List of Papers
This thesis is based on the following papers, which are referred to in the text by their Roman numerals.
I Ranlöf, M., Perers R. and Lundin U., “On Permeance Modeling of Large Hydrogenerators With Application to Voltage Harmonics Predic- tion”, IEEE Trans. on Energy Conversion, vol. 25, pp. 1179-1186, Dec. 2010. II Ranlöf, M. and Lundin U., “The Rotating Field Method Applied to Damper Loss Calculation in Large Hydrogenerators”, Proceedings of the XIX Int. Conf. on Electrical Machines (ICEM 2010), Rome, Italy, 6-8 Sept. 2010. III Wallin M., Ranlöf, M. and Lundin U., “Reduction of unbalanced mag- netic pull in synchronous machines due to parallel circuits”, submitted to IEEE Trans. on Magnetics, March 2011. IV Ranlöf, M., Wolfbrandt, A., Lidenholm, J. and Lundin U., “Core Loss Prediction in Large Hydropower Generators: Influence of Rotational Fields”, IEEE Trans. on Magnetics, vol. 45, pp. 3200-3206, Aug. 2009. V Ranlöf, M. and Lundin U., “Form Factors and Harmonic Imprint of Salient Pole Shoes in Large Synchronous Machines”, accepted for pub- lication in Electric Power Components and Systems, Dec. 2010. VI Ranlöf, M. and Lundin U., “Finite Element Analysis of a Permanent Magnet Machine with Two Contra-rotating Rotors”, Electric Power Components and Systems, vol. 37, pp. 1334-1347, Dec. 2009. VII Ranlöf, M. and Lundin U., “Use of a Finite Element Model for the Determination of Damping and Synchronizing Torques of Hydroelec- tric Generators”, submitted to The Int. Journal of Electrical Power and Energy Systems, May 2010. VIII Ranlöf, M., Wallin M. , Bladh J. and Lundin U., “Experimental Study of the Effect of Damper Windings on Synchronous Generator Hunting”, submitted to Electric Power Components and Systems, February 2011. IX Lidenholm J., Ranlöf, M. and Lundin U., “Comparison of field and circuit generator models in single machine infinite bus system simula- tions”, Proceedings of the XIX Int. Conf. on Electrical Machines (ICEM 2010), Rome, Italy, 6-8 Sept. 2010.
v X Wallin M., Ranlöf, M. and Lundin U., “Design and construction of a synchronous generator test setup”, Proceedings of the XIX Int. Conf. on Electrical Machines (ICEM 2010), Rome, Italy, 6-8 Sept. 2010.
Reprints were made with permission from the publishers.
vi Contents
1 Introduction ...... 1 1.1 Background ...... 1 1.2 Applications of Permeance Models of Salient-pole Generators . 2 1.3 Core Loss Prediction in Large Hydropower Generators ...... 3 1.4 Form Factors of Salient Pole Shoes ...... 3 1.5 Analysis of a PM Generator with Two Contra-rotating Rotors . 4 1.6 Electromechanical Transients - Simulation and Experiments . . 4 1.7 Outline of the Thesis ...... 5 2 Theory ...... 7 2.1 Salient-pole Synchronous Generators ...... 7 2.1.1 Main Construction Elements ...... 7 2.1.2 Grid-connected Operation ...... 9 2.2 Equivalent Circuit Generator Model ...... 10 2.2.1 P.U. Electrical Equations ...... 11 2.3 Finite Element Generator Model ...... 13 2.3.1 Calculation Geometry and Material Property Assignment . 13 2.3.2 Field Equation Formulation ...... 14 2.3.3 Finite Element Discretization ...... 16 2.3.4 Boundary Conditions ...... 17 2.3.5 Calculation of Air-gap Torque and Induced EMF ...... 18 2.4 Coupled Field-circuit Models ...... 19 2.4.1 Coupling Equations for Circuit-connected Conductors . . . 19 2.4.2 Rated Voltage No-load Operation Model ...... 20 2.4.3 Balanced and Unbalanced Load Models ...... 23 2.4.4 Grid-connected FE Model with Mechanical Equation . . . . 25 3 Applications of Permeance Models of Salient-pole Generators . . . . 27 3.1 Previous Work ...... 27 3.2 Permeance Model Implementation ...... 28 3.2.1 Coordinate System ...... 28 3.2.2 Field and Armature MMF Functions ...... 29 3.2.3 Pole Shape Permeance Function ...... 31 3.2.4 Saturation and Stator Slot Permeance Functions ...... 31 3.3 Damper Winding MMF and Circuit Equations ...... 33 3.3.1 Flux Density Harmonics ...... 34 3.3.2 Unitary Damper Loop MMF Functions ...... 36 3.3.3 Calculation of Damper Loop Currents ...... 37
vii 3.3.4 Resultant Damper MMF ...... 40 3.4 Selected Results ...... 41 3.4.1 THD of the Open-circuit Armature Voltage Waveform . . . 41 3.4.2 Damper Bar Currents at Rated Load Operation ...... 42 3.4.3 Reduction of the UMP by Parallel Armature Circuits . . . . 43 4 Core Loss Prediction in Large Hydroelectric Generators ...... 45 4.1 Previous Work ...... 45 4.2 Iron Loss Estimation ...... 45 4.2.1 Loss Separation ...... 45 4.2.2 Rotational Losses ...... 46 4.3 Study Summary ...... 48 4.4 Selected Results ...... 50 5 Form Factors of Salient Pole Shoes ...... 53 5.1 Background ...... 53 5.2 Pole Shoe Form Factors ...... 54 5.3 Study Summary ...... 55 5.3.1 Pole Face Contours ...... 56 5.3.2 Pole Shoe Variables ...... 57 5.4 Selected Results ...... 58 5.4.1 Effect of Pole Face Contour ...... 58 5.4.2 Linear Models with Saturation Considered ...... 59 5.4.3 Perspectives on Pole Shoe Shape Selection ...... 60 6 Analysis of a PM Generator with Two Contra-rotating Rotors . . . . . 61 6.1 Previous Work ...... 61 6.2 Generator Topology ...... 61 6.2.1 Dual Contra-rotating Rotor Topology ...... 61 6.2.2 Reference Machine Topologies ...... 62 6.3 Selected Results ...... 63 6.3.1 Characterization of the Inter-rotor Cross Coupling ...... 63 6.3.2 Synchronized Contra-rotating Load Operation ...... 66 7 Electromechanical Transients - Simulation and Experiments ...... 69 7.1 Previous Work ...... 69 7.2 Rotor Angle Oscillations ...... 69 7.2.1 The Swing Equation ...... 70 7.2.2 Damping and Synchronizing Torques ...... 71 7.3 Study Summary ...... 73 7.3.1 Torque Coefficient Determination from a Field Model . . . 73 7.3.2 Experimental Study ...... 73 7.4 Selected Results ...... 74 7.4.1 Comparison of Field and Circuit Model Responses ...... 74 7.4.2 Experimental Study ...... 76 8 Conclusions ...... 81 9 Suggested Future Work ...... 83 10 Summary of Papers ...... 87
viii 11 Summary in Swedish ...... 93 Acknowledgment ...... 95 References ...... 97
ix
List of Symbols and Abbreviations
Fields
Symbol Unit Definition A Tm Magnetic vector potential B T Magnetic flux density / induction H A/m Magnetic field J A/m2 Current density
Scalars
Symbol Unit Definition Az Tm Z-component of magnetic vector potential bp m Pole body width Bgm T Peak value of air-gap flux density wave Bmax T Peak flux density ΔBr T Radial flux density distortion ei V Induced EMF in armature phase i (i = a,b,c) (field model) ed p.u. Direct-axis armature voltage (equivalent circuit model) e fd p.u. Field voltage (equivalent circuit model) eq p.u. Quadrature-axis armature voltage (equi- valent circuit model) E V or p.u. Internal EMF f Hz Electrical frequency fa - Pole taper f0 Hz Hunting frequency hpp m Pole shoe height H s Inertia constant
xi Scalars (continued)
Symbol Unit Definition
i j A Current in armature phase j ( j = a,b,c) id p.u. Direct-axis armature current i fd p.u. Field winding current (equivalent circuit model)
iq p.u. Quadrature-axis armature current i1d p.u. Direct-axis damper current i1q p.u. Quadrature-axis damper current If A Field current J kgm2 Moment of inertia 4 kc Sm /kg Classical loss coefficient kd - Direct-axis armature pole shoe form factor 3 −0.5 −1 kE Am V kg Excess loss coefficient k f - Field winding pole shoe form factor 4 −1 kH Am (Vskg) Hysteresis loss coefficient kq - Quadrature-axis armature pole shoe form factor
Kd p.u. torque / Damping torque coefficient (rad/s)
Ks p.u. torque / rad Synchronizing torque coefficient le m Effective machine length Lad p.u. Direct-axis mutual inductance Laq p.u. Quadrature-axis mutual inductance Le H Armature end-winding leakage inductance L fd p.u. Field leakage inductance Ll p.u. Armature leakage inductance L1d p.u. Direct-axis damper winding leakage in- ductance
L1q p.u. Quadrature-axis damper winding leakage inductance
Ma A·turns Armature winding MMF MD A·turns Damper winding MMF Mf A·turns Field winding MMF
xii Symbol Unit Definition n rpm Rotational speed
Nd - Number of damper bars per pole Nf - Number of field winding turns per pole Np - Pole pair number q1 - Number of stator slots per pole and phase ptot W/kg Total specific iron loss Padd−dyn % Fractional loss increase due to rotational and harmonic fields
Padd−rot % Fractional loss increase due to rotational fields
Ra p.u. Armature phase resistance Rc Ω Inter-pole end-ring resistance Re Ω Armature end-winding resistance R fd p.u. Field winding resistance R1d p.u. Direct-axis damper winding resistance R1q p.u. Quadrature-axis damper winding resis- tance S m2 Conductor area
Te Nm or p.u. Electrical torque ΔTe p.u. Change in electrical torque Un V or p.u. Rated terminal voltage (RMS, line-to-line) V V Electric potential / applied voltage (field model)
Xd Ω or p.u. Direct-axis synchronous reactance Xq Ω or p.u. Quadrature-axis synchronous reactance Zb Ω Damper bar impedance Γ - Degree of rotation δ Elect. rad. Rotor (load) angle (Chapters 2 and 7) δ m Air-gap length (Chapter 5) Δδ Elect. rad. Rotor angle deviation θ Elect. rad. Electrical angular coordinate
θm Mech. rad. Mechanical angular coordinate Λ Vs/(Am2) Air-gap permeance function
Λecc - Eccentricity permeance function
xiii Scalars (continued)
Symbol Unit Definition −1 ΛP m Pole-shape permeance function Λsat - Saturation permeance function ΛSslot - Stator slot permeance function μr - Relative magnetic permeability μ0 Vs/(Am) Permeability of free space ν m/H Magnetic reluctivity σ S/m Electric conductivity
τD s Damping time constant τds - Damper slot pitch τp m Pole pitch τpc m/- Concentric pole shoe width τpp m/- Pole shoe width τs m Stator slot pitch φ Elect. rad. Power factor angle Ψ Wb turns / p.u. Flux linkage
Ψad p.u. Direct-axis mutual (air-gap) flux linkage Ψaq p.u. Quadrature-axis mutual (air-gap) flux link- age
Ψd p.u. Direct-axis armature winding flux linkage Ψ fd p.u. Field winding flux linkage Ψq p.u. Quadrature-axis armature winding flux linkage
Ψ1d p.u. Direct-axis damper winding flux linkage Ψ1q p.u. Quadrature-axis damper winding flux link- age ω Elect. rad/s Electrical angular frequency
ωm Mech. rad/s Mechanical angular frequency ωms Mech. rad/s Synchronous angular frequency ωs Elect. rad/s Synchronous angular frequency ω0 Mech. rad/s Hunting angular frequency Δω p.u. Angular frequency (speed) deviation
xiv Abbreviations
AC Alternating Current DC Direct Current EC Equivalent Circuit EMF Electromotive Force FE Finite Element FEA Finite Element Analysis FEM Finite Element Method MMF Magnetomotive Force PM Permanent Magnet p.u. Per Unit SiFe Silicon-Iron alloy SMIB Single Machine Infinite Bus THD Total Harmonic Distortion UMP Unbalanced Magnetic Pull
xv
1. Introduction
1.1 Background Large-scale exploitation of hydropower resources in Sweden started in the first decades of the 20th century. The clean and controllable supply of power from hydropower plants was vital for the electrification of the society and the de- velopment of the Swedish industry throughout the century. Today, hydropower still remains an essential ingredient in the national energy mix, and accounts for 46%1 of the country’s annual electricity production of 145 TWh [1]. Reli- able and efficient operation of the hydropower plants is crucial, and this calls for safe and professionally designed plant components. The generator is one of the key components of a hydropower plant, since it constitutes the site for the conversion between mechanical and electrical energy. The work presented in this doctoral thesis is a part of a research program devoted to hydropower generator technology at Uppsala University, initiated by The Swedish Hydropower Centre (Svenskt Vattenkraftcentrum, SVC). SVC is a national collaboration platform for power suppliers, manu- facturers of hydropower equipment, consulting agencies, The Swedish Energy Agency, The Swedish National Grid Agency and five technical universities. SVC’s vision is to promote the provision of qualified human resources to all branches of the national hydropower industry in order to secure an efficient and safe production of hydro electricity in the future, and to secure a main- tained dam safety2. The scientific aim of the doctoral project was to address subjects associ- ated with electromagnetic analysis of synchronous machines with a particular emphasis on grid-connected operation of hydroelectric generators. Because of the general formulation of scope of the project, the work comprises a set of diversified studies. The field of synchronous machine analysis encompasses both electric, mag- netic, thermal and mechanical aspects. As the title of the thesis indicates, the work presented here is largely limited to electric and magnetic phenomena. Electromagnetic analysis is here defined as the study of electric currents, mag- netic fields, electric voltages and power flows in an apparatus during steady- state and transient operating conditions. The scope of the work is somewhat
1Calculated average between the years 2000-2008. 2www.svc.nu. Accessed on January 12 2011.
1 extended with a simple model of electromechanical interaction in studies on synchronous machine hunting (see Chap. 7). The studies that are presented in this comprehensive summary can be di- vided into five main subjects. The ten papers, which constitute the founda- tion of the thesis, are in turn subordinate to either of these five subjects. The first main subject will be referred to as applications of permeance models of salient-pole generators. A series of papers (I, II, and III) fall under this subject. The second subject is core loss prediction. A single publication (Paper IV) be- longs to this category. The third subject is entitled form factors of salient pole shoes, and is represented by Paper V. The fourth main subject concerns a non-conventional permanent magnet (PM) generator topology and is labeled analysis of a PM generator with two contra-rotating rotors. Paper VI em- bodies this subject. The final subject is electromechanical transients. Various aspects of this topic are discussed in Papers VII, VIII, and IX. The last paper, Paper X, deals with design considerations for an experimental generator setup and is the only publication not to fall under any of the main subjects. In spite of the diversity of the addressed problems, some studies that be- long to different main subjects share common denominators. For instance, the damper winding end-ring connection is discussed both in terms of its impact on the armature voltage waveform distortion (Papers I and X) as well as its mitigating effects on rotor angle oscillations (Papers VII and VIII). Moreover, all studies but one (Paper VI), are concerned with the conventional vertical- axis machine topologies that are typically encountered in large hydropower plants. In the following sections, the five main subjects are briefly introduced and the objectives of the individual studies are stated. The chapter is concluded with a presentation of the outline of the thesis.
1.2 Applications of Permeance Models of Salient-pole Generators The rotating field method determines the air-gap flux density in an electrical machine as the product of a magnetomotive force (MMF) and a permeance function. A calculation scheme that uses this approach to derive the air-gap flux density is referred to as a permeance model. In combination with circuit equations that represent the damper winding, it is possible to determine ap- proximately the full air-gap flux density waveform, including the harmonic contribution of the damper reaction [2]. Different applications of the perme- ance modeling technique for synchronous generators with salient, laminated poles are explored in Papers I, II, and III. The aim of the study presented in Paper I was to develop a permeance model suitable for the calculation of open-circuit armature voltage harmonics. In the study summarized in Paper II, the objective was to explore the applicability
2 of the model in studies of steady balanced and unbalanced load operation. Finally, in Paper III, the objective was to assess the usefulness of the perme- ance model in predicting the effects of parallel armature circuits on a steady unbalanced magnetic pull (UMP).
1.3 Core Loss Prediction in Large Hydropower Generators In the conversion between mechanical and electrical energy that takes place in a generator, a certain amount of power is continuously converted to heat through various dissipation mechanisms. This is the loss of the conversion scheme. The term core loss refers to the power loss that is developed in the iron core of the stator. Core losses are fundamentally attributable to the eddy currents that arise in the stator laminations upon exposure of a time-varying magnetic flux. Besides the macroscopic eddy current loss, the internal mag- netic domain structure of the soft ferromagnetic steel used in stator lamina- tions gives rise to additional loss components - hysteresis and excess losses - that also add to the core loss. As high machine efficiency is a prioritized objective, iron loss studies con- tinuously generates many scientific papers. Recently addressed problems in this field include improved material modeling [3, 4], the influence of bidirec- tional magnetic fields (“rotational losses”) [5–8], time-saving analytical loss calculations [9–11], and loss predictions from 3-D magnetic field computa- tions [12]. The goal of the project that resulted in Paper IV was to evaluate the core losses in twelve large hydroelectric generator topologies, using iron loss pre- diction models of varying complexity. An equally important goal was to as- sess the importance of the additional loss introduced by bidirectional magnetic fields in these machines.
1.4 Form Factors of Salient Pole Shoes Hydroelectric generators are typically equipped with salient rotor poles, and the shape of the pole shoe directly affects the appearance of the air-gap flux density waveform. In order to determine the inductances of fundamental wave equivalent circuit representations of synchronous machines, the correlation between the fundamental wave amplitude and the maximum wave amplitude is required. To this end, pole shoe form factors are introduced in the math- ematical expressions of the different machine inductances. Form factors are defined for three reference cases of magnetic excitation and can be said to characterize the pole shoe shape.
3 In the technical literature, many studies on salient pole shoe design and form factors date from the first part of the 20th century [13, 14]. These early studies are founded on analysis techniques that neglect iron saturation and higher order harmonics of the impressed MMF waveforms. The validity of the results for all the practical pole face contour designs that are in use is also unclear. Even so, the results of these studies are usually cited in modern textbooks of synchronous machine design [15]. The primary objective of the work presented in Paper V was to study the effect of iron saturation on pole shoe form factors. The study was however extended to embrace a more general comparison of different pole face con- tour designs from a form factor perspective. Moreover, the harmonic imprint of different salient pole shoes on the air-gap flux density waveform was con- sidered.
1.5 Analysis of a PM Generator with Two Contra-rotating Rotors Hydraulic turbine concepts with two contra-rotating impellers have been pre- sented both for use in small-scale hydropower plants [16] and in tidal energy conversion schemes [17]. The benefits of employing a turbine with two contra- rotating stages include a near-zero reaction torque on the support structure, near-zero swirl in the wake and high relative rotational speeds. For a com- plete energy conversion system employing such a turbine, a generator with two contra-rotating rotors and one single stator winding is an interesting, but unexplored machine concept. Caricchi et al. performed one of the rare studies on this particular type of machine topology [18]. Their communication reports of an axial flux motor with two contra-rotating rotors designed to operate in a ship propulsion drive. Motivated by the possible applicability in small-scale hydro schemes as well as the relative sparsity of available information on electrical machines with contra-rotating rotors, a research project aimed at exploring further the operating characteristics of this machine topology was initiated. A selection of findings are reported in Paper VI.
1.6 Electromechanical Transients - Simulation and Experiments During perfect steady-state operation of a grid-connected synchronous gener- ator, the speed of the rotor is identical to the synchronous speed dictated by the mains frequency. The term electromechanical transient will be used here
4 to denote temporary rotor speed excursions around the synchronous speed, and the associated fluctuations in electrical torque. From a physical perspective, the grid-connected generator is in close anal- ogy with a mechanical arrangement consisting of a discrete mass attached to a wall through a spring and a damper. Electric spring and damper action during rotor swings results from the interaction between the rotor and sta- tor circuits, and is described in terms of synchronizing and damping torques. Because of their importance for stable operation of inter-connected power sys- tems, damping and synchronizing torques of synchronous machines have been extensively studied in the past [19–23]. While previous studies have addressed damping and synchronizing torque calculation with analytical formulae, the objective of the study presented here was to determine these machine properties from numerical field simulations. To this end, a coupled field-circuit model of the classical single machine infinite bus (SMIB) system was developed. Papers VII and IX describe the outcome of the numerical experiments performed with this model, while Paper VIII is concerned with the experimental determination of the natural damping properties of a laboratory generator. Particular attention is devoted to the effect of different damper winding configurations.
1.7 Outline of the Thesis Due to the diversity of the research studies, the author has preferred to devote one chapter to each main subject. Each subject chapter contains a description of the method of analysis and a few, selected results. This unconventional outline was deliberately chosen to facilitate for readers who take interest in one particular subject. The first part of Chapter 2 contains a short introduction on the function and the main construction elements of salient-pole synchronous generators. The second part of Chapter 2 discusses equivalent circuit (EC) and finite element (FE) models of synchronous electric machines. The chapter is then concluded with a presentation of the coupled field-circuit models that were used in the different studies. Next, Chapters 3-7 are devoted to the respective main sub- jects. Permeance model applications are treated in Chapter 3, core losses in Chapter 4, and pole shoe form factors in Chapter 5. Chapter 6 and Chapter 7 are devoted to analysis of a PM generator with two contra-rotating rotors and electromechanical transients respectively. Conclusions are presented in Chap- ter 8 and suggestions for future studies are given Chapter 9.
5
2. Theory
This chapter is intended to serve two purposes. The first purpose is to provide non-expert readers with some useful notions which will assist digestion of the contents of Chapters 3-7. The second purpose is to provide professional read- ers with comprehensive mathematical descriptions of the EC and FE models of synchronous generators that have been used in the different studies. In a spirit of compromise between these aims, some general information on EC and FE models of synchronous generators, which the author deemed manda- tory, is also provided. Section 2.1 describes the main construction elements of hydroelectric gen- erators. In Section 2.2, EC models of synchronous generators are briefly dis- cussed. Furthermore, the EC model structure used in Papers VII, VIII and IX is presented. Next, Section 2.3 provides an introduction to FE generator models. Section 2.4 finally presents the mathematical structure of the coupled field-circuit models that have been used in the different studies.
2.1 Salient-pole Synchronous Generators 2.1.1 Main Construction Elements The purpose of a generator is to convert mechanical energy, supplied from a prime mover via a rotating shaft, to electric energy, which is typically fed into the power grid. This electromechanical energy conversion is realized with the magnetic field inside the generator acting as an intermediate coupling. Most generators in large hydropower plants are synchronous generators with salient rotor poles. The word “large” here denotes a generator in the MW range. In the past, horizontal-axis units were common, but today, the majority of the hydro generating units are built as vertical-axis machines. The two main parts of a conventional hydroelectric generator are the stator and the rotor. The stator consists of a circular magnetic iron core, constructed from thin silicon steel sheets and supported by a steel frame. The inner sta- tor periphery holds uniformly stamped slots, where a three-phase winding is inserted. This is the armature or stator winding. The winding is typically com- posed of form-wound copper coils insulated with a high voltage mica-based insulation system. The rotor, or pole wheel, is attached to the rotating shaft. It consists of a frame, an iron ring made from stacked steel sheets, and rotor poles. The rotor
7 Figure 2.1: (a) Axial cross-section of a salient-pole synchronous machine with four poles. 1. Pole body. 2. Pole shoe. 3. Field coils. 4. Stator winding coils. 5. Damper winding. is separated from the stationary stator by an air-gap. The rotor poles, also constructed from laminated steel sheets, hold the field winding, that provides the fundamental magnetic field excitation. Fig. 2.1 shows the axial cross-section of a four-pole synchronous machine with salient poles. The part of the pole which is closest to the air-gap is re- ferred to as the pole shoe. The pole shoes of large synchronous machines typ- ically hold copper or brass bars. This is the amortisseur or damper winding. The bars in adjacent poles can be connected via a short-circuit ring in both machine ends. This configuration is referred to as a complete or a continuous damper winding.1 A damper winding that lacks the inter-pole connection like- wise has many designations in the technical literature. Any of the terms open, incomplete, non-continuous,orgrill damper winding can be used to denote this damper winding configuration. To deal with the asymmetric air-gap produced by the pole saliency, it is convenient to introduce two sets of rotor-fixed reference axes - the direct (d) and quadrature (q) axes (see Fig. 2.1). A d-axis is aligned with the center axis
1Some prefer to refer to this configuration simply as a squirrel cage winding.
8 of a north pole. The q-axes go through inter-polar gaps adjacent to and leading the d-axes.
2.1.2 Grid-connected Operation Most of the global electric energy generation is performed through synchronous generators connected to three-phase alternating current (AC) power grids. The rotational speed, n, of a grid-connected synchronous generator is given by f n = 60 · [rpm], (2.1) Np where f is the grid frequency and Np denotes the number of pole pairs in the generator. n is referred to as the synchronous or rated speed of the unit. During normal load operation, balanced three-phase currents in the arma- ture winding phases produce a magnetic field that rotates at synchronous speed. This field is called the armature reaction. The fundamental waves of the armature reaction and the rotor excitation field have the same number of poles and are at standstill with respect to each other. Through the interaction between these fields, a non-zero synchronous torque is produced which tend to align the fields with each other. During balanced load operation, the angle between the rotor and the armature fields is more or less constant, and the syn- chronous torque production is manifested as a continuous transfer of power to the AC grid. The steady active and reactive power productions, Pg and Qg, from a syn- chronous generator are approximately given by 3EU 3 2 1 1 Pg = sinδ + U − sin2δ Xd 2 Xq Xd (2.2) 2δ 2δ 3EU 2 cos sin Qg = cosδ − 3U + . Xd Xd Xq
In the above expressions, the resistive losses in the stator winding are ne- glected. E is the so called internal EMF (here, an RMS phase quantity in Volts), U is the terminal voltage (RMS phase quantity in Volts), and Xd (Ω) and Xq (Ω) denote the synchronous reactances in the direct- and quadrature axes respectively. δ is the load angle (or rotor angle), and corresponds to the phase angle between the voltages E and U. The function Pg(δ) is called the active power - load angle characteristics of the synchronous generator and is schematically illustrated in Fig. 2.2. During normal operation, the synchronous generator operates at a load an- gle that is considerably smaller than the critical load angle, δC. The angle δC corresponds to the maximal active power delivery at a given level of excita-
9 Figure 2.2: Active power versus load angle (synchronous generator). tion. The generator is considered to be stable with respect to slow shaft torque 2 or load variations as long as the load angle does not exceed δC [24].
2.2 Equivalent Circuit Generator Model In studies of the electromagnetic interaction between synchronous generators and other electrical equipment, the generators are frequently represented by a set of electrical circuit equations. A long tradition of elaborate refinement and adaptation of such circuit representations to fit almost any problem of interest, makes this the most established and accessible form of generator analysis. A number of factors determine the nature of a generator EC model. Some of the most important factors are briefly discussed in the following.
Nominal or P.U. Representation of Model Variables Model quantities can be represented with physical units (V, A, W e.t.c) or, alternatively, units are eliminated from the calculations by expressing all quantities in terms of fractions of specified base values. The latter approach is called per unit (p.u.) representation. The p.u. representation is convenient in power systems with many different voltage levels, and also facilitates the comparison of electrical arrangements with dissimilar power ratings.
Winding Representation The armature can be modeled by its three physical stationary armature phases A, B, and C or, alternatively, by means of fictitious rotor-fixed wind- ings. A stator-fixed representation is usually referred to as a phase domain model, while the rotor-fixed representation is called a dq0 or two-axis model.
2This is the static stability of the generator, and is defined as the ability of the generator to remain in synchronism with the power grid when subjected to slow shaft power or load varia- tions.
10 Two-reaction theory, which forms the basis of two-axis representations of syn- chronous machines, was originally worked out by Blondel [25]. The dq0-representation brings about numerous modeling advantages, such as time-independent circuit inductances and decoupling of the d- and q-axis circuits if iron saturation is neglected. The approach involves the application of the Park transformation to all stator quantities [26]. In power system analysis software, the internal electrical representations of synchronous machines are almost exclusively in dq0-coordinates. Applica- tions of physical armature phase representation in EC models however do ex- ist. A prominent example is the analysis of internal short-circuit faults [27,28]. The representation of the rotor windings primarily concerns the structure of the equivalent damper winding circuits [29, 30]. The level of modeling detail should be adjusted to the problem at hand and the required accuracy of the results.
Consideration of Non-Linear and Harmonic Effects Most dq0-models are fundamental wave models, that is, they only consider the dominating space fundamentals of the magnetic flux density waves inside the generator. Linear EC models either neglect iron saturation or represent the effect by parameter values appropriate at the studied point of operation (“saturated parameters”). Iron saturation can alternatively be accounted for with refined iterative methods [31]. It is also possible to account for some harmonic effects on generator performance [32].
Choice of Independent State Variables The selection of independent state variables depends on the circuit repre- sentation [33]. For fundamental parameter circuit representations, winding flux linkages and currents are preferably used. In some applications, a mixed or “hybrid” choice of independent variables may be the best choice [34].
2.2.1 P.U. Electrical Equations EC generator models were used in the studies presented in Papers VII, VIII and IX. The employed circuit model was a dq0-model with one damper circuit in each axis, which is customary for rotor angle stability studies of hydroelec- tric generating units. The model was represented in the conventional Lad-base reciprocal p.u. system. In Paper IX, the circuit parameters were derived from FE simulations of standard parameter determination tests [35]. In Papers VII and VIII, the parameters were calculated from generator design data. The em- ployed analytic parameter calculation formulae were taken from [15] and [36]. The p.u. electrical equations of the EC model are listed below. All variables, including time, are given in p.u. Zero-sequence equations are omitted, since only balanced generator operation was considered in the studies where a cir- cuit model was used. The system of differential-algebraic equations used to
11 Figure 2.3: Circuit representation of voltage and flux linkage equations. Top: d-axis circuit. Bottom: q-axis circuit. simulate the SMIB systems of Papers VII and IX, can be derived from the ex- pressions below, except for the two equations that describe the grid coupling. These equations are summarized in Paper IX. The listed equations can be represented with the equivalent d- and q-axis circuits shown in Fig. 2.3. The notation follows that used in IEEE Std. 1110- 2002 [29], but for completeness all symbols are also described in the List of Symbols.
Stator voltage equations dΨ e = d − Ψ ω − R i (2.3) d dt q a d dΨ e = q + Ψ ω − R i (2.4) q dt d a q Rotor voltage equations dΨ e = fd + R i (2.5) fd dt fd fd dΨ 0 = 1d + R i (2.6) dt 1d 1d dΨ 0 = 1q + R i (2.7) dt 1q 1q Stator flux linkage equations
Ψd = −(Lad + Ll)id + Ladi fd + Ladi1d (2.8) Ψq = −(Laq + Ll)iq + Laqi1q (2.9)
12 Rotor flux linkage equations
Ψ fd = −Ladid +(Lad + L fd)i fd + Ladi1d (2.10) Ψ1d = −Ladid + Ladi fd +(Lad + L1d)i1d (2.11) Ψ1q = −Laqiq +(Laq + L1q)i1q (2.12)
Air-gap torque Te = Ψadiq − Ψaqid (2.13)
2.3 Finite Element Generator Model In equivalent circuit models, the inherently distributed nature of the electro- magnetic interaction inside the generator is “lumped” into a fairly limited set of equations. We here define a field generator model as a model that deter- mines the electrical performance directly from the magnetic field distribution in the active parts (stator, air-gap, rotor) of the generator. The magnetic field distribution is determined from Ampères law, which needs to be appropriately formulated for the application at hand. The prob- lem of solving the field equations by means of digital computing can then be tackled with a variety of numerical methods. For electromagnetic analysis of electrical machines, the Finite Element Method (FEM) has emerged as the most widely applied numerical method. Its popularity is linked to its ability to handle the complicated calculation geometries presented by rotating machin- ery [37]. FEM was originally used to study problems in structural mechanics. Its employment for the solution of the electromagnetic vector field problems pre- sented by electric machinery became widely diffused in the 1980’s [38, 39]. Today, FE analysis is more or less a standard tool in electrical machine de- sign, and the method can be used to study problems of both electromagnetic, thermal, mechanical and coupled (“multiphysics”) nature. There exists a num- ber of commercial FE software packages specifically designed for analysis of electromagnetic field problems3.
2.3.1 Calculation Geometry and Material Property Assignment The problems addressed in this thesis have been analyzed with a two- dimensional field model. The magnetic field was determined with FEM, and therefore the terms field model and FE model will be used interchangeably to denote this generator modeling approach. The two-dimensionality of the
3http://www.ansys.com/Products/Simulation+Technology/Electromagnetics (accessed on Jan- uary 19 2011) http://www.cedrat.com/en/software-solutions/flux.html (accessed on January 19 2011) http://www.comsol.com/products/acdc/ (accessed on January 19 2011)
13 Iron
Conductor
Air
Figure 2.4: Calculation geometry example (one pole pitch of a hydroelectric genera- tor). model means that it is assumed that the magnetic field in the generator is perfectly parallel to the axial cross-section of the generator. For most problems, symmetry conditions allow for a radical reduction of the region where the magnetic field needs to be evaluated. Fig. 2.4 shows an example of such a reduced calculation geometry, corresponding to one pole pitch of a hydroelectric generator. Lines demarcate different subdomains of the calculation geometry. These regions represent the physical parts of the generator, such as rotor iron core, field winding conductors, stator teeth and stator winding conductors. The subdomains are allocated material properties relevant for the electromagnetic field problem, such as electric conductivity, σ, and relative magnetic perme- ability, μr. Non-linear ferromagnetic material properties are represented by single-valued B(H)-curves.
2.3.2 Field Equation Formulation The FE code used in the thesis solves Ampère’s law for the magnetic vec- tor potential, A. In the 2-D formulation of the problem, A has only an axial component, denoted Az. Az is related to the Cartesian components of the flux density B according to ∂A B = z (2.14) x ∂y ∂A B = − z (2.15) y ∂x Bz = 0. (2.16)
Hence, there is no axial component of flux density, as dictated by the 2-D nature of the field problem formulation. The magnetic vector potential inside
14 the cross-section of the generator is assumed to be governed by the following partial differential equation4:
In conductor subdomains: ∂ ∂ ( , , ) ∂ ∂ ( , , ) ∂ ( , , ) ∂ ( , , ) ν Az x y t + ν Az x y t = σ Az x y t + σ V x y t ∂x ∂x ∂y ∂y ∂t ∂z Elsewhere: ∂ ∂A (x,y,t) ∂ ∂A (x,y,t) ν z + ν z = 0 ∂x ∂x ∂y ∂y (2.17)
Here, 1 ν = , (2.18) μrμ0 denotes the reluctivity, μ0 is the permeability of free space and V is the electric potential. The right-hand side of (2.17) is the total current density. As seen in the equation, only subdomains that correspond to conductors are allowed to have a non-zero current density. The conductor subdomains are therefore referred to as the sources of the field problem. The total current density typically depends on the nature of the conductor subdomain and the circuit to which it is connected. Additional coupling equations are typically required to completely specify the field problem in a conductor.
Equation (2.17) warrants the following supplementary remarks:
σ ∂V (x,y,t) 1. The term ∂z denotes the applied current density while the term ∂ ( , , ) σ Az x y t ∂t denotes the induced current density. 2. The applied current density plays a key role when one or several conductors are connected in series. In such a situation, the induced current density may not be equal in the different conductors, but the net current must be the same in all conductors. The electric charge distribution introduced by the applied current density term then ensures that this condition is met [41]. The quantity V , which is referred to as the applied voltage, is constant over the conductor subdomain area, and is directly proportional to the potential difference between the (fictitious) ends of the conductor. 3. If the dynamic interaction between the magnetic field and the conduc- tors is to be disregarded, the conductor currents may be specified by pre- determined functional expressions. This is equivalent to connecting the conductors to ideal current sources.
4For a full derivation of this equation see [40] or any textbook on finite element analysis of electrical machines.
15 ∂Az(x,y,t) 4. The term ∂t only appears explicitly in conductor subdomains treated as solid conductors [42], where eddy currents provoke a non-uniform spa- tial current distribution in the conductor cross-section. 5. In the FE models used to study the subjects of this thesis, all conductor sub- domains have been treated as filamentary conductors. That is, the current calculated in a given time step is assumed to be uniformly distributed across the subdomain. In the coupled field-circuit models to be described subse- quently, the induced current density is nevertheless considered on average terms in additional coupling equations.
2.3.3 Finite Element Discretization There exist different techniques to solve (2.17). The starting point for most FE solvers is to reformulate the problem on a variational form. In essence, this means that the problem of finding a function Az(x,y,t) that satisfies (2.17) is transformed into the problem of finding a function Az(x,y,t) which is a sta- tionary point to some functional, F . For the problem at hand, F is typically set to the electromagnetic energy of the system: B F = H · dB − JA dS. (2.19) 0 S
Here, H denotes the magnetic field, J is the current density, and S refers to the area of the calculation geometry. The search for a solution is carried out with ∗ trial functions Az , N ∗ Az(x,y,t) = ∑ A jϑ j(x,y,t), (2.20) j=1 where A j are unknown coefficients and ϑ j are called base functions. The fundamental principle of the finite element method is to subdivide the calculation geometry into many small, non-intersecting elements and make use of base functions that are non-zero only within a single element. If the el- ements are sufficiently small, the base functions of (2.20) can be very simple, without much loss of computational accuracy. Typically, base functions that are linear or quadratic functions of the spatial coordinates x and y are used. The elements in 2-D FEM are usually shaped as triangles and the vertices of these triangles are referred to as nodes. The complete body of elements is called a mesh. A mesh of triangular elements is illustrated in Fig. 2.5. In the FE formulation of the variational problem, the coefficients A j denote the magnetic vector potentials in the nodes of the mesh. With a trial solution on the form presented in (2.20), it can be shown that the variational problem transforms into a system of ordinary differential-algebraic equations, with the node potentials as the unknown variables. Accordingly, an appropriate numer-
16 Figure 2.5: Triangular mesh in a part of the calculation geometry.
Figure 2.6: Boundary conditions for the example calculation geometry. ical integration method can provide a solution to the original field problem in (2.17). If the calculation geometry contains domains with non-linear magnetic properties, the field solution in every time step is computed through an itera- tive procedure that determines the element reluctivities.
2.3.4 Boundary Conditions For the field problem to be completely specified, the outer borders of the cal- culation geometry need to be assigned with appropriate boundary conditions. Fig. 2.6 exemplifies two boundary conditions that are frequent in finite ele- ment analysis (FEA) of electrical machines - the Dirichlet and the periodic boundary condition. A homogeneous Dirichlet boundary condition sets Az to 0. This is equi- valent to consider the material external to the boundary to have zero relative
17 permeability (a perfect “magnetic insulator”). The periodic condition exploits the repetitive features of the magnetic field inside the machine, and relates the values of Az on two boundaries. In Fig. 2.6, Az on the upper boundary is equal in magnitude but opposite in sign to Az on the lower boundary. Also indicated in Fig. 2.6 is a sliding mesh condition in the middle of the air-gap. This condition is used in time-stepped simulations to mimic rotor motion. In essence, the sliding mesh condition is the intersection between the interfaces of the separately meshed stator and rotor. The potentials of the rotor and stator nodes on the intersection are found through an interpolation procedure. This approach allows for the use of a variable integration time step.
2.3.5 Calculation of Air-gap Torque and Induced EMF It is possible to derive many different electric and magnetic quantities from the field solution. Here, the expressions for air-gap torque and induced winding EMF are provided. The air-gap torque of the field model is of relevance for Papers VI, VII, VIII, and IX. The induced winding EMF formula was used in the studies summarized in Papers I and VI. The air-gap or electrical braking torque in the generator is given by
Te = ler0 σt dγ, (2.21) Γ0 where le is the effective machine length, Γ0 is an arc in the air-gap, r0 is the arc radius and σt is the tangential stress. σt is given by 1 σt = BrBt , (2.22) μ0 where Br and Bt denote the radial and tangential flux density components respectively. The magnetic flux crossing a surface of effective length le and spanning between the points (x1,y1) and (x2,y2) is
Φ = le · (Az(x1,y1) − Az(x2,y2)). (2.23)
The flux linkage, Ψ, of an arbitrary machine winding can hence be calculated from the 2-D field solution as l Ψ = e ∑ A dS − ∑ A dS , (2.24) + z − z S n+ S n− S where n+ and n− are the total number of positively and negatively oriented winding conductors respectively, and S+ and S− are the corresponding con- ductor areas. It is assumed that S+ = S− = S.
18 The induced winding EMF is derived from the flux linkage as dΨ e = − . (2.25) w dt
2.4 Coupled Field-circuit Models The conductors in a generator field model are inter-connected to form com- plete windings. The terminals of the field and armature windings are addition- ally connected to external circuits. As the inclusion of conductor subdomains in windings and circuits affects the conductor currents, additional coupling and circuit equations are required for the field problem to be completely spec- ified in these subdomains. A model where field and circuit equations are solved simultaneously to pre- dict the behavior of an electric apparatus is usually referred to as a coupled field-circuit model. This section provides the circuit equations for the coupled field-circuit models used in the different studies of the thesis. The coupling equations needed to associate a set of conductor subdomains to a winding are also given.
2.4.1 Coupling Equations for Circuit-connected Conductors For a conductor subdomain that is a part of an electric circuit, the field equa- tion (2.17) in that subdomain is supplemented with the following coupling equations dAz σ dS − σψc = 0 (2.26) Sc dt ∂V σψ + S σ c + I = 0, (2.27) c c ∂z where Sc denotes the conductor area, Vc the applied conductor voltage, and I is the current in the conductor. ψc is the induced conductor EMF integrated over the conductor surface. The variables I and Vc needs to be determined from additional circuit equations, to be presented subsequently. The structure of (2.26) and (2.27) is the same for all conductor subdomains that are connected to circuits, regardless if the conductor is a part of the field, damper or armature winding. The exact formulation of the additional circuit equations for the field, damper and armature windings depends on the studied problem, as discussed in the following. Before the circuit equations are introduced, we state the expression for the total electric potential difference across a winding of series-connected con-
19 ductors:
Vw = le ∑ Vc − ∑ Vc . (2.28) c∈C + c∈C −
C + here denotes the set of positively oriented conductors, and C − is the set of negatively oriented conductors in the winding.
2.4.2 Rated Voltage No-load Operation Model Rated voltage no-load operation was studied in Papers I, IV and X. Simulation of rated voltage operation at no-load implies consideration of the requirement e2 + e2 + e2 a b c = U , (2.29) 2 n where ea, eb, and ec denote the induced armature phase EMFs and Un is the rated line-to-line voltage of the generator. The field voltage is adjusted such that (2.29) is met. A short numerical transient is to be expected before the problem converges.
Field Circuit Equation The additional circuit equations that complete the problem specification in field conductor subdomains at rated voltage no-load operation are
u fd0 −Vfd = 0 (2.30) i f + − i f − = 0. (2.31) u fd0 is the field voltage at no-load operation at rated armature voltage and speed and Vfd is the potential drop across the entire field winding. Vfd effectively provides the coupling to (2.17) and (2.26) - (2.27) through (2.28). i f + and i f − denote the currents in conductor subdomains on opposite sides of the pole body. The effects of end winding leakage flux are neglected.
Damper Circuit Equations The damper circuit equations are based on a work by Shen and Meunier [43]. Definitions of relevant quantities are shown in Fig. 2.7. To state the cir- cuit equations on a compact form, the following column vectors are intro- duced: T i =[i1 i2 ... in] (2.32) T j =[j1 j2 ... jn] (2.33) T Vb =[Vb1 Vb2 ... Vbn] (2.34) T ve =[ve1 ve2 ... ven] . (2.35)
20 (a)
(b) Figure 2.7: Damper winding equations in the field model. (a) Definition of bar and end-ring currents. (b) Definition of bar potentials and end-ring voltage drops.
21 Here, the integer n denotes the number of damper bars considered in the cal- culation geometry. For generators with integral slot armature windings, 2N (continuous damper winding) n = d (2.36) Nd (non-continuous damper winding), where Nd denotes the number of damper bars per pole. From Fig. 2.7, the following relations can be established between the bar cur- rent vector i, the end-ring current vector j, the bar potential vector Vb and the end-ring voltage vector ve: i = MT j (2.37) MVb = 2 ve (2.38) ve = Red j. (2.39)
M denotes the (n × n) matrix ⎡ ⎤ 1 −10...... 0 ⎢ ⎥ ⎢ 01−10... 0 ⎥ ⎢ ⎥ ⎢ − ... ⎥ ⎢ 001 10 ⎥ = ⎢ ⎥ M ⎢ ...... ⎥ (2.40) ⎢ . .01 . . ⎥ ⎢ ...... ⎥ ⎣ ...... ⎦ −10...... 01 and Red denotes the diagonal (n × n) matrix ⎡ ⎤ R 0 ...... ⎢ e1 ⎥ ⎢ ... ⎥ ⎢ 0 Re2 00 ⎥ ⎢ . ⎥ = ⎢ . .. .. ⎥. Red ⎢ .0 . . 0 ⎥ (2.41) ⎢ . . . . ⎥ ⎣ ...... 0 ⎦
0 ...... 0 Ren