ICEM 2006 – paper reference no 509 1

Abstract—A small power 9-phase axial-flux PM synchronous

generator was built to validate the design and optimization process of a 5MW direct-drive wind turbine. Two quality criteria are imposed for the system. The electromagnetic torque of the generator and the DC bus current of the rectifier must not contain low-order harmonics. Nominal load tests, where the generator is supplied via 9-phase IGBT rectifier and the control is realized using the multi-machine multi-converter systems concept, allow verifying these criteria. The parameters of the machine are determined using no-load tests. The results obtained by numerical and analytical models are compared with the experimental measurements.

Index Terms—Axial-flux, PM synchronous generator, polyphased machine, experiments.

I. INTRODUCTION HE wind generators, the so-called direct-drive wind

turbines, characterized by high torque and low speed parameters, represent one of the PM synchronous machines applications in the energy industry. The nowadays applications, using PM generators, have nominal power up to 5 MW for geared drive solutions. For the gearless drives, up to 3MW systems are reported for on-shore applications [1].

A PM machine was designed for a direct-drive 5 MW wind generator application. The machine is an axial-flux one, using a polyphased concentrated winding. Polyphased structures offer the advantage of modularity, with immediate consequences for the fabrication process, assembly, transportation and maintenance.

Several solutions can be considered: sinusoidal or trapezoidal waveform [2], 3-phase and polyphased systems. Different systems, with different number of phases and different shift displacement were compared in [3].

Numerical models were used for preliminary studies for both machine and converter. To investigate the performances of polyphased systems, a real time vector control is implemented. The configuration chosen based on the system specification and imposed quality criteria, consists in 9-phase

Manuscript received September 2, 2006. The work presented in this paper

was done within FuturElec2 Archimed Project, supported by the CNRT in Electrical Engineering and Nord-Pas de Calais Region in France.

D. Vizireanu, S. Brisset, and P. Brochet are with L2EP, Ecole Centrale de Lille, Cité Scientifique – BP 48, 59651 Villeneuve d’Ascq cedex, France.

X. Kestelyn and E. Semail are with L2EP, ENSAM, 8 bd Louis XIV, 59046 Lille cedex, France.

system, where each star is supplied via a 3-phase back-to-back converter, and the converters are parallel connected via the DC bus.

The first quality criteria is the electromagnetic torque, and the interest is to minimize torque oscillations, which cause lower mechanical stability, audible noise and accelerated ageing of the machine due to vibrations.

The DC bus current represents the second quality criterion. To avoid over-voltage and to protect the transistors, the control of the inverter imposes a constant DC voltage. If the DC voltage is maintained constant, the DC current waveform will give an indication about the power transfer. Reducing harmonic content of the DC bus current will allow reducing the size of the DC bus filter and the harmonic filter at the output of the converter.

To validate the analytical and numerical models, a small-scale prototype was built. The paper describes the fabrication technique and the different tests realized to verify the quality criteria. In terms of modeling, the multi-machine multi-converter systems (MMS) concept is used to realize the control of the system in order to maximize the regenerated power.

II. THE PROTOTYPE The generator is a polyphased one, where the 9 phases form

a triple-star configuration where the windings are uniformly distributed. Its architecture consists on a discoid, axial flux, two outer stators and one interior PM rotor (Fig. 1a and 1b), where the magnets, placed face-to-face on both sides of the rotor, have opposite polarities (North South – NS – topology).

Each stator has 9 slots. The stators are realized from one long rolled metal sheet (fig. 1b), and the slots are cut-off with a constant width. The winding is a concentrated one (fig. 1d), which avoids coils overlapping. It represents a very promising solution because the coils are very easy to insert in the slots. The end-windings are shortened compared to a distributed winding. Therefore, the radial bulk and the copper losses are reduced.

The prototype has 10 poles. The magnets have trapezoidal shape (fig. 1e) and the magnet pole arc to pole pitch ratio is 0.8.

The construction allows variation of the air-gap thickness and variation of the mechanical angle between the two stators, which allows investigating their influence on parameters of

Experimental Tests on a 9-phase Direct-Drive PM Axial-Flux Synchronous Generator Darius Vizireanu, Xavier Kestelyn, Stéphane Brisset, Pascal Brochet, and Eric Semail

T

ICEM 2006 – paper reference no 509 2

the machine like the no-load back electromotive force (EMF), and the cogging torque. The windings of the same phase placed on different stators are connected in series. Table I gives the main parameters of the prototype.

III. EXPERIMENTAL BENCH The experimental test bench is presented in fig. 2. A

classical 3-phase synchronous machine is used to drive the polyphased generator with variable speed. The machines are coupled on the same shaft and a torque sensor is used to measure the mechanical torque. The driving machine is supplied from the grid using a classical 3-phase AC-DC-AC

converter. A speed control is implemented for the driving machine. The driving machine is equipped with a position encoder, and the information supplied by the encoder is emulated and used for the current control of the 9-phase generator. The generator is supplied via a 9-phase IGBT rectifier. The converters have parallel-connected DC busses. Six current sensors are used, to measure two currents for each star of the generator. The real-time current control of the generator is implemented using a dSPACE control board.

IV. NO-LOAD TESTS The parameters like phase resistances, phase inductances

and mutual inductances are determined by specific no-load tests. Supplying, one by one, the windings of each phase by a DC source and measuring the currents, the phase resistance is determined.

To determine the inductances, one phase of the generator is supplied using an AC source, connected in series with a resistor in order to limit the current level. The circuits of the other phases are open. Measuring the current and voltage across the supplied phase, the phase inductance (sum of gap and leakage inductances) is determined. The mutual inductances are calculated measuring the voltages induced on the other phase windings. The values of measured resistances and inductances are presented in Table II. The measured values are closed to the values estimated with the analytical

a)

b)

c) Half-machine d) Stator e) Rotor

Figure 1. 9-phase axial flux prototype

Table I. Characteristics of the 9-phase prototype

Nominal torque [Nm] 40 Nominal speed [rpm] 200

Number of phases 9 Number of poles 10

Outer diameter [mm] 250 Inner diameter [mm] 150

Stator yoke thickness [mm] 35 Slots height [mm] 40

Rotor yoke thickness [mm] 50 Number of turns per coil 747 Air-gap thickness [mm] 4 Magnet thickness [mm] 10

Figure 2. Experimental test bench

Outer stator Outer stator

Inner PM rotor

Rolled sheet

Tooth winding

3-phase driving machine

9-phase prototype

3 phase converter

9-phase converter

6 current sensors

dSPACE control and acquisition board

Torque sensor

ICEM 2006 – paper reference no 509 3

model. In table II, M1, M2, M3 and M4 are respectively the mutual inductance between two phases of the same star corresponding to a phase shift of 120°, 40°, 80°, and 160°.

For the phase resistance, there is a good correlation between measurements and analytical calculations. On the other hand, a difference of 25% occurs on the phase inductance, due to the end winding inductance, which is not taken into consideration in the analytical model. Concerning the mutual inductances, their values depend on the gap inductance and the error is inferior to 10%.

Other no-load test allows determination of the cogging torque and no-load phase EMF. The generator is rotating at nominal speed. The torque sensor measures the no-load torque, which has a continuous component due to mechanical and iron losses of the prototype. The driving machine has a very low cogging torque. After subtracting the continuous component, the cogging torque of the generator is obtained (fig. 3). An FFT analysis shows that the spectrum of the cogging torque contains harmonics related to the number of slots (30 Hz and multiples). The 60Hz harmonic has the highest amplitude, due to a small dissymmetry of the rotor. Its amplitude is about 0.238Nm (0.17Nm RMS), which represents 0.4% of the nominal torque.

For the same conditions (no-load and the machine turning at nominal speed), the no-load EMF voltages are measured. Figure 4 shows the 9-phase no-load EMF, for the case where the windings of the stators are connected in series. It can be seen that the 9-phase system is well balanced. The FFT analysis (fig. 5) shows that the resulting no-load phase EMF has a 5% 3rd harmonic, while the others are less than 0.5%.

The same RMS values are obtained with the finite element (FE) model and the measurements, but the measured third harmonic has a lower value than the one predicted using FE analysis, i.e. 8%.

V. DYNAMIC TESTS If 3-phase machines are well known in terms of modeling

and control, the multi-phase machines require a specific attention. In fact, multiplication of the phase number makes the analysis and the achievement of the control scheme delicate. To simplify the study of multi-phase machines, several methods have been proposed. Among these methods, the MMS concept [4]-[5] transforms the real m-phase drive into several fictitious one-phase or two-phase systems. Using a space vector analysis, the overall analysis and control of the m-dimensional system becomes then much simpler.

Using the MMS concept, the 9-phase machine is equivalent to a set of five fictitious machines: four two-phase and one one-phase. Table III gives the harmonic set associated to each fictitious machine. Due to the symmetry of system, only the odd harmonics are considered.

It can be observed that if sinusoidal currents are imposed, only the main machine contributes to the torque production. The torque ripple occurs if the EMF contains 17th or 19th order harmonics. To obtain torque using the second and the fourth machine, 7th respectively 5th current harmonics must be injected. The gain will be an increase of power density, unfortunately accompanied by supplementary losses associated to the current harmonics. In the case of the 9-phase prototype, the EMF spectrum does not contain significant 5th and 7th harmonics and no power increase will be obtained. Consequently, a torque reference will be imposed only for the

Table II. Electrical parameters of the machine

Measure Analytical Rph [Ω] 22 Ω 22 Ω

Lph 0.49 H 0.4 H M1 -33 mH -36 mH M2 -24 mH -30 mH M3 -27 mH -33 mH M4 -69 mH -65 mH

Figure 3. Cogging torque of the prototype and its FFT analysis

-300

-200

-100

0

100

200

300

-0,1032 -0,0532 -0,0032 0,0468 0,0968

EMF1 EMF7 EMF8 EMF9 EMF6 EMF2 EMF3 EMF4 EMF5

Figure 4. The 9-phase EMF waveforms

Figure 5. FFT analysis of the no-load phase EMF

ICEM 2006 – paper reference no 509 4

main machine. Since the no-load EMF of each fictitious machine is

sinusoidal, rotation operation on each EMF leads to constant quantities (so-called d-q components). As in the case of 3-phase machines, the implemented algorithm for each fictitious machine is similar, the only difference is that rotating reference frames of each machine are deduced using the following rotation operations: R(ωt), R(3ωt), R(5ωt) and R(7ωt) [5]. Thus, simple PI controllers and PWM converters can be used. Figure 7 shows the control structure for a nine-phase machine.

The DC buses of the converters are parallel connected and regenerated power is injected in the DC bus of the primary converter. Thus, the grid supplies only an energy equivalent to the losses of the system. The general electric connections diagram is presented in Fig. 8.

To maximize the generated power, the currents (fig. 10) are imposed in phase with the no-load phase EMF. The results for the nominal conditions (200 rpm and 40 Nm) are presented. The measured torque corresponds to the torque partition

between different fictitious machines. The main machine is loaded with nominal current Iq1 = 1.5A, corresponding to 0.5A RMS phase current (fig. 11), while for the other machines no load is imposed (fig. 12 and 13).

The DC bus current obtained for a 9-phase is the sum of the DC bus currents of each 3-phase rectifier associated to each star, parallel connected through the DC bus. The measurement of the DC bus current shows a good power balance. An FFT analysis of the DC bus current (fig. 14) shows that its spectrum contains only harmonics related to the PWM frequency. No low frequency harmonics appears. The mechanical power is 840W. The DC bus current between the two converters has a mean value of 1.37A. The DC bus voltage is 576V. Therefore, the DC bus power is 790W. The difference between them represents the copper losses of the generator, i.e. 50W. The same value is found using the number of phases (9), their resistance (22ohms) and the phase current (0.5A). The power balance confirms that the iron losses can be neglected for this type of machine, due to fabrication technology (one rolled sheet) and due to low frequency regime (16.7Hz). However, significant differences appear between measurements and simulations on the power factor of the generator. The FEA and analytical models lead to the same value of the power factor, i.e. 0.99, while 0.95 is measured. This difference may come from a bad calculation of the cyclic inductance. Indeed, the hypothesis of linear magnetic material is used both in analytical model and MMS concept. The end-winding inductance is also neglected in FE analysis and analytical model.

Figure 9 – Measured torque for nominal conditions

Figure 10 – The 9 phase currents for nominal conditions

Table III. Multi-machine characterization of a nine-phase machine

Fictitious machine Harmonic order Zero-sequence 9n(9,18,27,...)

Main 9n±1(1,17,19,...) 2nd 9n±2(7,11,25,…) 3rd 9n±3(3,15,21,…) 4th 9n±4(5,13,23,…)

Figure 7. The control structure for the 9-phase machine

Figure 8. Electrical basic diagram of the experimental bench

ICEM 2006 – paper reference no 509 5

VI. CONCLUSION A small scale 9-phase axial flux PM generator was built to

validate the design and optimization processes of a 5MW wind turbine. A comparison is done between the results obtained from numerical and analytical models on one hand, and experimental measurements on the other hand. A good correlation was found for the phase no-load EMF, phase resistance, mutual inductances, cogging torque and DC bus current. However, significant differences occur for the cyclic inductance and the power factor. The analytical conception does not take into consideration the saturation effect, end-windings inductance, which seems to be important for machines with short stator yoke length.

The dynamical performances are evaluated using MMS concept, which allows the transformation of a polyphased

machine into several fictitious one-phase or two-phase machines, leading to an easier analysis and control. The experimental results validate the MMS method for a 9-phase PM axial flux synchronous generator and confirms the results obtained for the quality criteria by numerical simulations.

ACKNOWLEDGMENT The work presented in this paper was done within

FuturElec2 Archimed Project, supported by the CNRT in Electrical Engineering and Nord-Pas de Calais Region in France.

REFERENCES [1] A. Binder, T. Schneider, “Permanent magnet synchronous generators for

regenerative energy conversion – a survey”, Proceedings of EPE2005, Dresden, Germany.

[2] D. Vizireanu, S.Brisset, P.Brochet, Y. Milet, D. Laloy, “Investigation on brushless dc machine suitability to direct-drive generator wind turbine”, Electromotion Quaterly Journal, Vol 12, No. 4, Oct-Dec. 2005.

[3] D. Vizireanu, X. Kestelyn, S. Brisset, P. Brochet, Y. Milet, D. Laloy, “Polyphased Modular Direct-Drive Wind Turbine Generator”, Proceedings of EPE2005, Dresden, Germany.

[4] E. Semail, A. Bouscayrol, J.P. Hautier, “Vectorial formalism for analysis and design of polyphase synchronous machines”, EPJ AP (European Physical Journal-Applied Physics), vol. 22 no 3, June 2003, pp. 207-220.

[5] X. Kestelyn, E. Semail, A. Bouscayrol, “Right Harmonic Spectrum for the Back-Electromotive Force of a n-phase Synchronous motor”, IAS 2004 W.-K. Chen, Linear Networks and Systems (Book style). Belmont, CA: Wadsworth, 1993, pp. 123–135.

Fig. 14 – DC bus current and its FFT analysis

Fig. 11 - d, q currents of the main machine

Fig. 12 - d,q currents of the second fictitious machine

Fig. 13 - d,q currents of the fourth fictitious machine