modes and characteristics of turbogenerator

The following operating modes can be calculated using TurboTCM GUI:
Figure 3.14: Air-gap characteristic viewer
• single operating mode of turbogenerator, defined by stator voltage and current and
power factor;
• no-load characteristic;
• load characteristic of turbogenerator working on a simple charge;
• regulation characteristic of turbogenerator working for network;
• V-shaped characeristics of turbogenerator working for network;
• angular characteristics of turbogenerator working for network.
For each calculated characteristic the following information can be obtained:
• flux density and flux of air-gap, pole and rotor teeth, stator teeth and yoke;
• magnetic potentials of the air-gap, pole, rotor central core, stator and rotor tooth part,
stator yoke;
• electromagnetic torque and load angle;
• parameters xd and xq ;
• etc...
Figure 3.15: All characteristics viewer
Besides, the following parameters can be obtained for each point of each characteristic
or operating mode:
• air-gap flux density spacial distribution;
• graphical view of the magnetic field parameters (magnetic potentials, flux density,
fluxes) according to each element of PN.
The following sections provide the description of using GUI of TurboTCM to obtain the
particular results. The last are presented further for different machines in Chapter 4.
Operating mode calculation

The window allowing to calculate the operating mode of turbogenerator (Figure 3.16) can
be called by chosing the menu item Calculations → Mode Calculation in the main window
(or corresponding button in the toolbar).
This window allows to calculate one operating mode of the turbogenerator. In order
to start the calculation the following fields must be set:
• Mode type – Stator voltage, current and power factor – calculation of the mode defined
by stator voltage, current and power factor. In this case the excitation current will be
obtained as a result.
• Unit choice – Units or PerUnit – defines if the calculation should be performed in

regular units or per unit.
Figure 3.16: Operating mode calculation window
• Mode definition – defines the values of the mode: stator voltage and current, power
factor (PF), rotation speed and load type. These fields can be filled down by rated
values pressing the button Set to rated data. Pressing the button View rated data...
allows to view the rated data of the turbogenerator.
Upon the definition of the values of the mode the calculation can be performed pressing
Calculate button. A vector diagram on the top-right of the window will show the iteration
process in “real-time”. The information about convergence will be displayed in Output
window.
Finally the vector diagram will show the vectors of stator current Isc , stator voltage
Usc and EMF “beyond the end-winding leakage” Esc , obtained from the calculation.
As an example, several values can be observed upon the end of the calcualtion, like
potential distribution in the elements of PN, i.e. elements’ ampere-turns (Figure 3.17, top)
and flux density on the middle line of air-gap (Figure 3.17, down).
Figure 3.17: Ampere-turns of PN elements (top) and air-gap flux density (down)
After closing the calculation window, the results will be added to the result list of the
main window.
No-load characteristic
The no-load characteristic of turbogenerator is calculated as E1 = f (If ), where E1 – stator
winding EMF, If – turbogenerator excitation current. The rotation speed is constant. This
is a direct problem, thus there is no need in additional iterative process.
Chosing the menu item Calculations → No-load characteristic opens a no-load charac-
teristic parameters window (Figure 3.18). The excitation current must be set up as a vector,
rotation speed – scalar. All the values must be set in per unit. The reference value for the
excitation current is that calculated after the PN-model generation for the rated load. The
calculation will start as soon as OK button is pressed.
Figure 3.18: Input data for no-load characteristic
In case of succeffull calculation the no-load characteristic will be added to the result
list. It can be viewed by double-clicking on it in the result list.
Note: setting vector values. Vector values are widely used while setting the input data
for calculations. The vector must be set using standard Matlab notation. Here are several
examples:
• [0.1:0.2:1.2] – means a range of values from 0.1 to 1.2 with step 0.2.
• [0.1:0.05:0.6 0.7:0.1:1 1.2] – a range of values from 0.1 to 0.6 with step 0.05,
then from 0.7 to 1.0 with step 0.1 and finally a single value 1.2.
For detailed information see the Matlab help.
Load characteristic
The load characteristic is calculated on a single load for defined values of stator current,
excitation current and power factor as Us = f (Is ), where Us – stator voltage, Is – stator
current.
The calculation will start upon chosing the Calculations → Load characteristic menu
and setting input data in the window shown on Figure 3.19). This window allows to define
the vector of stator current values, excitation current, PF and rotation speed. All data must
be defined in per unit.
Attention! The load type (inductive or capacitive) is defined by the sign of the PF:
positive PF means active-inductive load, negative – active-capacitive. For the PF equal to
1 its sign is ignored.
Figure 3.19: Input data for load characteristic
Upon the end of calculation the characteristic will be added to the end of the result
list and can be viewed from there.
Regulation characteristic of turbogenerator

The regulation characteristic of turbogenerator on parallel work with network is calculated
for fixed value of stator voltage and power factor as If = f (Is ), where If – excitation current,
Is – stator current.
The calculation is launched from the menu item Calculations → Regulation charac-
teristic. The input data window will show up (Figure 3.20) and allow to enter the valuse of
stator current vector, stator voltage, PF and rotation speed. All data must be defined in
per unit. The rules of setting load types were discribed in section 3.3.4.
The calculated characteristic will be added to the end of the result list.
V-shaped characteristics calculation

A V-shaped characteristic of turbogenerator is calculated for fixed values of stator voltage
and active power as Is = f (If ), where If – excitation current, Is – stator current. The
excitation current is given two ranges: underexcited and overexcited.
The menu item Calculations → V-shaped characteristic opens the input dialog (Fig-
ure 3.21) which allows to define the following values (per unit):
• vector of stator current values in underexcited zone;
• vector of stator current values in overexcited zone;
• stator voltage;
Figure 3.20: Input data for regulation characteristic
• active power;
• rotation speed.
Note that the values of stator current in p.u. must not be less than defined active
power in p.u. since such operating modes don’t exist and these points will be excluded from
the calculation.
Figure 3.21: Input data for V-shaped characteristic calculations

3.4. CONCLUSION 97
Angular characteristic calculation

The angular characteristic of turbogenerator is calculated for defined stator voltage and
excitation current as P = f (θ), where P – active electric power, θ – load angle.
Similarly to the previous characteristic, that is launched from the menu item Calcula-
tions → Angular characteristic and input data dialog window (Figure 3.22).
The results of calculation are similarly stored in the end of the result list. Additional
characteristics are shown in the results of the calcualtion (Figure 3.23): active and reactive
power, stator current and PF as function of load angle. The red cross on angular charac-
teristic marks its maximum, which load angle θmax corresponds for. That angle determines
the range of stable work of turbogenerator. The value of θmax is shown under the active
power graph.
3.4 Conclusion
This chapter presented the approach to software implementation of the methods expounded
in Chapter 2. A description of the software package was given along with its capabilities.
The software was designed on two different levels: simulation core and almost inde-
pendent graphical user interface. This conception allows to separate the tasks, gives the
user flexible and convenient methods to create a project, enter a huge amount of machine’s
parameters, perform some calculations, view the results. From the other hand, it offers very
flexible programming interface aimed to thoroughly use of all the capabilities of the model,
creating user-defined scripts. The encapsulation of the calculation core gives the possibility
to couple TurboTCM with the optimization supervisor (see Section 4.4 for examples) or any
other software written in Matlab.
Figure 3.22: Input data for angular characteristic
Figure 3.23: Results of angular characteristic calculation

Chapter 4
Application
The proposed approach was tested on several kinds of synchronous machines during the
evolution of the software. Since the used method is rather universal, in the beginning
a small synchronous electrical machine (3 kW) was produced in order to perform some
adjustment of the method. The fluxes in different parts of the machines were calculated
by TurboTCM and then compared with those from OPERA. No-load, short-circuit and load
characteristics were compared with experimental results.
After the application to a small machine, the software was developed in order to be
applied to two-poles turbogenerator. We were provided by Jeumont with experimental data
in no-load and short-circuit cases for several machines. These results were compared with
TurboTCM calculations.
The software allows calculating synchronous non-salient pole machines with any num-
ber of poles. In 4-pole turbogenerator case we had to extend a little bit the typical PN of
the machine in order to reflect that particular case. The modified PN, some calculation and
experimental results are also presented in this chapter.
4.1 Synchronous machine of 3kW

In the beginning, in order to “tune-up” the model of a synchronous machine, to reflect and
verify its particularities, a small machine – prototype – was studied. The machine had the
properites that are particular to turbogenerators, like different rotor slot pitches. For that
machine the primary version of the software was realized. It took into account not so many
input parameters as for turbogenerators but helped to define and optimize main structure
of the software by means of try-and-error method.
4.1.1 Machine description

The synchronous machine used in the study was constructed based on an asynchronous
motor of 3 kW. The stator was left the same, while the rotor was producted complying with
the specification reflecting the particularities of a turbogenerator.
The main parameters and dimensions of the machine are presented in Table 4.1. The
stator of the machine has trapezoidal slots with the pouring winding. The rotor has slots
with different slot pitches and is presented on Figure 4.1.
99
100 CHAPTER 4. APPLICATION
Rated data
Total power 3000 VA
Stator phase voltage 220 V
Power factor 0.8
Number of poles 2
Stator data Rotor data
Outer diameter 164 mm Outer diameter 92 mm
Inner diameter 94 mm Shaft diameter 36 mm
Core length 110 mm Core length 110 mm
Number of slots 24 Number of slots 16
Slot type trapezoidal Slot type rectangular
Slot height 15.52 mm
Slot height 14 mm
Slot width on top 9.45 mm Slot width 4 mm
Slit width 2.7 mm
Slit height 0.68 mm
Number of turns 100 Number of turns 440
Slot pitches 17.14◦ /16.36◦
Table 4.1: Dimensions of the 3kW machine
Figure 4.1: Rotor dimensions

4.1. SYNCHRONOUS MACHINE OF 3KW 101
The machine’s shaft is made of a soft steel. It differs from the other part of the rotor.
The BH-curves of stator, rotor and shaft materials are presented on Figure 4.2.
Figure 4.2: Steel BH-curves: rotor (left), stator(right) and shaft (bottom).
4.1.2 Machine’s permeance network

The permeance network of the machine is generated using the method discussed in Chap-
ter 2. The difference can be found in rotor: the shaft is modelized by an additional layer of
nonlinear permeances.
Figure 4.3 shows a simplified rotor scheme, outlining the particularities of this model.
Teeth permeances are shown on the figure while slot permeance branches are omitted.
Permeances Λyh1 − Λyh8 represent the D-axis oriented branches corresponding to the yoke
part of the rotor. Permeances Λsh1 − Λsh4 represent D-axis oriented branches corresponding
to the shaft part of the rotor. Branches Λyv1 −Λyv5 and Λsv1 −Λsv3 represent Q-axis oriented
flux corresponding to the yoke and shaft parts respectively. All D-axis oriented branches
also contain the MFF-sources which are omitted on the figure.
4.1.3 Calculation results

In order to validate the model, the obtained results were compared with FE-calculation
software OPERA-RM. A fragment of the OPERA model with generated mesh is shown
on Figure 4.4. The mesh is rather complicated and irregular. The main part of energy
conversion – the air-gap – has the finest mesh, since the gradient of flux density there is
the highest. Special techniques must be applied to reflect the mutual movement of stator
Figure 4.3: Rotor simplified scheme
and rotor cores, otherwise remeshing should be performed as soon as rotor position changes.
From the other hand, the parts like stator and rotor teeth and yokes have also relatively
high meshing, though flux density gradient is relatively small there. These factors increase
calculation time while using finite element method, especially in load cases, when several
rotation periods should be calculated to obtain only one point of the characteristic.
Figure 4.4: OPERA-RM model (fragment).

Flux comparison
The flux comparison was performed in static mode calculation for different field currents.
We compared the fluxes in the middle section of stator tooth aligned with rotor pole and the
part of stator yoke where the flux was maximal. The results of calculations are presented
in Table 4.2 and Figure 4.5. The maximal errors are 4, 927% for tooth fluxes and 2, 649%
for yoke fluxes.
Teeth Yoke
J,A Opera TurboTCM Error Opera TurboTCM Error
0,1 3,2660E-05 3,4264E-05 4,911% 1,1440E-04 1,1734E-04 2,567%
0,2 6,5310E-05 6,8528E-05 4,927% 2,2910E-04 2,3467E-04 2,433%
0,5 1,6330E-04 1,7132E-04 4,911% 5,7300E-04 5,8668E-04 2,388%
1,0 3,2660E-04 3,4264E-04 4,911% 1,1460E-03 1,1734E-03 2,388%
1,5 4,8910E-04 5,1278E-04 4,843% 1,7160E-03 1,7556E-03 2,305%
2,0 6,4890E-04 6,7881E-04 4,609% 2,2759E-03 2,3232E-03 2,076%
4,0 1,07700E-03 1,09277E-03 1,464% 3,63550E-03 3,59360E-03 1,153%
6,0 1,19500E-03 1,20593E-03 0,914% 3,95670E-03 3,88127E-03 1,906%
8,0 1,25800E-03 1,26192E-03 0,312% 4,15470E-03 4,05694E-03 2,353%
10,0 1,30450E-03 1,30199E-03 0,192% 4,30900E-03 4,19483E-03 2,649%
Table 4.2: Flux calculation results
Figure 4.5: Fluxes comparison: TurboTCM – OPERA RM
Static characteristics comparison

No-load and load characteristics are shown on Figure 4.6. TurboTCM result is compared with
OPERA RM one. That finite element package performs the calculation of the EMF on a
full period of rotation. The result is RMS value of the obtained EMF for different field
currents. Maximal error is 6, 39% for the field current of 20A.
For a fixed field current we obtained the load characteristic of the machine Vs (Is )
(Figure 4.6, right). The maximal error is 4, 14% for stator current of 13A.
Figure 4.6: Noload (left) and load (right) characteristics: TurboTCM – OPERA RM
Dynamic simulation
Figure 4.7 presents dynamic no-load calculations performed in OPERA RM and TurboTCM.
EMF curves are shown on the left while their spectrum on the right. The data were obtained
for 180 positions of the rotor in TurboTCM and 100 positions in OPERA RM for the field
current of 10A. OPERA RM limits 100 positions per simulation, thus to obtain 180 positions
we should have performed two simulations one by one.
Figure 4.7: Dynamic calculation: TurboTCM – OPERA RM
Table 4.3 shows the calculation times in OPERA RM and TurboTCM for different
field currents. The results show that even without saturation the time gain is 266, 13 times,
while in high saturated zone the gain becomes 672, 13 times.
J, A TurboTCM OPERA RM Difference, times

1 12,4sec 0h55min 266,13
10 18,1sec 3h06min 616,57
20 18,3sec 3h25min 672,13
Table 4.3: Calculation time comparison
Experimental results
Experimental workbench shown on Figure 4.8 allows to obtain the experimental results for
different modes. It consists of a generator prototype driven by a variable speed asynchronous
motor with inverter which permit to run the machine on speed up to 3000 rpm. The
acquisition equipment form the data channel to the computer.
Figure 4.8: Experimental workbench
Two characteristics – noload and regulation – were obtained for the comparison (Fig-
ure 4.9). Simulation data show rather good correlation. But the remanent magnetization of
the rotor causes additional error, which can be easily seen on a no-load characteristic and
also affects the regulation characteristic. Besides, we had no information about the mag-
netic properties of the rotor and stator material, thus the BH-curves used for simulation
were estimated.
Figure 4.10 presents the comparison of the EMF in dynamics. To the left one can see
the EMF and its spectrum obtained by simulation while to the right – the experimental
results. The amplitude, form and harmonic orders (h3 , h21 and h23 ) are almost the same.
Harmonics h21 and h23 are caused by tooth effects, since rotor slots are opened.
Figure 4.9: Static characteristics: comparison with experiment
Figure 4.10: Rotation effect: experimental comparison

4.2. TWO POLE TURBOGENERATOR 107
4.2 Two pole turbogenerator

This section presents the results of application of the approach to a two poles turbogenerator
of 10-100 MW. Some of these results were presented in [42, 38, 40, 39, 41].
4.2.1 Machine description and its permeance network

The turbogenerators are driven by gas turbines having constant synchronous speed of 3000
or 3600 rpm. The range of machines of 10-100 MW have direct air cooling of rotor winding
and indirect air cooling of stator.
Rotor has concentric winding stacked in massif magnetic shaft. The end windings are
maintained by non-magnetic bands. Rotor slots have under-winding ducts and perforated
conductors, thus having air flow from the cooling duct directed to the air-gap. Stator has
three phase bar winding (two bars per slot) connected in the manner to realize necessary
circuit and to obtain rated output voltage with ideally magnetized magnetic circuit and
minimized voltage harmonics.
The end windings and coupling circuits are firmly supported and wedged to avoid the
vibrations that damage the isolation. Copper magnetic shield protects the magnetic system
of the induced currents caused by the magnetic field of end windings.
The PN of the turbogenerator is built using the principles explained in Chapter 2.

In order to validate the model, the obtained results were compared with experimental data
in no load case and short circuit case1 . Since the experimental result under load were not
provided, the rated mode was compared with other calculations by Potier method and by
means of native Jeumont software TurboB. TurboB, TurboB2 and TurboB4 are the software
packages realized in Fortran and devoted to the calculation of 2-poles, 4-poles and n-poles
turbogenerators. They implement analytical approach of turbogenerator’s characteristics
calculation and are in industrial use in Jeumont.
In order to study several turbogenerators Jeumont proposed an internship to the stu-
dent of ENSEM2 Ryth LY. The cases included the machines complying the range of 10-
100 MW were studied.
The parameters of studied machines are shown in the following table:
No-load case
No-load characteristic of a machine present stator voltage regarding excitation current with-
out any load.
Comparison data of no-load characteristics of the machines are presented in Ap-
pendix B in tables B.1-B.2, where: If – excitation current, A; Eexp – measured EMF,
V; Etcm – calculated EMF, V; Error – relative error, per cent. Graphical representation of
the characteristics is given on figures B.1-B.8.
1
Here and further all the measurements and machine parameters were provided by Jeumont Industry
Framatome ANP.
2
Ecole Nationale Superieure d’Electricite et de Mecanique Nancy.
Name Full power Stator voltage Frequency Power factor

Machine 1 63.000 MVA 11500 V 50 Hz 0.80
Machine 2 33.625 MVA 11000 V 50 Hz 0.80
Machine 3 50.900 MVA 11000 V 50 Hz 0.80
Machine 4 34.375 MVA 13800 V 50 Hz 0.80
Machine 5 51.375 MVA 11500 V 50 Hz 0.80
Machine 6 66.667 MVA 11000 V 50 Hz 0.90
Machine 7 41.688 MVA 11000 V 50 Hz 0.80
Machine 8 31.529 MVA 13800 V 60 Hz 0.80
Table 4.4: Studied machines’ parameters.
First points of all characteristics show big difference (the error is up to 92 percent)
between calculation and experimental results, caused by remanent magnetization of the rotor
which was not taken into account. In some cases (machines 1 and 3) the magnetization level
is so high that leads to completely absent correlation between measured and calculated data.
In fact, the direction of the remanent magnetization is always unknown, thus unpredictably
affecting the whole characteristic.
Another difference can be found in the saturated part of the no-load curve (for example,
Figure B.7). That difference occurs because of the insufficient data points of the BH-curve
of stator and rotor steels: only 3-5 points of the curves were provided.
Mean absolute values of the errors are shown for each machine in the following table.
Cases 1 and 3 have the biggest errors in the set (highlighted in bold).
Machine: 1 2 3 4 5 6 7 8
Error: 7,11% 2,41% 16,46% 1,03% 1,35% 1,67% 3,45% 2,47%
Short-circuit case
Short circuit characteristic of a machine is obtained by measuring stator current while
applying different excitation currents and stator voltage equal to zero (short circuit). In
this case the field of the machine is orientated in q-direction.
Comparison data of short circuit characteristics of the machines are presented in Ap-
pendix B in tables B.3-B.4, where: Is – stator current, A; If exp – measured excitation cur-
rent, A; If tcm – calculated excitation current, A; Error – relative error, per cent. Graphical
representation of the characteristics is given on figures B.9-B.16.
In short circuit mode the machine’s magnetic system stays mostly linear, thus the
calculation error is diminished. The comparison of the mean values of the errors is shown
in the following table3 :
Machine: 1 2 3 4 5 6 7 8
Error: 1,96% 0,38% 6,72% 2,71% 1,08% 1,21% 0,84% 0,37%
Rated mode comparison

Since we had not been provided with experimental data for any mode in charge, the calcu-
lation results obtained in TurboTCM were compared with that of TurboB and that obtained
3
The value of the error for the machine 8 is obtained without taking into account the first point since
excitation current less than 1 A cannot be measured. See tables B.3-B.4 for details.
by Potier method. The table 4.5 shows the comparison for the rated mode, where: Vs –
stator phase voltage, Is – stator phase current, COSϕ – power factor, N – rotation speed.
The results show rather big errors between TurboB and TurboTCM. From the other
hand, the error between Potier calculation and TurboTCM is small. The effect might be
caused by a difference of the calculation methods of TurboB and TurboTCM, in particular,
the method of taking into account nonlinear BH-curve of the machine.
In addition, it’s possible to determine electromagnetic torque calculated by TurboTCM:
1 X T ∂Λ
TT CM = · U · ·U (4.1)
2 ∂θ
where: U – magnetic potential differences, Λ - permeances, θ - rotor angle.
It’s possible to compare the result with the classical relation for the rated point ne-
glecting losses:
S · PF
TEM = (4.2)
Ω
where: S – full power, VA, P F – power factor, Ω – angular rotation speed, rad/sec.
The calculation error for rated mode is negligible:
TT CM = 1.006 · TEM (4.3)
Working mode characteristics

TurboTCM allows to calculate a big variety of static characteristics of turbogenerators. Here
some calculation results are presented without their validation.
Regulation characteristic is calculated for a constant stator voltage and load type
varying stator current. Figure 4.11 shows three characteristics for different load types –
capacitive, active and inductive – and rated voltage. Here: If – excitation current, Is –
stator current.
V-shaped characteristic is calculated for a constant active power and stator voltage,
varying stator current. A set of V-shaped characteristics for machine 1 for different values
of active power is shown on Figure 4.12, where: If – excitation current, p.u., Is – stator
current, p.u., P – active power, p.u. The calculation time of that whole set of characteristics
is less than 12 minutes on a computer Pentium IV, 1300 MHz, 784 MB RAM.
Additional results can be obtained for any point of any characteristics. It was interesting
to illustrate the variance of xd and xq parameters of the turbogenerator. These values
were calculated by the method described in Section 2.5.2. The regulation characteristic of
machine 1 for inductive load with power factor 0,8 and corresponding variation of xd and xq
are shown on Figure 4.13. The resulting dependence of these parameters on the operating
mode of the machine illustrates that real geometry of the turbogenerator was taken into
account.
110
Machine Vs , V Is , A cos ϕ N , rpm Excitation current, A Error

TurboB TurboTCM Potier TurboB Potier
Machine 1 6639,50 3162,90 0,8 3000 729 651,80 -10,59%
Machine 2 6350,90 1764,90 0,8 3000 756 736,10 -2,63%
Machine 3 7967,40 1438,10 0,8 3600 505 451,51 -10,59%
Machine 4 6350,90 2671,60 0,8 3000 720 703,12 716,87 -2,34% -1,82%
Machine 5 6639,50 2579,20 0,8 3000 662 595,73 607,14 -10,01% -1,88%
Machine 6 6350,90 3499,10 0,9 3000 696 630,46 -9,42%
Machine 7 6350,90 2188,10 0,8 3000 599 520,46 -13,11%
Machine 8 7967,40 1319,10 0,9 3600 591 548,61 551,72 -7,17% -0,56%
Table 4.5: Rated mode calculation comparison.
CHAPTER 4. APPLICATION
Figure 4.11: Regulation characteristic of machine 1.
Figure 4.12: V-shaped characteristic of machine 1.

Figure 4.13: Regulation characteristic of machine 1 and variation of its xd and xq parameters.
Flux density distribution in the air-gap sometimes can be very useful for an engineer,
for example, in surface loss calculations. Although TurboTCM does not use that distribution
and its calculation is base only on winding flux linkages, it implements that option. The
following figures show flux density distribution in the middle of the air-gap for three different
cases: no-load (Figure 4.14), rated mode (Figure 4.15) and short-circuit mode (Figure 4.16).
The flux distribution is shown on the left side of each figure along with its spectrum under.
The pictures to the right show the distribution of the ampere-turns in the elements of the
PN of the machine corresponding to each case. In all cases on the spectrum one can see
the influence of stator and rotor tooth harmonics (39 and 41). Rather symmetric picture
in no-load mode is changed in rated mode, thus reflecting the armature reaction, which is
even bigger in short-circuit mode.
Figure 4.14: No-load air-gap flux density distribution.

Figure 4.15: Rated mode air-gap flux density distribution.
Figure 4.16: Short-circuit mode air-gap flux density distribution.

4.3 Multi-pole turbogenerator

This section presents the application of TurboTCM to the 4-pole machine of the same power
range 10-100 MW. The construction of the PN of multi-pole generator was discussed Sec-
tion 2.2.3.
One of the advantages of TurboTCM is the possibility of incorporating different ge-
ometries of the machines and their parts. Adaptive PN allows to take into account almost
any slot types, rotor forms without additional programming or with small add-on code in
Matlab, depending on the complexity of the problem.
Here the PN of a specific 4-pole case is shown and some calculation results are pre-
sented.
4.3.1 Jeumont 4-pole turbogenerator

Although the PN of the 4-pole turbogenerator doesn’t differ too much from that of n-
pole generator, there are some particularities in its rotor, which are specific only for 4-pole
configuration. The machine which parameters were provided by Jeumont Industry has
laminated core with a hole in the center, where it is attached to the solid shaft with arms
(see Figure 4.17). That hole is used for rotor ventilation in radial direction. The arms and
shaft can not be neglected since in 4-pole configuration the flux also passes by them.
Figure 4.17: Rotor of the 4-pole turbogenerator.
The method of taking into account that special rotor configuration was realized by
David Bonnet4 , whose internship was devoted to the implementation of 4-pole machine as
a module of TurboTCM.
Figure 4.18 shows the definition of the rotor zones. The minimal and maximal diam-
eters define the limits of the new yoke zone. Additional BH-curve is given for the arm and
4
David BONNET was a student from Ecole Nationale Superieure d’Ingenieurs de Limoges, who passed
his internship in Jeumont Industry and Ecole Centrale de Lille in 2005.
4.3. MULTI-POLE TURBOGENERATOR 115
Figure 4.18: Taking into account rotor arms
shaft zones. The yoke zone in its turn is connected with tooth zone of the rotor which was
not changed regarding the method described in Chapter 2. The distribution of the elements
in the yoke, arm and shaft zones is illustrated on Figure 4.19. Red horizontal permeances
implement the network of the yoke part between the poles (MMF sources are not shown
here since they are placed regularly). Blue vertical permeances implement the arms. The
common MMF source which combines all the currents of the above slots is placed between
adjacent arms.
Figure 4.19: 4-pole rotor permeance network.
Figure 4.20 presents some no-load calculation results for the machine of full power
S = 66667kV A, stator voltage U = 11kV , power factor cos ϕ = 0.9 and rotation speed
3000rpm. The comparison was performed between linear and nonlinear calculations and
experiment. The BH-characteristics of the arms and rotor were unknown and estimated,
thus giving rather big error in non-linear case.
Figure 4.20: No-load characteristic comparison for 4-pole machine.

4.4. OPTIMIZATION APPLICATION EXAMPLE 117
4.4 Optimization application example

In order to study the variation of different operating values (like EMF harmonics, torque,
etc) depending on machine’s geometrical parameters (like machine’s length, tooth and yoke
dimensions, pole dimensions and etc) TurboTCM was coupled with Sophemis. The last tool
was designed in Ecole Centrale de Lille, L2EP by Stephan Vivier (see, for example, [57])
and implements experimental design approach.
The term “Experimental Design” is a methodology that is used for behavioral charac-
terization of a system. It is based on modification and measurement of proper variables of a
considered object, which comprises studied values along with their variation sources. Thus
his method allows to determine the relationship between two types of variables: the response
– studied physical values, and the factors – physical values that are defined by experimenter
and influence the responses. To be more precise, it shows the links between responses and
factors along with those between responses and factor interactions. This permits to build a
model expressing the response in function of the factors.
First we present the objective of the study followed by some results. Other results can
be found in [43].
4.4.1 Objective
Our objective was to demonstrate the use of TurboTCM together with an optimization su-
pervisor, thus rather simple variables were chosen:
• hs1 – stator tooth height;
• bs1 – stator tooth width;
• Di1 – stator boring diameter (affects air-gap value);
• Tp1 – rotor pole width.

Their variation limits are presented in the following table:
Factors Lower bound Upper bound
hs1 hs1 − 10% hs1 + 10%
bs1 bs1 − 10% bs1 + 10%
Di1 Di1 − 5% Di1 + 5%
Tp1 Tp1 − 10% Tp1 + 10%
The chosen responses are:
• KhB3 – 3rd order harmonic of the air-gap flux density in no-load;
• KhE3 – 3rd order harmonic of the stator EMF in no-load;
• KhE1 – the fundamental of the no-load stator EMF;
• If – excitation rotor current in no-load.

The chosen factor set is a complete set of 3 levels with 4 factors which leads to 81
experiments.
Figure 4.21: KhB3 for Tp1 min (left) and KhB3 for Tp1 max (right)

Results are obtained for a fixed diameter Di = 1.05Di1 and, thus, maximal value of the
air-gap. Therefore, one can see that if the pole width increases, the 3rd harmonic of air-gap
flux density decreases. The right picture on the Figure 4.21 shows very interesting result of
the minimum of the 3rd harmonic.
The following results present the analysis of the 3rd and 1st harmonics of stator EMF.
Figure 4.22 shows the influence of the stator tooth height and width on these values. Lower
surfaces on that figure correspond to smaller Tp1 .
It took 25 minutes to perform 81 simulations on Pentium IV, 1.9GHz and that time
includes also the network regeneration time, since the parameters of its elements change
depending on the variable values.
The calculation of excitation current If was performed for the same 4 parameters with
only two extreme values of the diameter and the results are shown on Figure 4.23. The
lowest surface corresponds to the minimal value of the pole width. The results are rather
obvious: excitation current depends on the value of the air-gag, the pole width is in inverse
propotion with that current. The value of If increases with increase of the stator tooth
height since (stator yoke decreases in that case).
4.5 Conclusion
The aim of this chapter was to validate the approach and the software and to present some
additional possible application.
On the very first stage of the thesis we implemented a model of a small generator,
which had a particular rotor with slot distribution specific for turbogenerators. The results
were validated by experiment and compared with the calculation by FEM. The comparison
4.5. CONCLUSION 119
Figure 4.22: KhE3 (left) and KhE1 (right) for different Tp1 .
Figure 4.23: If for Di1 min (left) and Di1 max for different Tp1 .
gave an acceptable results while the gain in calculation time regarding FEM-calculation was
of order of 600 times. We obtained all the expected effects in the simulation like tooth
harmonics of EMF, saturation, etc. From the other hand, we found out that the method is
very sensitive to the material characteristics, thus the last must be obtained and measured
carefully.
The software was then developed to implement a 2-pole turbogenerator model taking
into account all its parameters including axial heterogeinity effects. The calculation results
were compared with experimental data for eight different machines in no-load, short-circuit
cases and rated operating mode. In some cases the provided test results were obtained on
machines with remanent rotor magnetization, which affected the experimental data leading
to significantly bigger errors while comparing them with simulation. Also we presented
the results of calculation of different machine’s characteristics. Although nowadays regu-
lar methods use linear dependency for xd and xq , our calculations show their variation in
different operating modes.
An application of the approach and the software to 4-pole turbogenerator demanded
additional changes to the PN. The shaft and support arms were taken into account along
with variable section of the rotor yoke. It highlights the modifiability of TurboTCM, the
easiness of reflecting different geometries of a machine using adaptive PN. Some comparison
results were given.
Finally, we introduced the coupling of TurboTCM with optimization supervisor Sophemis,
implementing experimental design method. That coupling allows to determine influencing
factors, that can help to limit the number of variables for turbogenerator computer aided
design. Using the response surface permits to determine an equivalent model, which can be
used in optimization. The results and calculation time open good perspectives for further
use of the model coupled with optimization software.
Conclusion and perspectives
In this work we proposed a rapid, flexible optimization-oriented approach and realized a tool
to calculate the operating modes and characteristics of turbogenerators 10-100 MW using
the concepts of “virtual prototyping”. That approach is based on a theoretically validated
tooth contour method. The method was developed in order to implement the complex
geometrical structure of modern turbogenerators, take into account their particularities,
generalize and automatize the approach of PN generation. The software TurboTCM was
realized, which gives flexible possibilities to turbogenerator characteristics calculation and
can be coupled to an optimization supervisor or other calculation tool in Matlab.
In the first chapter we presented the overview of the problem of electromagnetic ma-
chines’ analysis, existing methods that solve different aspects of the problem: from basic
conventional approaches to modern numeric solutions. Since our method is based on per-
meance networks, we paid more attention on highlighting the existing solutions in that
field.
The second chapter describes general approach in details. The theoretical basis and the
way of permeance network construction, calculation of its parameters were shown without
links to any particular type of the machines. However, we couldn’t completely abstract away
from the particularities of turbogenerators, the complexity of their structure, thus giving
variable topology model of the rotor and some calculation algorithms in respect to them.
The main idea of the approach is to exploit the particularities of a machine to min-
imize the number of the network elements. The formed calculation method represents a
development of the combined approach to electrical machine analysis existing from the
dawn of electromechanics. On the first stage we divide the volume of the machine onto
zones with independent magnetic fields and well-expressed boundaries. Then we form a
non-planar discrete scheme of the magnetic system. On a coupling stage we introduce the
MMF sources representing slot currents into the scheme. And on final stage we perform
iterative calculation of the flux linkages.
The precision of air-gap permeances calculation dictates the precision of the whole
model. The approach doesn’t insist on the way of that calculation: it can be either finite
element or analytical or any other method. We introduced the analytical approximative
approach giving the possibility to calculate air-gap permeance characteristics rapidly and
with high precision.
We showed that special techniques must be implemented to properly reflect axial
heterogeinity of magnetic cores while using a model “in-plane”. Although general way of
taking it into account is rather simple, one should remember that different parts of the
machine are differently influenced by that heterogeinity. We showed five characteristic zones
of a machine and described the implementation of the approach on the level of permeance
network parameters. However, the proposed method is rather general and can be easily used
121
with other methods of machine’s characteristic calculations, even FEM (in plane).
The use of adaptive PN construction helps to take into account complicated geometries
of magnetic cores, irregular slot structure and distribution. The implementation of the
algorithm allows to pass to other slot geometries and even rotor or stator geometries with
minimal software modification using “add-on” principle.
Fixed rotor algorithm described in Section 2.5 offers rapid steady-state calculations,
as a compromise between precision and calculation time. Anyway, for thorough harmonic
analysis rotor rotation is easily implemented.
Chapter 3 presents the implementation of the software TurboTCM, created in Matlab.
The separation of the software on two levels – mathematic core and graphical user interface
– leads to the advantage of using the core in a couple with any other software, like opti-
mization supervisors, leaving the function of project creation and results management to
the user interface. Besides, the core itself gives very flexible control of the whole model and
its behavior. The separation of the objects which are the circuit containers and couplers
facilitates the future coupling with other possible models.
In the beginning the software was tested on a small 3 kW synchronous generator,
which was specially constructed to give the possibility of model tuning. The results were
compared with finite element analysis and experimental data and showed good correlation.
Then a two-pole turbogenerator model was realized and tested on several different
machines of the selected power range. The no-load, short-circuit characteristics and some
modes under load were compared with experimental results and gave relatively small errors
in most cases. The cases of great errors are explained by high remanent magnetization of the
rotors of these machines which can be easily seen in the beginning of their no-load curves.
Besides, we realized the model of a four-pole turbogenerator with a particular construction
of the rotor, which demanded additional permeance network modification.
Finally, we realized the coupling of TurboTCM with the optimization supervisor and
performed some tests on simple variables. The simplicity of model creation and coupling
with other tools proved the effectiveness of the approach and the modularity of the software
in the tasks of optimization, demonstrating high speed of response surface determination.
The developed approach and designed software give the possibilities to further devel-
opment and extension of the model by multi-physical phenomena. The architecture of the
tool is open for its future extension by vibrational, acoustic and especially thermal models.
The following points show our perspectives in the nearest future:
• Expand the application of the approach and designed software to other types of ma-
chines. The implementation of TurboTCM already allows to take into consideration
cores of any geometry.
• Concerning asynchronous machines: interesting results can be achieved by coupling
the magnetic permeance network with thermal equivalent circuits and their joint cal-
culation. This work is already an object of a CDD (Contrat de Durée Déterminé)
“Futurelec3 – operation 1-bis” which started in October 2006. Normally, multiphys-
ical models are based on linear magnetic models which take into account only the
permeances of air-gap. It seems to be more realistic to replace them by a PN model
of an asynchronous motor, which allows to avoid many restrictions.
• Improve the calculation of slot leakage permeances and air-gap permeances. There
are many methods of air-gap permeance determination exist. They must be compared
4.5. CONCLUSION 123
together within unique calculation tool. This is rather big domain of work since the
tool must be rebuilt generic.
Taking into account the rotation allows to extend the model to perform in future tran-
sient analysis of the turbogenerators taking into account the Foucault currents in massive
parts of the machines.
Appendix A
Matlab Solver functions
This chapter presents a list of main functions of TurboTCM and their brief descriptions.
A.1 General functions
General functions concern first of all TurboTCM project management. An empty project
can be created by TCMProject_Create function. If the project was generated by GUI and
saved in the file, then TCMProject_Load function can be used to load the project.
In order to use the saved project and calculate the model, ModelData must be gener-
ated first. And it must be regenerated each time the input parameters are changed. The
following example shows how to use it:
Prj = TCMProject_Load(’..\2p - TURBO.tcmprj’); % load a project

P = Prj.Parameters; % Extract the parameters to the variable P
MD = GenerateModelData(P); % Generate model data
%--- Calculate nominal point

[MD, If, Teta, exitflag, Esc, Usc, Isc, Psi] = ...
MD_Calc_UsIsPF(MD, Usnom, Isnom, N.PF, Rs, Xse, N.Speed,’Teta’,pi);
%--- Obtain a TORQUE
M = MD_GetTorque(MD);
%--- Change any parameter you want
P.Stator..... = ....
%--- Regenerate model data
MD = GenerateModelData(P);
%--- Obtain another TORQUE
M = MD_GetTorque(MD);
%--- Etc......
125
126 APPENDIX A. MATLAB SOLVER FUNCTIONS
A.2 Calculation functions

A.2.1 Mode calculation functions
The following functions perform mode calculation based on fixed-rotor model described in
Section 2.5.
MD_CalcEMF – calculates the no-load mode defined by excitation current, stator current,
its angle and speed.
MD_Calc_IfIsPF – calculates the mode defined by excitation current, stator current
and power factor. The values of the stator voltage and load angle are returned.
MD_Calc_UsIsPF – calculates the mode defined by stator voltage, current and power
factor. The values of the excitation current and load angle are returned.
MD_CalcFluxLinkageFixedRotor – calculates phase and complex flux linkages for
the specified excitation current, stator current and its angle. (This function is used in
MD_CalcEMF).
A.2.2 Characteristic calculation functions

These functions return the characteristic in the form of TRMD – TCMResultsMagneticData
structure. It contains one or several calculation points of the magnetic system and has the
following fields:
• Basic fields:
Title – a title of the characteristic or mode point;
Type – the type of the characteristic, which may be one of the following: ’point’,
’noload’, ’load’, ’regulation’, ’vshaped’, ’angular’;
bPerUnit – shows if input data and results are perunit; can be either true or false
(default);
nCharCount – number of sequently calculated characteristics. Normally, it’s 1 but
can be more for a set of characteristics with different constants;
• Mode determining values:

If – excitation current, A;
Isc – complex value of stator current, A(rms) d-q;
Usc – complex value of stator phase voltage, V(rms) d-q;
Esc – complex value of stator phase EMF, V(rms) d-q;
• Magnetic circuit fluxes and potentials:

Um – magnetic branches’ voltages, A;
Im – magnetic branches’ fluxes, Wb;
Vm – magnetic nodes’ potentials, A;
A.3. LOW-LEVEL CALCULATION FUNCTIONS 127
• Calculation results:
Fdelta – airgap flux on a pole pitch, Wb;
Fpole – flux of the pole coming from airgap, Wb;
Bdelta – airgap flux density on a pole pitch, T;
Bys – stator yoke flux density, T;
Bzs(:) – flux density of a stator tooth (array), T;
Bpole(:) – flux density of a pole (array), T;
Byr(:) – flux density in a pole yoke (array), T;
Uagp – airgap ampere-turns, A;
Uys – stator yoke ampere-turns, A;
Uzs – stator teeth ampere-turns, A;
Upole – rotor pole ampere-turns, A;
Uyr – rotor yoke ampere-turns, A;
MMFex – MMF of the excitation winding, A;
Te – electromagnetic torque, N*m;
Xd – synchronous reactance in d-axis, Ohm;
Xq – synchronous reactance in q-axis, Ohm.
MD_CalcLoadChar - calculates the characteristic Vs(Is) for the specified speed and
excitation current.
MD_CalcRegChar - calculates the characteristic If(Is) for the specified speed and stator
voltage.
MD_CalcVChar - calculates the V-shaped characteristic - Is(If) - for underexcited and
overexcited stator currents specified, active power specified and the voltage specified.
A.3 Low-level calculation functions

These functions can be used to set up and calculate a user-specified functioning mode of
the machine.
MD_SetStatorCurrents - sets up the phase currents of the stator for RMS value and
the angle.
MD_SetRotorCurrents - sets up the rotor excitation current.
MD_SetTheta - sets up rotor position angle.
A.4 Post-processing functions

These functions can be used to obtain additional values after calculating one point of the
functioning.
MD_GetAGFluxDensity - calculates the distribution of the normal component of a flux
density in the air-gap.
128 APPENDIX A. MATLAB SOLVER FUNCTIONS
MD_GetPotentials - calculates stator/rotor teeth potentials.

MD_GetPotentialsComplexForm - calculates stator/rotor teeth potentials in a complex
form in plane DQ thus allowing to plot them using COMPASS function.
MD_GetTeethFluxes - returns the teeth fluxes of stator/rotor in the specified layer.
MD_GetTorque - calculates the electromagnetic torque.
Appendix B
2-pole turbogenerator result comparison
This appendix presents the graphical and table representation of the two-pole turbogenerator
calculation results. Table B.1-B.2 and Figures B.1-B.8 show no-load cases. Tables B.3-B.4
and Figures B.9-B.16 give short-circuit results and comparison.
129
130 APPENDIX B. 2-POLE TURBOGENERATOR RESULT COMPARISON
Machine 1 Machine 2
If , A Etcm , V Eexp , V Error If , A Etcm , V Eexp , V Error
1,0 760 60,76 92,00% 37,5 2145 2116,32 1,34%
65,0 3380 3949,18 -16,84% 87,5 4789 4926,31 -2,87%
140,0 7600 8288,01 -9,05% 124,2 6706 6926,52 -3,29%
170,0 9140 9716,43 -6,31% 154,4 8202 8477,21 -3,36%
249,0 11500 11678,85 -1,56% 180,1 9309 9650,87 -3,67%
315,0 12430 12653,01 -1,79% 238,7 11004 11343,77 -3,09%
298,2 12022 12176,83 -1,29%
425,3 13258 13322,04 -0,48%
555,9 13992 14169,80 -1,27%
Machine 3 Machine 4
7,2 1860 569,24 69,40% 1,0 77,5 55,38 28,55%
28,9 3200 2284,87 28,60% 47,4 2746 2624,85 4,41%
36,1 3786 2854,11 24,61% 100,2 5580 5537,07 0,77%
72,3 6160 5714,16 7,24% 140,0 7674 7674,62 -0,01%
108,4 8360 8511,13 -1,81% 163,8 8875 8885,12 -0,11%
144,5 10600 11115,49 -4,86% 185,3 9920 9923,00 -0,03%
180,6 12678 13230,59 -4,36% 212,0 11000 11046,67 -0,42%
224,0 13800 14777,92 -7,09% 244,2 12150 12029,43 0,99%
310,7 15560 16656,80 -7,05% 261,0 12638 12448,56 1,50%
404,6 16600 18193,04 -9,60%
Table B.1: No-load data comparison: machines 1-4.

131
Machine 5 Machine 6
1,0 135 62,63 53,61% 1,0 141 58,46 58,54%
36,8 2325 2304,90 0,86% 35,0 2115 2046,11 3,26%
41,5 2595 2599,28 -0,16% 66,5 3900 3887,19 0,33%
59,5 3660 3726,60 -1,82% 99,5 5718 5786,33 -1,19%
83,0 5055 5192,75 -2,72% 132,0 7455 7574,81 -1,61%
114,5 6900 7103,36 -2,95% 157,5 8700 8857,16 -1,81%
130,0 7773 7996,17 -2,87% 180,0 9615 9795,60 -1,88%
185,5 10395 10401,32 -0,06% 230,0 11000 11086,14 -0,78%
225,0 11500 11514,86 -0,13% 296,3 12060 12082,87 -0,19%
275,3 12450 12530,27 -0,64% 394,5 12922 13178,19 -1,98%
358,5 13317 13486,03 -1,27% 490,0 13530 14022,50 -3,64%
Machine 7 Machine 8
1,0 118 61,03 48,28% 1,0 180 69,92 61,16%
49,5 3027 3020,84 0,20% 32,7 2336 2286,41 2,12%
70,3 4233 4288,75 -1,32% 78,3 5340 5470,54 -2,44%
100,6 6000 6096,91 -1,62% 95,5 6454 6653,73 -3,09%
140,7 8183 8329,65 -1,79% 125,1 8370 8644,76 -3,28%
179,5 9750 10102,90 -3,62% 138,9 9364 9549,76 -1,98%
233,0 11000 11493,53 -4,49% 170,9 11114 11452,15 -3,04%
289,0 11745 12312,42 -4,83% 200,2 12460 12893,70 -3,48%
390,0 12614 13403,34 -6,26% 242,0 13800 14170,33 -2,68%
432,7 12906 13804,36 -6,96% 315,3 15250 15383,07 -0,87%
445,0 16520 16808,55 -1,75%
Table B.2: No-load data comparison: machines 5-8.

Figure B.1: No-load characteristic comparison: Machine 1.

133


135

Machine 1 Machine 2
Is , A If exp , A If tcm , A Error Is , A If exp , A If tcm , A Error
936 140,00 132,42 -5,42% 521 156,20 158,48 1,46%
1620 230,00 229,19 -0,35% 954 288,20 290,20 0,70%
2256 320,00 319,16 -0,26% 1241 376,90 377,51 0,16%
2808 400,00 397,26 -0,69% 1567 476,20 476,67 0,10%
3163 455,00 447,46 -1,66% 1765 536,50 536,91 0,08%
3552 520,00 502,52 -3,36% 1938 590,10 589,53 -0,10%
2102 639,80 639,42 -0,06%
2307 704,30 701,79 -0,36%
Machine 3 Machine 4
364 59,00 76,11 29,00% 675 134,50 132,67 -1,36%
536 99,00 112,07 13,21% 1364 274,50 268,10 -2,33%
762 152,00 159,33 4,83% 1884 380,60 370,31 -2,70%
906 178,00 189,44 6,43% 2146 434,00 421,81 -2,81%
1051 211,00 219,76 4,15% 2418 488,90 475,28 -2,78%
1160 237,00 242,55 2,34% 2677 542,40 526,20 -2,99%
1280 263,00 267,64 1,77% 2948 597,90 579,50 -3,08%
1438 303,00 300,68 -0,76% 3207 651,00 630,45 -3,16%
1575 336,00 329,33 -1,98% 3330 676,20 654,65 -3,19%
1778 382,00 371,78 -2,67%
Table B.3: Short circuit data comparison: machines 1-4.
137
Machine 5 Machine 6
363 55,50 56,85 2,45% 470 62,30 65,55 5,21%
593 92,00 92,88 0,96% 1011 140,80 140,99 0,14%
921 144,50 144,26 -0,16% 1552 217,00 216,44 -0,26%
1298 204,20 203,31 -0,43% 1968 276,00 274,45 -0,56%
1678 264,50 262,83 -0,63% 2462 346,00 343,35 -0,77%
2025 320,00 317,19 -0,88% 2812 395,00 392,16 -0,72%
2374 376,30 371,85 -1,18% 3157 444,80 440,27 -1,02%
2579 410,00 403,97 -1,47% 3499 492,50 487,97 -0,92%
3000 475,90 469,91 -1,26% 3874 547,00 540,27 -1,23%
3260 518,00 510,64 -1,42% 4077 576,00 568,58 -1,29%
Machine 7 Machine 8
493 78,30 79,89 2,03% 1 1,00 0,30 -70,23%
1100 178,50 178,26 -0,14% 377 110,80 112,24 1,30%
1522 247,50 246,64 -0,35% 556 164,50 165,53 0,63%
1920 313,50 311,14 -0,75% 787 234,00 234,21 0,09%
2188 357,00 354,57 -0,68% 1027 305,50 305,75 0,08%
2417 395,00 391,68 -0,84% 1111 331,00 330,76 -0,07%
2629 429,50 426,03 -0,81% 1246 372,50 370,95 -0,42%
2850 467,20 461,85 -1,15% 1364 407,00 406,08 -0,23%
1488 444,00 443,00 -0,23%
1550 463,00 461,45 -0,33%
Table B.4: Short circuit data comparison: machines 5-8.
Figure B.9: Short circuit characteristic comparison: Machine 1.

139


141

Bibliography
[1] Amine Ait-Hammouda. Predimensionnement et etude de sensibilite vibro-acoustique

des machines a courants alternatifs et a vitesse variable. PhD thesis, Ecole Centrale de
Lille, March 2005.
[2] L. Albert. Modelisation et optimisation des alternateurs a griffes. Application au do-

maine automobile. PhD thesis, INPG, Grenoble, 2004.
[3] Joao Pedro A. Bastos and Nelson Sadowski. Electromagnetic Modeling by Finite Ele-
ment Methods. Marcel Dekker, Inc, New York, 2003.
[4] Bekkouche Benaissa. Calcul du champ magnetique par la methode des permeances
de contour de dent. Application au demarrage des moteurs asynchrones. PhD thesis,
Universite des Sciences et de la Technologie d’Oran, Oran, Algerie, May 1995.
[5] Grady Booch. Object-Oriented Analysis and Design. Addison Wesley Longman, second
edition, 1994.
[6] D. Box. Essential COM. Addison Wesley Longman, 1998.
[7] Muzaffer Canay. Equivalent circuits of synchronous machines as well as pre-calculation

of their parameters with examples. PhD thesis, Ecole Polytechnique de l’Universite de
Lausanne, Lausanne, Swiss, 1968.
[8] C.J. Carpenter. Magnetic equivalent circuits. Proc. IEE, (115):1503–1511, October
1968.
[9] Augustin Delale, Laurent Albert, Laurent Gerbaud, and Frederic Wurtz. Automatic
generation of sizing models for the optimization of electromagnetic devices using reluc-
tance networks. IEEE Transactions on Magnetics, 40(2):830–833, March 2004.
[10] C. Delforge. Modelisation d’un actionneur asynchrone et de sa commande vectorielle

par reseaux de permeances. PhD thesis, Universite des science et technique de Lille,
January 1995.
[11] A. Demenko and J. Sykulski. Magneto-electric network models in electromagnetism.

ISEF, September 2005.
[12] V. Devanneaux. Modelisation de la machine asynchrone par reseau de permeances,

validation par le diagnostic. PhD thesis, Universite de Picardie Jules Verne, November
2002.
143
144 BIBLIOGRAPHY
[13] B. Du Peloux de St Romain. Modelisation des actionneurs electromagnetiques par

reseaux de reluctances. PhD thesis, INPG, Grenoble, 2006.
[14] B. Ducharme. Caracterisation et modelisation des circuits magnetiques - du global vers

le local. PhD thesis, Universite de Claude Bernard, Lyon I, 2003.
[15] Jacek F. Gieras and Mitchell Wing. Permanent Magnet Motor Technology. Design and
Application. Marcel Dekker, Inc., New York, second edition, 2002.
[16] R.N. Hasanah, M. Markovic, M. Jufer, and Y. Perriard. Slot-leakage inductance deter-
mination using schwarz-christoffel mapping. Electrimacs-2005, 2005.
[17] Michel Hecquet. Contribution a la modelisation des systemes electrotechniques par la

methode des schemas equivalents magnetiques. Application a l’alternateur automobile.
PhD thesis, Universite des science et technique de Lille, January 1995.
[18] Michel Hecquet. Contribution a l’instrumentation de la demarche de conception des

machines electrique. (HDR). PhD thesis, L’Universite des Sciences et Technologie de
Lille, Decembre 2006.
[19] Michel Hecquet and Pascal Brochet. Modelisation de l’alternateur automobile par un
reseau de permeances couple a des circuits electriques. Journal de Physique, (3):1099–
1116, August 1996.
[20] Michel Hecquet and Pascal Brochet. Time variation forces in a synchronous machine
using electric coupled network model. IEEE Transactions on Magnetics, 34(5):3656–
3659, September 1998.
[21] A. V. Ivanov-Smolensky. Tooth contour method and its application to a non-saturated

electrical machine taking into account double-sided saliency. Electrichestvo, 9:18–28,
1976.
[22] A. V. Ivanov-Smolensky. Electrical Machines. Energy, Moscow, 1980.
[23] A. V. Ivanov-Smolensky, J. V. Abramkin, A. I. Vlasov, and V. A. Kuznetsov. Universal

Method Of Electromagnetic Processes Calculation In Electrical Machines. Energoat-
omizdat, Moscow, 1986.
[24] H. Jerance. Reseaux de reluctances et diagnostic des machines electriques. PhD thesis,
ENSIEG, Grenoble, November 2002.
[25] Bo-nan Jiang. The Least-Squares Finite Element Method. Theory and Applications
in Computational Fluid Dynamics and Electromagnetics. Springer-Verlag, New-York,
1998.
[26] V.A. Khvostov. Computer calculation of the magnetic field in the active zone of a
turbogenerator taking into account saturation. PhD thesis, Moscow Power Engineering
Institute, Moscow, 1977.
BIBLIOGRAPHY 145
[27] V. A. Kuznetsov and Pascal Brochet. A general numerical modeling of electromagnetic

process in electromechanical systems. The International Journal for Copmutation and
Mathematics in Electrical and Electronic Engineering (COMPEL), 22(4):1142–1154,
2003.
[28] Vyacheslav Kuznetsov. General Method of Calculation of Magnetic Fields and Processes
in Electric Machines Having Discrete Space-Distributed Windings. PhD thesis, Moscow
Power Engineering Institute, Moscow, 1990. Technical Science Doctoral Thesis.
[29] Lev D. Landau and Evgeniy M. Lifshitz. The classical Theory of Fields, volume 2 of
Course of Theoretical Physics. Nauka, Moscow, seventh edition, 1988.
[30] M. Latreche. Modelisation magnetique et thermique des machines electriques par des
schemas equivalents. PhD thesis, INP Toulouse, 1987.
[31] M. Liwschitz. Calcul des machines electriques. SPES, II, 1967.
[32] M. Markovic, M. Jufer, and Y. Perriard. Analyzing an electromechanical actuator by

schwarz-christoffel mapping. IEEE Transactions on Magnetics, 40(4), July 2004.
[33] F. Marthouret. Developement de modeles dynamiques comportementaux de circuits

magnetiques. PhD thesis, INSA de Lyon, Lyon, 1995.
[34] K. Nakamura, K. Kimura, and O. Ichinokura. Electromagnetic and motion coupled

analysis for switched reluctance motor based on reluctance network analysis. EPE -
PEMC, 2005.
[35] V. Ostovic. Dynamics Of Saturated Machines. Springer-Verlag, 1989.
[36] V. Ostovic, J.M. Miller, V.K. Garg, R.D. Schult, and S.H. Swales. A magnetic equiva-
lent circuit based on performance computation of a lundell alternator. IEEE Transac-
tions on Industry Applications, 35(4), July/August 1999.
[37] J. Perho. Reluctance network for analysing induction machines. PhD thesis, University
of Helsinki, Helsinku, Finland, December 2002.
[38] Dmitry Petrichenko, Michel Hecquet, Pascal Brochet, Vyacheslav Kuznetsov, and
Daniel Laloy. Numerical modelling of a turboalternator using tooth contours method:
Cad application. ICEM, October 2004. CD-ROM.
Daniel Laloy. Design and simulation of turbo-alternators using coupled permeance
network model. COMPUMAG, June 2005. CD-ROM.
Daniel Laloy. Development of the turbo-alternator simulation software. ELECTRI-
MACS, April 2005. CD-ROM.
Daniel Laloy. Tooth contour method applied to design of turboalternators: devel-
opement of cad software. COMPUMAG, June 2005. CD-ROM.
146 BIBLIOGRAPHY
Daniel Laloy. Design and simulation of turbo-alternators using coupled permeance
network model. IEEETransMag, 42:1259–1262, April 2006.
Daniel Laloy. Permeance network method coupling with experimental design in or-
der to optimize turbogenerators. ICEM, September 2006. CD-ROM.
[44] William H. Press, Saul A. Teukovsky, William T. Vetterling, and Brian P. Flannery.
Numerical Recipes in C. The Art of Scientific Computing. Cambridge University Press,
second edition, 1999.
[45] M. Rakatovao. Un modele operationnel complet pour l’alternateur a griffes dans le

domaine automobile. PhD thesis, INP, Grenoble, 1996.
[46] Carlos Ramirez. Plate-forme numerique d’essais pour machines electriques de puis-
sance. PhD thesis, Ecole Polytechnique Federale de Lausanne, Lausanne, 2003.
[47] Herve Roisse. Contribution a la modelisation des systemes electrotechniques par la

methode des reseaux de permeances couples. Application aux machines synchrones a
aimants permanents. PhD thesis, L’universite des sciences et technologies de Lille,
Lille, January 1998.
[48] H. Roters. Electromagnetic devices. John Wiley and Sons, New-York, USA, 1941.
[49] Filtz R.V. Mathematical basis of the electromechanical converters theory. Naukova
Dumka, Kiev, 1979.
[50] G. Shepherd and B. King. Inside ATL. Microsoft Press, 1999.
[51] G. A. Sipajlov, E. V. Kononenko, and K. A. Khorkov. Electrical Machines (Special

Course). Vysshaya Shkola, Moscow, second edition, 1987.
[52] F. Sixdenier, B. Ducharne, J.P. Masson, L. Morel, and M.A. Raulet. Coupled pro-
gramming: “finite elements” – “dynamical hysteresis flux tubes” for non-homogeneous
circuits modeling. COMPUMAG, July 2003.
[53] G.R. Slemon. An equivalent circuit approach to analysis of synchronous machines

with saliency and saturation. IEEE Transactions on Energy Conversion, (5):538–545,
September 1990.
[54] S. Srairi, J. Farook, A. Djerdir, and Miraoui A. The magnetic network modeling of
a permanent magnet motor with consideration of the saturation and heating effects.
ISEF, September 2005.
[55] Bjarne Straustrup. The C++ Programming Language. Addison Wesley Longman, third
edition, 1997.
[56] J. Turowski. Ch.4, “Reluctance networks” in Computational Magnetics, J. Sykulski.

Chapman and Hall, New York, USA, 1996.
BIBLIOGRAPHY 147
[57] Stephan Vivier, Michel Hecquet, Amine Ait-Hammouda, Mark Goueygou, Pascal Bro-
chet, and A. Randria. Experimental design model applied to a multiphysical model:
treillis designs for a multidimensional screening study. The International Journal for
Copmutation and Mathematics in Electrical and Electronic Engineering (COMPEL),
24(3):15, 2005.
[58] Jiri Vlach and Kishore Singhal. Computer Methods For Circuit Analysis And Design.
Van Nostrand Reinhold Company Inc., New York, 1983.
R-1
Contribution à la modélisation et à la conception optimale des turboalternateurs.

(résumé en français)
Par Dmitry Petrichenko
C.N.R.T. Réseaux et Machines Electriques du Futur (Projet ‘Archimed’).

L2EP, Ecole Centrale de Lille
JEUMONT SA (AREVA NP)
Mots-Clefs : réseau de perméances, méthode de contours de dents, turboalternateurs, CAO des machines.
Keywords : permeance network, tooth contour method, turbogenerator, CAD of electrical machines.
I. Introduction
Contexte :
Dans les complexes industriels, les cahiers des charges sont de plus en plus exigeants en terme de
performances et d’optimisation économique des installations de production électrique. Dans certains cas, on
cherchera à minimiser les courants de court-circuit pour soulager les organes de coupure ; au contraire dans
d’autres cas, on veillera à limiter les chutes de tension transitoires. Les moyens de calcul rapides permettant
de dimensionner les machines, basés sur des modèles analytiques n’ont pas toujours la précision nécessaire.
Les méthodes de calcul par Eléments Finis permettent d’optimiser des parties locales de la machine, mais
deviennent trop lourds d’utilisation s’agissant de dimensionner des équipements pour répondre aux appels
d’offre de matériel de moyenne puissance.
La méthode proposée par L2EP de Lille et le MPEI de Moscou, basée sur l’emploi des réseaux de
perméances, a retenu l’attention de Jeumont SA pour développer un nouveau logiciel dans le cadre du
CNRT « Réseaux et Machines du Futur » de Lille. Cette méthode permet de conserver la rapidité
d’exploitation et d’améliorer la précision de calcul en adaptant facilement le réseau de perméances à la
configuration étudiée.
Le travail développé est ouvert aux méthodes d’optimisation ainsi qu’à la prise en compte de régimes
dynamiques.
Objectif :
L’objectif de ce travail concerne le développement d’un outil complètement ‘automatisé’ permettant
dans un premier temps la modélisation des turbo-alternateurs.
Notre idée est de se reposer sur un réseau de perméances paramétrées établi à partir de calculs
analytiques et en ce qui concerne l’entrefer, l’utilisation de la Tooth Contour Method (TCM) [1], [2] ou la
méthode de contour de dents. Le couplage électrique, magnétique ainsi que mécanique, la saturation, le
mouvement et les effets tridimensionnels sont pris en considération.
L’objectif à long terme est de mettre au point d’un outil de conception optimale électromagnétique de
turbo-alternateurs de 10 à 100 MW, d’un modèle numérique d’une part, d’un progiciel sur PC d’autre part.
Cet outil, développé sous Matlab, nous fournit très rapidement le circuit magnétique, en quelques
secondes, et la possibilité d’effectuer des calculs :
- en statique, exemple : calcul des grandeurs comme la tension et le courant débité pour un point
de fonctionnement nominal,
- ou en dynamique, exemple : détermination de l’allure temporelle de la tension de sortie.
R-2
La réalisation de l’outil est présentée, incluant la méthode de calcul du circuit, développée en C++, et le
solveur développé sous Matlab. Celui-ci génère automatiquement le réseau de perméances paramétré.
Difficultés liés à l’étude des turbo-alternateurs :
Le premier point a été la prise en compte du nombre de paramètres géométriques et électriques qui
varient entre 250 et 300 en fonction de la finesse retenue pour l’étude. Sur la figure 1, une partie du rotor est
présentée où l’on doit intégrer les différentes formes d’encoches, mais aussi la possibilité de relever le flux
et l’induction en de nombreux points.
La deuxième difficulté est le nombre de phénomènes à prendre en compte. Les principaux concernent:
- La prise en compte des évents (canaux de ventilations),
- les têtes de bobines et des écrans magnétiques,
- l’association des différentes parties mécaniques (joints…)
- le bobinage (prise en compte des isolants et des cales entre plans)
- la présence ou non de cales d’encoche magnétique ou semi-magnétique.
Ainsi, un réseau de perméances adapté à ces paramètres a été défini. Ce réseau devait etre déduit
automatiquement. La démarche de constitution du réseau de perméances est rappelée ci-dessous.
Flux dans le pτ le
Flux de fuite
Flux dans l' armature.
Fig. 1 : ¼ du rotor d’un turbo-alternateur
II. Définition du réseau de perméances
Choix de la méthode :
Différentes méthodes existent pour la modélisation des machines électriques et le calcul de champ en
commençant par la méthode des éléments finis : précise mais pour des modélisations 3D avec prise en
compte du mouvement sur ce genre de machine les temps de calculs deviennent très pénalisants, supérieur à
plusieurs heures.
Notre objectif à long terme étant de développer un outil de CAO des turbo-alternateurs, il est
préférable de privilégier un outil avec des temps de calculs réduits (de l’ordre de la minute).
Ainsi, une autre approche que l’on peut classer entre les méthodes numériques et analytiques concerne
la Permeance Networks Method (PNM) ou la méthode des réseaux de perméances. L’avantage de la PNM
R-3
est de réaliser un bon compromis entre le temps de calcul et la précision. Différentes techniques de calculs
du réseau existent, surtout en ce qui concerne les perméances d’entrefer :
- celles utilisant des méthodes analytiques uniquement, par exemple [3], [4],
- celles reposants sur des calculs E.F. en statique et permettant d’avoir une grande précision sur le
réseau de perméances obtenu, surtout en ce qui concerne des machines complexes et 3D [5], [6],
[7].
- et celles utilisant la T.C.M. (Tooth Contour Method).
La T.C.M. ou méthode de contour de dents est basée sur une représentation théorique du champ dans
l’entrefer à partir de la somme des champs magnétiques obtenus par des éléments spéciaux que l’on appelle
« contour de dent ». Sur l’exemple donné Fig. 2, le ‘contour de dent’ stator concerne uniquement une dent
stator entourée par des conducteurs disposés dans les encoches adjacentes. Au rotor, la discrétisation ou
‘contour de dent’ dépend de la géométrie, celle-ci imposant un réseau de perméances plus ou moins
complexe.
La détermination du champ autour d’une dent est obtenue en utilisant des conditions limites particulières,
en supposant que la perméabilité du fer stator et rotor est nettement supérieure à celle de l’air. Uniquement
trois à quatre dents sont choisis pour la détermination des perméances en plaçant nos conditions limites
directement sur les bords de la dent (Fig.2). Ainsi, la détermination est très rapide et ne nécessite pas la
machine complète.
L’allure présentée λsr(bsr) de la perméance mutuelle entre le contour de dent stator s et le contour non
alimentée du rotor r (potentiel scalaire égal à 0) est obtenue comme une fonction de la position relative
λsr(bsr) [8]. Afin de valider cette allure, des calculs sur OPERA-2D (code éléments finis) [9] en statique ont
été nécessaires [10].
λ(b)
Fig. 2 : Représentation du contour de dent (1 seule dent ‘alimentée’)
La saturation est ensuite insérée au niveau de chaque perméance dans notre réseau et la méthode de
Newton-Raphson permet de déterminer le flux ou le potentiel magnétique dans le circuit magnétique.
Présentation du réseau de perméances :

Le réseau de perméances du système magnétique est généré automatiquement à partir des paramètres
géométriques. La structure générale du stator et du rotor est présentée sur la figure suivante (Fig. 3).
R-4
a) Au stator b) Au rotor
Fig. 3: Partie du réseau de perméances
Le nombre de subdivisions de la culasse statorique et des dents est variable. Cette variation sur la densité
du réseau est possible au stator et au rotor. Elle permet d’intégrer plus précisément le problème des
saturations locales. Cette décomposition a une influence, bien évidemment, sur le temps de calcul.
Le rotor d’un turbo-alternateur (Fig. 1) peut être très sophistiqué avec différentes formes d’encoches et
de dents avec des profondeurs variables. Ces différentes formes sont prises en considération par une analyse
fine des zones afin de trouver le meilleur découpage.
La Fig. 3.b montre le circuit magnétique équivalent de dents rotorique. L’utilisateur définit uniquement
le nombre de subdivisions. La compléxité de géometrie est traité par l’outil.
Couplage électrique – magnétique :

Les sources de force magnétomotrice (MMF) sont placées au niveau des branches horizontales : Fig. 4
(non représentées sur les deux figures précédentes). Elles permettent le couplage électrique-magnétique.
MMF
source 1
MMF
source 2
I
II MMF
source 3
MMF
source 4
Fig. 4: Source de force magnétomotrice
Pour exemple, la source MMF1 est déterminée à partir de la ‘région’ de bobinage I, puis la source
MMF2 à partir de la région de bobinage I & II, etc.…
Des détails du circuit complet avec le couplage et la description du système d’équations sont donnés dans
la référence [10].
Notre réseau est construit sur une période électrique uniquement et dans ce modèle équivalent, les
courants induits au rotor ne seront pas considérés.
R-5
Prise en compte des effets 3D :
Dans le cas de l’étude des turbo-alternateurs, la prise en compte de la 3ème dimension est nécessaire en
vue d’intégrer les phénomènes suivants comme les canaux de ventilations ou les effets d’extrémités. En
effet, l’influence des canaux de ventilations situés au stator et/ou au rotor est à prendre en considération dans
le réseau de perméances.
Pour cela, un calcul de la longueur réelle du fer et l’utilisation de relations analytiques pour la prise en
compte des effets d’extrémités sont placés directement dans le calcul des perméances. Ce point est détaillé
dans la référence [11].
III. Réalisation de l’outil
Une présentation des différentes parties de notre outil, appelé « TURBO-TCM » est donnée sur la Fig. 5.
En ce qui concerne les données d’entrée, la fiche électrique et magnétique du turbo-alternateur est nécessaire
et doit contenir : les dimensions stator et rotor, le nombre d’encoches stator et rotor, leurs dimensions, la
description complète du bobinage stator et rotor, les caractéristiques des matériaux, les spécifications du
circuit électriques, etc.
A partir de toutes ces informations, un circuit magnétique équivalent paramétré est défini, mais aussi le
circuit électrique et les équations de couplage. La définition de ce réseau couplé peut être réalisée
automatiquement ou manuellement (par exemple, durant la phase de test de l’outil).
Electric circuit Turboalternator parameters

parameters
Electric circuit Winding Magnetic circuit

description description description
Electric part equations Magnetic part

equations
Coupling equations
SOLVER
Calculation results
Fig. 5 : Structure de l’outil.
Ainsi, le système d’équations matricielles [10] peut être traité. Ces différentes ‘parties’ correspondent à
un modèle objet défini sous Visual Studio C++ .NET et utilisant la technologie ‘Component Object Model’
(COM). Celles-ci sont ensuite compilées comme une librairie ‘type library : TCMLib’.
L’avantage de cette démarche est de pouvoir générer et manager en temps réel le réseau paramétré tel
que les éléments ‘caractéristiques électrique ou magnétique’, les paramètres du circuit, etc. Ces composantes
‘TCM-Lib’ peuvent être traitées par n’importe quel solveur. Dans notre cas, nous utilisons Matlab.
En ce qui concerne l’implantation sous Matlab, une interface graphique (Graphic User Interface) a été
réalisée et sur la figure suivante, nous présentons une partie de l’interface utilisée pour les ‘Données
d’entrée’ :
R-6
(pour des raisons de confidentialité, les données ont été supprimées)

(exemple : encoche statorique)
Fig. 6 : Interface graphique : données d’entrée.
Différents algorithmes existent permettant de calculer les régimes en statique et dynamique de la

machine. L’effet de saturation est réalisé par un algorithme bien connu : Newton-Raphson.
IV. Application à une machine synchrone 2 pôles
Une première machine disposant d’un rotor identique à celui d’un turbo-alternateur a été conçue en vue
de valider notre modèle. La machine a une puissance de 3kW avec un rotor modifié. Celui-ci a la
configuration suivante :
Fig. 7 : rotor 2 pôles (prototype de faible puissance)
La machine a un entrefer d’1mm et est composée de 24 encoches au stator et 14 au rotor avec une
distribution non-régulière. Les caractéristiques b(h) sont différentes au stator et au rotor.
Nos simulations sont comparés aux résultats obtenus par un logiciel éléments finis, vue de la ½ machine
Fig. 8 en utilisant OPERA-2D-RM : rotating machine, afin de valider le flux traversant une dent par
exemple, mais aussi aux résultats expérimentaux.
R-7
Fig. 8 : ½ machine 2 pôles (prototype de faible puissance)
Une première comparaison entre les simulations OPERA et le TURBO-TCM des flux au niveau de la
carcasse et dans une dent en statique à vide et en charge a été réalisée. La Fig. 9 compare les flux au niveau
d’une dent statorique alignée avec l’axe du pole rotorique, et au niveau de la culasse où le flux est maximal.
5,0E-03
4,5E-03
Yoke fluxes
4,0E-03
3,5E-03 OPERA 2D - yoke

Turbo-TCM - yoke
3,0E-03
OPERA 2D - tooth
Flux, Wb
2,5E-03 TurboTCM - tooth
2,0E-03
Tooth fluxes
1,5E-03
1,0E-03
5,0E-04
0,0E+00
0,0 2,0 4,0 6,0 8,0 10,0 12,0
Excitation current, A
Fig. 9 : Comparaisons Turbo-TCM et OPERA : flux de dent et de culasse.
L’écart est relativement faible, l’erreur maximale sur le flux de dent est 4.9% et de 2.6% pour le flux de
culasse. La figure suivante présente la caractéristique à vide E (la force électromotrice) fonction du courant
d’excitation j. L’erreur maximale est de 6.6% obtenu pour le courant j maximal (20A).
Le FEM est obtenu par la méthode de rotor fixé. Cette méthode est utilisée pour obtenir tous les
caractéristique de la machine en statique.
R-8
No-load characteristic
120
100
OPERA
Eeff , V 80 TurboTCM
60
40
20
0
0 2 4 6 8 10 12 14 16 18 20
J, A
Fig. 10 : Caractéristique à vide E(j) de la machine synchrone
A présent, la caractéristique en charge est présentée.

La comparaison en charge est donnée sur la Fig. 11 présentant la caractéristique Us, tension statorique
entre 2 phases, fonction de Is, courant dans une phase statorique. L’erreur maximale est de 4.1% pour un
courant stator de 13A.
Les simulations obtenues sont très satisfaisantes et présentent un bon compromis temps de calcul –
précision. En effet, le temps de calcul pour obtenir la caractéristique E(j) ou Us(Is) est de 10-15 secs.
Load characteristic
100
95
90
TurboTCM
85
Us, V
80
75
70 OPERA
65
0 2 4 6 8 10 12 14
Is, A
Fig.11 : Caractéristique en charge Us (Is)
La comparaison en ‘dynamique’ : avec prise en compte du mouvement est donnée sur la Fig. 12. Le
mouvement demande plusieurs points de simulations. Avec TURBO-TCM, 180 positions sont considérées
pour une période. Avec OPERA-RM, seulement 100 positions sont données, le nombre de points de sortie
R-9
étant limité pour un calcul. La figure présente la fem à vide E (Volt) en fonction du temps (sec.) pour un
courant d’excitation de 12 A (machine saturée).
250
OPERA
200 TurboTCM
150
100
50
Ea, V
-50
-100 150
OPERA
TurboTCM
-150
100
-200
-250
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
50 Time, sec
0
-5 0 5 10 15 20 25
Fig. 12 : Résultats de simulations avec le mouvement.
La Table 1 présente le temps de calcul entre OPERA and TurboTCM sachant que le calcul par réseau de
perméances compte 180 positions au lieu de 100 pour OPERA.
Table 1 : Temps de calcul
Rapport sur
J, A TurboTCM Opera le temps de
calcul
1 12,4 sec 0h 55min 266,13
10 18,1 sec 3h 06min 616,57
20 18,3 sec 3h 25min 672,13
Ci-dessous, quelques résultats expérimentaux viennent renforcer la validité de notre modèle ‘réseau de
perméances’. La comparaison de la courbe à vide calculée et expérimentale est présentée sur la Fig. 13, à
gauche. Le résultat pour caractéristique de réglage est presenté sur la Fig. 13, à droite.
Fig. 13 : Comparaison de résultats de simulation et expérimentaux.

R-10
IV. Application au turbo-alternateur 2 pôles
Les turbo-alternateurs de la gamme industrielle (10 à 100 MW) sont entrainés par des turbines à gaz ou à
vapeur tournant aux vitesses de synchronisme, 3000 ou 3600 tr/min. Le refroidissement par air est direct
pour le bobinage rotorique dont les conducteurs sont ajourés pour laisser l’air se propulser depuis un canal
sous encoche vers l’entrefer. Le stator est en refroidissement indirect, l’air circulant au travers des évents
ménagés entre les paquets de tôles du circuit magnétique (figure ci-dessous).
Les bobinages rotoriques sont concentriques, logés dans des encoches taillées dans l’arbre massif
magnétique. Les têtes de bobines sont maintenues par des frettes en acier amagnétique.
Le bobinage statorique triphasé est réalisé en barres (2 par encoches) connectées entre elles de façon à
réaliser les circuits souhaités pour obtenir la tension de sortie nominale avec un circuit magnétique
idéalement magnétisé et pour minimiser les harmoniques de tension.
Les barres isolées de la masse par un mur isolant micacé sont constituées de brins de cuivre isolés les uns
des autres, de dimensions optimisées et disposés en compensation Roebel pour minimiser les pertes Joules
supplémentaires de type field et de circulation.
Les développantes et les circuits de couplage sont fermement supportés et calés pour éviter les vibrations
qui endommagent les isolants.
Le circuit magnétique est réalisé en tôles à très faibles pertes, isolées et posées à joints décalés. Des
tirants et les doigts d’extrémité permettent d’obtenir un serrage de la tôlerie magnétique entre deux plateaux
de forte épaisseur pour éviter les vibrations et le bruit associé. Les plateaux en acier magnétique sont
protégés du champ magnétique développé par les courants dans les têtes de bobines par des écrans de flux
en cuivre .
Les caractéristiques dimensionnelles et les propriétés physiques de ces éléments constructifs constituent
les données d’entrée du code de calcul développé.
Des extensions du code permettront également de dimensionner des machines à rotor cylindrique à
polarité multiple (4 pôles, N pôles).
Carcasse et stator d’un alternateur 4 pôles

Rotors bobinés d’un alternateur 4 pôles
(en premier plan) et d’un turboalternateur
2 pôles (en arrière plan)
Caractéristiques à vide :
Quelques résultats obtenus en régime permanent sont présentés ci-dessous. La méthode utilisée est
détaillée dans les références [11], [12].
La Fig. 14 présente la tension à vide obtenue par Turbo-TCM et obtenue expérimentalement (mesures
réalisées par Jeumont SA.). L’influence des effets 3D est donnée. L’erreur maximale est de 6,8% et de 0,3%
pour le courant nominal.
La difficulté dans ce cas, est la prise en compte de la courbe b(h) « réelle » ayant une saturation
importante au rotor mais aussi l’effet de magnétisation que l’on retrouve sur la caractéristique à vide réelle.
R-11
La Fig. 15 présente la caractéristique obtenue en court-circuit représentant le courant rotorique (If)

fonction du courant d’induit ou statorique (Is). Les résultats de simulation donnent de très bons résultats,
erreur < à 3.5%. En effet, dans ce cas, il n’y pas d’effet de saturation.
Courant rotorique fonction du courant statorique Is
0.85
0.77
Expérimental
0.69
Tension de phase par rapport au courant rotorique
TurboTCM
1.217 0.61
If, A
1.043 (p.u.)
Experiment
0.87
0.38
0.3
Us, V TurboTCM without 3D effects
p.u. 0.22
0.15
TurboTCM with 3D effects 500 1000 1500 2000 2500 3000 3500 4000
0.348 Is, A
2
0.174
Error, %
0
0
0 50 100 150 200 250 300 350
If, A
-2
-4
936 1620 2256 2808 3163 3552
Is, A
Fig. 14 : caractéristique à vide Fig. 15 : caractéristique en court-circuit.
L’analyse en dynamique (prise en compte du mouvement) est détaillée dans la référence [12]. L’intérêt
est de pouvoir effectuer une analyse harmonique afin de vérifier par exemple que la tension de sortie de
l’alternateur est parfaitement sinusoïdale ou de supprimer un harmonique en particulier au niveau de la fft de
l’induction dans l’entrefer (Fig. 16).
No-load case
1
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
-4 -3 -2 -1 0 1 2 3 4
Angle, rad
Fig. 16: Allure de l’induction dans l’entrefer (à vide)
Le temps de calcul pour une période est de l’ordre de la minute sur un pentium IV- 300MHz.
R-12
Caractéristiques en charge :
Les figures suivantes (17 et 18) présentent les caractéristiques en charge d’un turbo-alternateur du marché
industriel. Sur la Fig. 17, la tension de sortie Us est fournie fonction du courant débité Is pour 3 valeurs de
cos ϕ (PF). Les courbes en V sont données pour une tension de sortie nominale et pour différentes valeurs
de puissance (Ps).
Fig. 17: Courbes U(I) pour différentes valeurs de PF Fig. 18. Courbes en V pour différentes valeurs de puissance Ps
Comme sur la Fig. 16, l’induction dans l’entrefer dans le cas d’un fonctionnement en charge est présentée sur
la Fig. 19.
Nominal mode
1.5
0.5
Bairgap, T
-0.5
-1
-1.5
-4 -3 -2 -1 0 1 2 3 4
Angle, rad
Fig. 19. Induction dans l’entrefer fonction de l’angle.
De plus, il est possible de determiner le couple d’origine électromagnétique par l’outil ‘TurboTCM’:
1 ∂Λ
TTCM = ⋅ ∑U T ⋅ ⋅U
2 ∂θ
Avec U – difference de potential magnétique, Λ - perméances, θ - angle du rotor
Il est possible de comparer avec le calcul “classique” pour le point nominal (en négligeant le couple de
pertes):
S ⋅ PF
TEM =
Ω
Avec S – la puissance apparente, PF – facteur de puissance, Ω - vitesse de rotation en rad/sec
R-13
La différence entre ces deux calculs pour le point nominal est quasiment négligeable :
TTCM = 1.006 TEM
V. Etude de sensibilité
Un autre avantage de ‘TurboTCM’, ayant un bon compromis entre le temps de calcul et la précision,
il est possible d’étudier les variations de grandeurs d’entrée comme le diamètre, la longueur, etc…, fonction
des caractéristiques de sortie.
Pour exemple, le couple électromagnétique produit par le turbo ainsi que la variation du courant
rotorique sont considérés comme réponse. Seulement deux paramètres sont considérés :
- la hauteur de la culasse liée dans notre cas au diamètre (dia),
- la longueur de la machine (L).
Le domaine d’étude est défini par des intervalles de variations (Table 2).
TABLE 2: INTERVALLES DE VARIATION

Facteurs Niveau bas Niveau haut
dia dia – 5% dia + 5%
L L – 5% L + 5%
Les figures suivantes (20 et 21) donnent l’influence de ces facteurs fonction des 2 réponses. Le
nombre de points de simulations considérées pour représenter est de 5*5, soit 25 simulations. Le temps de
calcul est de l’ordre de 5 minutes pour obtenir ces surfaces de réponses.
Courant rotorique (A) en p.u.

Couple (Nm) en p.u.
1.12
1.01
1.1
1.0095
1.08
1.009 1.06
If
1.04
C
1.0085
1.02
1.008
1
1.0075 1.8
1.85
1.007
1.9 2.6
2.6 2 1.95 2.55
2.55 1.95 2.5
2.5 1.9
1.85 2 2.45
2.45
1.8 dia L
L dia
Fig. 20. Couple électromagnétique en p.u. Fig. 21. Courant rotorique fonction en p.u.
Ces simulations sont réalisées dans des conditions nominales. Il est ensuite possible de choisir très
rapidement en fonction du cahier des charges la hauteur de la culasse et la longueur du fer la plus
intéressante par rapport au cahier des charges.
R-14
VI. Conclusion & perspectives
Dans ce travail, un modèle de turbo-alternateurs a été présenté, basé sur la méthode de contour de
dents. L’outil ‘TurboTCM’ développé sous Matlab permet de générer de façon automatique le réseau de
perméances. Des simulations ont été comparées aux résultats expérimentaux et ont montré la bonne
correspondance des résultats. Puis, certaines caractéristiques de sortie sont données montrant les possibilités
de l’outil.
En réalisant un bon compromis entre le temps de calcul et la précision, nous pouvons envisager le
couplage avec des outils d’optimisation. Dans un premier temps, une étude de sensibilité est fournie sachant
que celle-ci peut être étendue à n facteurs influents.
Les perspectives sont nombreuses sur l’application d’un tel outil. Il nous reste à valider celui-ci sur
‘n’ turbo-alternateurs de puissances différentes mais aussi avec un nombre de pôles différents. Un de nos
objectifs à court terme est de coupler l’outil avec la méthode des plans d’expériences [13], [14], permettant
de considérer un grand nombre de facteurs, une première étape de ‘Screening ou criblage’ permettant de
déduire les facteurs influents.
R-15
REFERENCES
[1] Ivanov-Smolenskii A., Kuznetsov V., “Universal Numerical Method for Simulating Electromechanical
Converters and Systems”, Elektrichestvo, Vol. 2000, No. 7, pp 24-34, July 2000, Moscow, Russia.
[2] Kuznetsov V., Brochet P., “A general numerical modeling of electromagnetic process in electromechanical
systems”, International Conference Electrical Machines (ICEM 2002), August 2002, Bruges, Belgium.
[3] Ostovic V., “Dynamics of saturated machines”, Springer-Verlag, 1989.
[4] Worotynski J., Turowski M., and Mendrela E.A., "The accuracy of calculation of magnetics fields, inductance
and forces in electromagnetic devices using the reluctance network method ", Proc.ISEF’93, Warsaw, pp159-162,
1993.
[5] M. Hecquet, " Contribution à la modélisation des systèmes électrotechniques par la méthode des schémas
équivalents magnétiques. Application à l'alternateur automobile", Thèse, Université des Sciences et Techniques
de Lille, Janvier 1995.
[6] M. Hecquet, P. Brochet, “Modélisation d’un alternateur automobile par un réseau de perméances couplé à des
circuits électriques ”, Journal of applied Physics. III, August 96, pp 1099-1116.
[7] Hecquet M., Brochet P., “Time variation forces in a synchronous machine using electric coupled network model”,
IEEE Trans. On Magnetics, Vol. 34, No 5, pp 3656-3659, September 1998.
[8] Ivanov-Smolenskii A. and authors, “Universal Method of Electromagnetic Processes Calculations in Electrical
Machines”, Energoatomizdat, 1986, Moscow, Russia.
[9] Référence Manual of Vectors Fields : OPERA-2D RM; www.vectorfields.com;
[10] D. Petrichenko, M. Hecquet, P. Brochet, V. Kuznetsov, D. Laloy, ‘Numerical modelling of a turboalternator using
Tooth Contours Method: CAD application’, ICEM 2004, CD-Rom, Cracovie, 5-7 Oct. 2004.
[11] D. Petrichenko, M. Hecquet, P. Brochet, V. Kuznetsov, D. Laloy, ‘Design and simulation of turbo-alternators
using coupled permeance network model’, IEEE Trans on Magnetics, Vol. 42, pp. 1259-1262, April 2006.
[12] D. Petrichenko, M. Hecquet, P.Brochet, Vyacheslav Kuznetsov (MPEI), D. Laloy (Jeumont-Framatome),
‘Design and simulation of turboalternators using Tooth contours method applied to design of turboalternators :
developpement of CAD Software, ISEF'2005, Baiona, Espagne, 15-17Septembre 2005.
[13] S. Vivier, “Stratégies d'optimisation par plans d'expériences et Application aux dispositifs électrotechniques
modélisés par éléments finis”, Thèse de doctorat, Université des Sciences et Techniques de Lille, Juillet 2002.
[14] S. Vivier, M. Hecquet, A. Ait-Hammouda, M. Goueygou, P. Brochet, A. Randria, ‘Experimental design method
applied to a multiphysical model : treillis designs for a multidimensional screening study’, ”, The International
Journal for Computation and Mathematics in Electrical and Electronic Engineering (COMPEL), Special Issue,
Vol. 24, No. 3, 15p, 2005.

These DPetrichenko VF PDF

Uploaded by

Copyright:

Available Formats

These DPetrichenko VF PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

These DPetrichenko VF PDF

Uploaded by

Copyright:

Available Formats

Calculation and Simulation of Turbogenerators Using

Permeance Network. Optimization Application.

To cite this version:

HAL Id: tel-00346766

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

ÉCOLE CENTRALE DE LILLE

Doctorat délivré par l’École Centrale de Lille.

Contribution à la modélisation et à la conception

Soutenu le 10 juillet 2007 devant le jury d’examen :

Rapporteurs : L.GERBAUD Professeur au LEG (Grenoble)

Thèse préparée dans le laboratoire, L2EP, à l’École Centrale de Lille.

To my wife’s mother Elena

1 Problem definition and overview 5

Conclusion and perspectives 121

A Matlab Solver functions 125

B 2-pole turbogenerator result comparison 129

1 4-poles rotor and stator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1 Synchronous model equivalent circuit . . . . . . . . . . . . . . . . . . . . . . 8

2.1 A part of magnetic system’s equivalent circuit in terms of SMP . . . . . . . 27

3.1 General structure of TurboTCM . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4.1 Rotor dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

B.1 No-load characteristic comparison: Machine 1. . . . . . . . . . . . . . . . . . 132

The development of new methods of investigation of electromechanical systems, consisting

Figure 1: 4-poles rotor and stator

Figure 2: 2-poles rotor of turbogenerator

Problem definition and overview

• Regulation, V-shape and Power-load angle characteristics.

• P-Q diagrams for different restrictions.

• Xd, Xq parameters for every calculated mode.

A modern turbogenerator is a complex, high-technological product which is described

• complex geometry of the turbogenerator’s rotor, its tooth zone;

1.1 Existing methods

where Φi – magnetic flux; Fi – magneto-motive force; Λi – magnetic permeance of the i-th

1.1.1 Conventional methods of calculations

• magnetic circuit saturation is neglected;

• hysteresis and eddy currents effects are neglected;

• electromotive forces (EMF) are supposed to be proportional to the excitation current;

• the machine is supposed to have non-salient poles.

Figure 1.1: Synchronous model equivalent circuit

This model allows using the following equation in a complex form:

Ė1 − R1 I˙1 − jx1 I˙1 − V˙1 = 0 (1.5)

R1 – active resistance of a stator phase;

x1 = xa + x1σ – full synchronous reactance of a stator phase, where Xa – stator winding

Ė1 – EMF of a stator phase;

I˙1 – stator current;

Figure 1.2: Synchronous model phasor diagram

x1σ = xσslot + xσdif f + xσend (1.6)

Figure 1.3: Equivalent circuit for Pottier diagram

Figure 1.4: Potier diagram

Figure 1.5: Frame conversion diagram

If the stator winding of a synchronous machine is connected in star with neutral or

Ψf , if – Rotor excitation winding flux linkage and current respectively.

Ld – An inductance of a stator phase with symmetric sinusoidal current for a synchronous

Lq – An inductance of a stator phase with symmetric sinusoidal current for a synchronous

L0 – An inductance of a stator phase with zero-sequence current. The corresponding

The voltages can be described by the following system of equations:

Figure 1.6: Phasor diagram of a synchronous generator working on an active-inductive load

The appropriate precision of the electromagnetic calculations is achieved using special

A lot of parameters and characteristics of electrical machines, which are significantly