Design of High Performance Permanent-Magnet Synchronous Wind Generators

Abstract

This paper is devoted to the analysis and design of high performance permanent-magnet synchronous wind generators (PSWGs). A systematic and sequential methodology for the design of PMSGs is proposed with a high performance wind generator as a design model. Aiming at high induced voltage, low harmonic distortion as well as high generator efficiency, optimal generator parameters such as pole-arc to pole-pitch ratio and stator-slot-shoes dimension, etc. are determined with the proposed technique using Maxwell 2-D, Matlab software and the Taguchi method. The proposed double three-phase and six-phase winding configurations, which consist of six windings in the stator, can provide evenly distributed current for versatile applications regarding the voltage and current demands for practical consideration. Specifically, windings are connected in series to increase the output voltage at low wind speed, and in parallel during high wind speed to generate electricity even when either one winding fails, thereby enhancing the reliability as well. A PMSG is designed and implemented based on the proposed method. When the simulation is performed with a 6 Ω load, the output power for the double three-phase winding and six-phase winding are correspondingly 10.64 and 11.13 kW. In addition, 24 Ω load experiments show that the efficiencies of double three-phase winding and six-phase winding are 96.56% and 98.54%, respectively, verifying the proposed high performance operation.

Full text

Energies 2014, 7, 7105-7124; doi:10.3390/en7117105
energies
ISSN 1996-1073
www.mdpi.com/journal/energies
Article
Design of High Performance Permanent-Magnet Synchronous
Wind Generators
Chun-Yu Hsiao *, Sheng-Nian Yeh and Jonq-Chin Hwang
Department of Electrical Engineering, National Taiwan University of Science and Technology,
Taipei 10607, Taiwan; E-Mails: snyeh@mail.ntust.edu.tw (S.-N.Y.);
jchwang@ee.ntust.edu.tw (J.-C.H.)
* Author to whom correspondence should be addressed; E-Mail: D9607101@mail.ntust.edu.tw;
Tel.: +886-9-2113-5781; Fax: +886-2-2737-6699.
External Editor: Simon Watson
Received: 27 August 2014; in revised form: 21 October 2014 / Accepted: 22 October 2014 /
Published: 4 November 2014
Abstract: This paper is devoted to the analysis and design of high performance
permanent-magnet synchronous wind generators (PSWGs). A systematic and sequential
methodology for the design of PMSGs is proposed with a high performance wind generator
as a design model. Aiming at high induced voltage, low harmonic distortion as well as high
generator efficiency, optimal generator parameters such as pole-arc to pole-pitch ratio and
stator-slot-shoes dimension, etc. are determined with the proposed technique using Maxwell
2-D, Matlab software and the Taguchi method. The proposed double three-phase and
six-phase winding configurations, which consist of six windings in the stator, can provide
evenly distributed current for versatile applications regarding the voltage and current
demands for practical consideration. Specifically, windings are connected in series to
increase the output voltage at low wind speed, and in parallel during high wind speed to
generate electricity even when either one winding fails, thereby enhancing the reliability as
well. A PMSG is designed and implemented based on the proposed method. When the
simulation is performed with a 6 Ω load, the output power for the double three-phase winding
and six-phase winding are correspondingly 10.64 and 11.13 kW. In addition, 24 Ω load
experiments show that the efficiencies of double three-phase winding and six-phase winding
are 96.56% and 98.54%, respectively, verifying the proposed high performance operation.
OPEN ACCESS

Energies 2014, 7 7106
Keywords: high performance; permanent-magnet synchronous wing generator (PSWG);
Taguchi method
1. Introduction
The permanent-magnet synchronous generator (PMSG), which is less noisy, high efficiency and has
a long life span, has becomes one of the most important types of equipment in wind turbine systems.
In 2008, Bumby designed and fabricated a 5 kW, 150 rpm axial vertical permanent-magnet (PM) generator
driven directly by wind and a water turbine, where the generator uses trapezoidal shaped magnets to
enhance the magnetism over conventional circular magnets. The average efficiency is 94% under no core
losses and only limited eddy current loss [1]. In 2011, Maia used finite element method (FEM) software
to analyse the operation characteristics of an axial PM wind turbine with rated output power of 10 kW
while running at the speed of 250 rpm [2]. He [3] indicated that the electromagnetic properties of the
permanent-magnet machine are highly dependent on the number of slots per pole, phase, magnet shape,
the stator slots and the slot opening. Eriksson described that a winding scheme having slots per pole ratio
of 5/4 will reduce the cogging torque and suppress unwanted harmonics. The third harmonic distortion
measured is 5.8% under no-load and load of a 12 kW direct driven PM synchronous generator [4].
Erickson also designed a PM generator rated at 10 kW of telecommunication tower. The generator’s
electrical efficiency is 94.3% using a 2-D FEM model simulation [5]. From the literature to date, a
systematic procedure and methodology concerning the determination of optimal structural parameters
of PM wind generator leading to desired efficiency seems lacking. This motivated the study of this paper.
Axial flux PMSGs are widely used for vertical-axis wind turbines [6–10], however, since the axial
structure magnet is placed on the inner surface of the rotor without slot and facing the stator, it will
lengthen the distance between upper and lower magnets, which in turn requires much more magnet
material and cost to improve the operational efficiency.
This paper adopts the radial flux type PMSG, which has its stator windings placed in the inner core,
needs less magnet, and therefore resulting in cost reduction as well as heat loss elimination. In addition,
it applies vertical-axis wind turbine and direct drive structure and blade coupling generator. Therefore it
doesn’t need to speed up/slow down the gearbox to decrease the noise caused and reduce the machinery
wastage. It also presents a systematic approach for the design of permanent-magnet synchronous wind
generator. To exemplify the feasibility of the proposed methodology, analysis and design of a high
performance, multi-pole, gearless PMSG aiming at high efficiency and low voltage harmonics
simultaneously using less material and having lighter structure is given. In addition, a novel six stator
winding structure is also proposed to yield versatile applications. During low wind speed, the two
three-phase windings are connected in series to increase the output voltage, while at high wind speed,
the six windings are connected in parallel to provide more current and ensure power continuity when
either one of the two three-phase windings fails [11].
The paper is organized as follows: Section 2 is concerned with the dimension of rotor design;
Section 3 describes the use of Maxwell 2-D to determine the suitable ratio of pole-arc to pole-pitch, and
the use of the Taguchi method to find the stator-slot-shoes dimension; the designed parameters for high

Energies 2014, 7 7107
performance PMSG is also given in this section. Section 4 presents experimental results and their
comparison with Maxwell 2-D simulations; finally, conclusions are given in Section 5.
2. Design Dimension of the Rotor
Both the designed double three-phase and six-phase winding configurations of PMSG consist of six
composite windings with the advantage of high conductor utilization rate and lower torque ripple [11].
Meanwhile, it can disperse the armature current evenly under the same voltage and power conditions.
Figure 1 shows a simple and generally used surface mount with the basic geometric structure. It has a
complete magnetic flux circuit, half N-pole and half S-pole, the magnetic flux travels from the rotor surface
through the airgap, the magnetic silicon steel in the stator, the airgap, and then back to the rotor to form a
complete closed-loop.
Figure 1. Schematic diagram of permanent-magnet synchronous generator (PMSG):
magnetic flux path.
The equivalent magnetic circuit can be modeled as shown in Figure 2a. The stator yoke width should
be selected properly for reducing flux leakage and preventing magnetic saturation due to too small width
of the yoke, or over weight and dimension coming from thicker yoke [12].
Figure 2. A magnetic circuit model for the proposed structure: (a) complete magnetic circuit
model; (b) simplified magnetic circuit model.
(a)
(b)
The airgap flux can be written as ϕg = Kl·ϕ, where the leakage factor Kl is typically less than unity.
For rapid analysis of the magnetic circuit, leakage magnetic reluctance
l

is ignored as shown in
g
l
m
l
Stator
Airgap
Rotor

Energies 2014, 7 7108
Figure 2b. In addition, since the steel reluctance (
rs
 
) is small relative to the airgap reluctance
g

,
the steel reluctance can be eliminated by introducing a reluctance factor Kr having its value chosen to be
a constant slightly greater than unity to multiply the
g

to account for the neglected (
rs
 
). For the
machine with surface magnets under consideration, the leakage and reluctance factors are typically in
the ranges of 0.9–1.0 and 1.0–1.2, respectively, while the flux concentration factor is ideally 1.0.
The magnetic flux can be derived as [12]:
r m r m r
g
m g s r m r g
r
m
22
2 2 2 2 1
KK
    
    
          
(1)
Since the relationship between permeance coefficient (Pc) and airgap flux density is nonlinear,
doubling Pc does not double Bg. Doubling Pc, however, means doubling the magnet length, which
doubles its volume and associated cost accordingly. Using the relation:
mg
c
gm
lA
PlA

(2)
and:
m
m
r 0 m
l
A


,
g
g
0g
l
A


,
g
g
g
BA


,
r
r
m
BA


(3)
Equation (1) becomes:
rm
gg
r
r
c
1
BA
BA
KP


(4)
Assuming that all the magnetic fluxes leaving the magnet through an airgap go into the stator core, then:
gm
AA
(5)
Since, as indicated above, Kr is slightly greater than unity, it is further assumed that Kr = 1. Thus,
substituting Equation (5) into Equation (4) results in:
g
g
r
r
m
1
1
B
l
B
l


(6)
Determination of the airgap length lg depends on the gap magnetic flux density and the processing of
machine structure. If the airgap length is too short, it will cause serious eccentric force at high speed. Wider
airgap length, however, will reduce the gap magnetic flux density and lower efficiency. The optimal ratio
between magnet thickness and the airgap is usually selected in the range of 4–6 as shown in Figure 3 [12].
Usually generator designer determines magnet thickness in accordance with this search range. Meanwhile,
production and installation tolerances must be considered to decide the eventual airgap length in order

Energies 2014, 7 7109
to avoid motor assembly complexity and operating problems. Equation (6) will be used to decide the
initial airgap length shown in Table 2 of Section 3.
Figure 3. Relationship between normalized airgap flux density and permeance coefficient.
Ignoring the magnetic effects caused by the stator teeth, the distribution of airgap flux density can be
illustrated by Figure 4. The ratio αp-p [13] between the width of the magnet and the pole-pitch of rotor core,
written as the ratio of pole-arc αarc to pole-pitch αpitch, is defined in Equation (7), where αarc and αpitch are
the angular span of any single magnet and that between the center lines of any two adjacent magnetic
poles, respectively. It is related to flux density. Specifically, the greater ratio, i.e., the longer magnet arc
length, will result in higher flux density. If the gap magnetic flux density waveform is closer to sinusoidal,
then the induced voltage harmonics content will be smaller:
arc
p-p
pitch



(7)
when αp-p is unity, the N and S poles of the magnet are consecutive, i.e., without a gap in between,
the gap flux density is a square wave. In general, for 0 ≤ αp-p ≤ 1, Fourier series expansion of the flux
density at any electrical degree θe in the airgap can be derived as:
h
h
g,peak p-p
g h h e
1h
21
1 ( 1) cos[ ( ) 180 ]sin
2
k
k
B
B k k
k




   

(8)
where Bg,peak is the maximum flux density of airgap and kh is the k-th harmonic.
Figure 4. Relationship between αp-p and airgap flux density: (a) αp-p = 1; (b) αp-p = 0.5.
(a)
(b)
m
g
l
l
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
g
r
B
B
10 2 3 4 5 6 7 8 9 10
N
N S S
ge
()B
e

0
ge
()B
0
e

N NS S
arc

pitch


Energies 2014, 7 7110
Equation (8) yields the k-th harmonic flux ratio:
h
h
p-p
h
h
1
1 ( 1) cos[ ( ) 180 ]
2
k
k
Bk
k


  
(9)
It is seen from Equation (9) that the airgap flux density ratio of harmonic is determined by different
αp-p, as shown in Table 1. For balanced three-phase, the third harmonic can be eliminated by using
Y-connected wiring. The harmonic distortion is not proportional to the phase and line voltages, but
depends on the third harmonic of the phase voltage. Even if the third harmonic of the phase voltage will
not appear in the line voltage, it will cause losses within each phase winding. It is obvious from Table 1
that αp-p = 0.800 will yield the smallest 3rd and smaller 5th harmonics. This choice of αp-p for the smallest
5th harmonic will further be confirmed by the finite element analysis given in Section 3.
Table 1. Relationship between αp-p and airgap flux density ratio.
αp-p
h1k
B
Harmonic content (
h h p-p
1, α1
/
kk
BB
  
)
1st (pu)
3rd (pu)
5th (pu)
7th (pu)
11th (pu)
0.667
1.7322
0.8661
0.0000
0.1732
0.1237
0.0788
0.800
1.9022
0.9511
0.1960
0.0000
0.0840
0.0865
0.857
1.9498
0.9749
0.2605
0.0866
0.0000
0.0712
0.909
1.9796
0.9898
0.3032
0.1511
0.0771
0.0000
1.000
2.0000
1.0000
0.3333
0.2000
0.1429
0.0909
3. Simulation Results and Discussion
The finite element method from the Maxwell-2D software will be used throughout the analysis and
design in this section. The relevant specifications and dimensions of the generator under design are listed
in Table 2 with the sketched schema graph of PMSG and magnet dimensions given in Figure 5.
As will be illustrated in the following, to design PMSG for double three-phase and six-phase winding
connections, αp-p will first be determined to obtain the greatest induced voltage and its smallest
harmonic content.
Figure 5. Schema graph of PMSG and magnet dimension: (a) structure of a 78-pole, 72-slot
PMSG; (b) magnet dimension.
(a)
(b)
so
R
ri
R
ro
R
si
R
Stator
Magnet
g
l
m
l
m
h
gl
l

Energies 2014, 7 7111
The optimal size of the stator shoes is then decided by using the Taguchi method. Moreover, the
induced voltage and current values can be determined by using different loads at the specific rotor speed
and then calculate power output. Maxwell_2D and Fourier transform are applied to analyze PSMG-based
electrical characteristics.
Table 2. Parameters of wind generator to be designed.
Parameter
Value
Pole number
78
Slot number
72
Number of phase
double three-phase and six-phase
Rated speed (rpm)
90
Rated output power (W)
10,000
Rated torque (N·m)
1,061
Induced voltage (V)
150
Estimated efficiency (%)
90
Winding turns per slot (turns)
25
Winding conductor diameter (m)
(0.0012) × 4
Stator radius: inner/outer (m) Rsi, Rso
0.3065/0.3550
Rotor radius: inner/outer (m) Rri, Rro
0.2820/0.3045
Airgap length (m) lg
0.002
Magnet thickness (m) lm
0.008
Stack length (m)
0.12
Core material/Permanent magnet
50H400/NdFeB42H
End airgap length of rotor magnet (m) lgl
0.0026
3.1. Optimal Sizing of Rotor Magnet
Aiming at high induced voltage and low harmonic distortion, sizing of rotor magnet will be conducted
by the best pole-arc to pole-pitch ratio, αp-p. Fixing the internal and external diameters of stator and rotor
as given in Table 2, double three-phase winding are used.
Assuming open stator slot as above, finite element analyses using Maxwell 2-D software for five
different αp-p, i.e., αp-p = 0.667, 0.800, 0.857, 0.909, and 1.000 are given. Figure 6 shows the various phase
voltage Vp and line voltage Vl values obtained by Maxwell 2-D with different αp-p. When αp-p = 0.800,
the values for Vp and Vl are 243.3 V and 458.4 V, respectively.
Figure 6. Induced voltages of PMSG with different αp-p by Maxwell 2-D: (a) Phase voltage;
(b) Line voltage.
(a)
(b)
00.005 0.01 0.015 0.02 0.025 0.03
-250
-200
-150
-100
-50
0
50
100
150
200
250
0.035
Time (s)
Vp (V)
= 0.6667
p-p

= 0.8
= 0.857
= 0.909
= 1
0.025 0.027 0.029
230
240
250
(A)
Time (s)
Vp (V)
(A)
p-p

p-p

p-p

p-p

0.005 0.01 0.015 0.02 0.025 0.03
-400
-300
-200
-100
0
100
200
300
400
-500
500
0 0.035
Time (s)
Vl (V)
0.0235 0.025 0.0265 0.028
360
420
480
(B)
Time (s)
Vl (V)
(B)
= 0.6667
= 0.8
= 0.857
= 0.909
= 1
p-p

p-p

p-p

p-p

p-p


Energies 2014, 7 7112
The results shown in Table 3 reveal that αp-p = 0.800 will yield the smallest 5th harmonic voltage
(0.14/0.13) with sufficiently high induced voltage. This agrees with Table 1 that αp-p = 0.800 will yield the
smallest 5th and smaller 3rd harmonics. Besides, as mentioned in Section 2 that for balanced three-phase,
the third harmonic voltage can be eliminated by using Y-connected wiring.
Table 3. Induced voltages and total harmonic distortion (THD) of 72-slot, 78-pole PMSG
with different αp-p.
αp-p
Vp/Vl (V)
THD (%)
Each order harmonic (Phase/Line)
3rd
5th
7th
9th
11th
0.667
237.7/421.7
1.17/1.18
0.07/0.02
1.11/1.11
0.17/0.22
0.05/0.06
0.01/0.02
0.800
243.3/458.4
7.28/0.41
7.27/0.01
0.14/0.13
0.15/0.19
0.09/0.03
0.02/0.01
0.857
244.8/465.7
9.26/0.54
9.24/0.05
0.25/0.29
0.08/0.07
0.06/0.07
0.03/0.02
0.909
245.6/470.7
10.38/0.67
10.36/0.03
0.46/0.48
0.03/0.03
0.09/0.09
0.12/0.08
1.000
246.7/472.2
11.00/0.72
10.98/0.03
0.60/0.63
0.11/0.07
0.50/0.06
0.05/0.02
3.2. Determining the Optimal Stator-Slot-Shoes Dimension Using Taguchi Method
Conventionally, to find the most appropriate parameters for the optimum design, analyses or tests are
usually conducted by changing one factor at a time for each experiment while other factors are kept fixed.
The basic principle of Taguchi method is, however, to choose several control factors and use an
orthogonal array to get useful statistic information with the fewest experiments. Taguchi method has
been widely used in motor design since it provides engineers with a systematic and efficient method to
administer numerical experiments and acquire the optimal parameters quickly [14–20].
Consider the block diagram shown in Figure 7, where system represents a product or manufacturing
process, such as PMSG in our case here, y refers to the quality characteristics or response, signal factors
(M), controllable factors (Z) and noise factors (X) are factors that can affect y [16].
Figure 7. Product/process parameters.
The effect of many different parameters or control factors on the performance characteristic in a
condensed set of experiments can be examined by using the orthogonal array experimental design proposed
by Taguchi [14,15]. Knowing the number of parameters and the number of levels, the proper orthogonal
array can be selected as shown in Table 4, which is adopted in this paper with the slot opening length,
the height of the shoe portion, the magnet length, and the tooth width chosen as the control factors.
System
Signal factors
(M)
Quality
characteristics (y)
Control factors (Z)
Noise factors (X)

Energies 2014, 7 7113
Table 4.
5
16 (4 )L
Array.
()
c
a
Lb
The number of level
The number of control factor
L is the first word by Latin Square
The number of experiments
Experiment
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
A
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
B
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
C
1
2
3
4
2
1
4
3
3
4
1
2
4
3
2
1
D
1
2
3
4
3
4
1
2
4
3
2
1
2
1
4
3
E
1
2
3
4
4
3
2
1
2
1
4
3
3
4
1
2
One of the key features of Taguchi method is the use of signal-to-noise, S/N, ratio to transform the
performance characteristic in the optimization process. The signal represents the mean performance while
the noise signifies the variance. No matter what the applications, the method of measurement, or the units
in which the results are expressed, there can be three types of S/N with the criteria: bigger is better,
smaller is better and nominal is the best.
Taguchi strongly recommends the use of S/N ratio, expressed as a log transformation of the mean-squared
deviation, as the yardstick for analysis of experimental results. i.e.,
);;();( ZMXeZMgy 
(10)
where the signal g(M; Z) and the noise e(X; M; Z) are usually in the predictable and unpredictable sections,
respectively. The objective of the design is to maximize the predictable and minimize the unpredictable
parts, correspondingly.
For the-larger-the-better (LTB) target, the S/N ratio is:
2
1
11
/ 10log ( )
n
ii
SN ny


  


(11)
For the-smaller-the-better (STB) quality characteristic, S/N ratio is defined as:
2
1
1
/ 10log ( )
n
i
i
S N y
n

  


(12)
The performance index follows LTB for this paper, it can be expressed as:
22
B
2
p
1
/ 10 log( ) -10 log( ) -10 log( )S N THD T
V

(13)
The first mathematical term can be considered the induced voltage which in value is relatively higher;
the next two terms respectively represent the total harmonic distortion (THD) and flux density of tooth
width, both having a smaller value.
As mentioned above, four control factors, namely that slot opening length, external height of shoes
(Shoes’_ho), internal height of shoes (Shoes’ hi) and tooth width are chosen in spite of the fifth control
factor [21,22]. Moreover, four levels are applied for each factor to maximize the induced voltage of
generator and keep the flux density of stator teeth lying within the saturation point of their laminations.
Details are given in the following section.
In this section, the Taguchi method will be used to determine the dimension of stator shoes. Consider
Figure 8. Four parameters, slot-opening length, ho, hi, and tooth width are chosen as control factors,
which are subdivided into four levels to yield four row data shown in Table 5 for the orthogonal array

Energies 2014, 7 7114
L16(45) to process 16 simulations with Maxwell-2D and Matlab software. The results for no-load analysis
of double three-phase winding configuration are summarized in Table 6 and Figure 9.
Figure 8. Slot dimensions.
Table 5. Four control factors and levels of stator shoes.
Level
Control factor
A
B
C
D
Slot opening length (mm)
Shoes’_ho (mm)
Shoes’_hi (mm)
Tooth width (mm)
1
1
0
1
7
2
2
1
2.5
9
3
3
2
4
11
4
4
3
5.5
13
Table 6. Performance statistics for induced voltage, THD and flux density of tooth width.
Optimized value: Vp(V): 208, THD(%): 0.39, TB: 1.2655, S/N: 52.49.
No.
1
2
3
4
5
6
7
8
Vp (V)
149
163
191
207
179
185
158
186
THD (%)
1.58
2.69
1.71
0.65
1.84
2.85
3.83
2.18
TB (Tesla)
1.6489
1.5282
1.3050
1.2582
1.3491
1.2148
1.6823
1.5833
S/N
35.15
31.97
38.65
48.07
37.16
34.56
27.79
34.63
No.
9
10
11
12
13
14
15
16
Vp (V)
191
192
171
160
174
154
201
197
THD (%)
1.86
1.01
1.29
3.53
1.42
1.99
0.97
1.24
TB (Tesla)
1.2257
1.1457
1.5825
1.6956
1.5848
1.7199
1.2539
1.4692
S/N
38.46
44.40
38.46
28.54
37.77
33.06
44.36
40.68
Table 6 indicates that the best S/N ratio among the 16 simulations is obtained during the 4th run
(S/N = 48.07), with THD and TB equal to 0.65 and 1.2582, respectively. Besides, It is seen from Figure 9
that the dimensions of slot-opening length, Shoes’_ho, hi, and tooth width for the optimized S/N ratio are
4, 3, 5.5 and 13 mm, respectively; the corresponding optimized values are Vp (V): 208, THD (%): 0.39,
TB: 1.2655 and S/N: 52.49. When 50H400 silicon steel sheet is used with maximum saturation point,
14 mm
Slot opening length
Shoes_ho
Shoes_hi
Tooth width
Open slot
,,
Shoes_hi
Tooth width
Slot opening length
14 mm
open slot
Shoes_ho
，，

Energies 2014, 7 7115
1.6300 Tesla, as a limit, the optimized value from Figure 10 (1.2655 Tesla) still does not exceed the
threshold from simulations to avoid producing the problems of high temperature and the
demagnetization phenomenon while the flux density of tooth width is near 1.6300 Tesla.
Figure 9. Taguchi analysis for different control factors.
Figure 10. Flux plot of the optimum stator-slot-shoes dimension by Maxwell-2D.
1.6300e+000
1.5621e+000
1.4942e+000
1.4263e+000
1.3583e+000
1.2904e+000
1.2225e+000
1.1546e+000
1.0867e+000
1.0188e+000
9.5083e - 001
8.8292e - 001
8.1500e - 001
7.4708e - 001
6.7917e - 001
6.1125e - 001
5.4333e - 001
4.7542e - 001
4.0750e - 001
3.3958e - 001
2.7167e - 001
2.0375e - 001
1.3583e - 001
6.7917e - 002
0.0000e+000
B [T]
38.46
33.53
37.47
38.97
37.13
36.00
37.3237.98
37.21
35.51 36.20
39.51
31.14
35.71
40.22
41.36
30
32
34
36
38
40
42
A1
(1mm)
A2
(2mm)
A3
(3mm)
A4
(4mm)
B1
(0mm)
B2
(1mm)
B3
(2mm)
B4
(3mm)
C1
(1mm)
C2
(2.5mm)
C3
(4mm)
C4
(5.5mm)
D1
(7mm)
D2
(9mm)
D3
(11mm)
D4
(13mm)
Slot opening length Shoes’_ho Shoes’_hi Tooth width
Average Effects
Control Factors
Taguchi analysis
Slot opening length
Shoes’_ho
Shoes’_hi
Tooth width

Energies 2014, 7 7116
3.3. Finite Element Analysis of the Optimum Case
The electrical characteristics as mentioned refers to the analysis for the induced voltage, which may be
further divided into the no-load and load tests. At different rotor speeds the corresponding induced voltage
varies, and the rotor speed is proportional to the induced voltage. Furthermore, the induced voltage and
current, as well as its output power can be determined at the specific rotor speed by using different loads.
In this section, Maxwell-2D is used to analyze PSMG-based electrical characteristics. A Fourier transform
with the results by using Matlab is then performed to obtain THD.
Based on the optimal parameters obtained above, finite element analysis with Maxwell-2D for 72-slot,
78-pole PMSG under seven different loads is conducted. The rotor is divided into two cylindrical
segments of equal length, and adjacent segments are rotated on their axis by a constant angle of the
generator. The results for induced voltages and THD of double three and six phase winding
configurations shown in Table 7 indicate that THD of both induced voltage and current are less than
1.4%, complying with IEEE Standard 519.
Table 7. Simulated results by Maxwell-2D.
Parameter
Vp (V)/THD (%)
Vl (V)/THD (%)
Ip (A)/THD (%)
Po (W)
Double
three-phase
winding
No load
208
0.39%
361
0.39%
-
-
Load 48 Ω
205
0.31%
355
0.31%
4.3
0.33%
2644.50
Load 24 Ω
199
0.31%
345
0.31%
8.3
0.31%
4955.10
Load 12 Ω
182
0.33%
316
0.33%
15.2
0.33%
8299.20
Load 6 Ω
146
0.38%
252
0.38%
24.3
0.34%
10643.40
Load 3 Ω
95
0.82%
164
0.86%
31.6
0.28%
9006.00
Six-phase
winding.
No load
215
1.41%
375
1.39%
-
-
Load 48 Ω
212
1.20%
355
0.31%
4.4
1.24%
2798.40
Load 24 Ω
206
1.08%
356
1.05%
8.6
1.08%
5314.80
Load 12 Ω
188
0.95%
336
0.90%
15.7
0.95%
8854.80
Load 6 Ω
149
0.63%
258
0.64%
24.9
0.63%
11130.30
Load 3 Ω
97
0.36%
167
0.84%
32.0
0.36%
9312.00
Ip: current per phase.
Figure 11 shows the induced voltages for the electrical angle equaling to 0° between Va1 and Va2 for
the double three-phase winding; whereas the phase difference between Va and Vx phases is 30° for the
six-phase winding. In addition, Table 7 shows that Vp are equal to 208 V and 215 V, respectively,
for double three-phase winding and six-phase winding for no-load. This can also be verified from the
near-sinusoidal voltage and current waveforms for no-load case from Figure 11. When the speed is 90 rpm,
from Table 7, the 6 Ω load receives generator outputs of 10.643 kW and 11.130 kW for the double
three-phase winding and six-phase winding, respectively.

Energies 2014, 7 7117
Figure 11. Analysis of induced voltage at no-load: (a) double three-phase winding;
(b) six-phase winding.
(a)
(b)
From the frequency domain analysis for Va1, Figure 12 shows that the peak values for Va1 under 24 Ω
load for the double three-phase winding and six-phase winding are 159.24 V and 160.28 V, respectively,
with the corresponding peak values of 7.05 A and 6.43 A for Ia1.
Figure 12. Induced voltage and current at 0, 6 and 24 Ω-load: (a) double three-phase winding;
(b) six-phase winding.
(a)
(b)
4. Experimental Results and Discussion
This section presents the experimental results conducted on the proposed high performance PMSG.
The generator and its measurement platform are built and shown in Figure 13, which contains a 6-pole,
50 horsepower (50 hp = 37.3 kW) induction motor as the prime mover.
00.005 0.01 0.015 0.02 0.025 0.03 0.035
-250
-200
-150
-100
-50
0
50
100
150
200
250
Time(s)
Voltage(V)
Va1
Vb1
Vc1
Va2
Vb2
Vc2
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
-250
-200
-150
-100
-50
0
50
100
150
200
250
Time(s)
Voltage(V)
Va
Vb
Vc
Vx
Vy
Vz
0 20 40 60 80 100 120 140 160 180 200
0
20
40
60
80
100
120
140
160
180
Freq. (Hz)
Va1 (V)
Load : 0 Ω load
Load : 24 Ω load
Load : 6 Ω load
Va1, 6 Ω load peak (120.36 V)
Va1, 24 Ω load peak (159.24 V)
Va1, 0 Ω load peak (166.6 V)
0 20 40 60 80 100 120 140 160 180 200
0
5
10
15
20
25
Freq. (Hz)
Ia1 (A)
Ia1, 24 Ω load peak (7.05 A)
Load : 0 Ω load
Load : 24 Ω load
Load : 6 Ω load
Ia1, 6 Ω load peak (21.18 A)
Ia1, 0 Ω load
020 40 60 80 100 120 140 160 180 200
0
20
40
60
80
100
120
140
160
180
Load : 0 Ω load
Load : 24 Ω load
Load : 6 Ω load
Freq. (Hz)
Va1 (V)
Va1, 24 Ω load peak (160.48 V)
Va1, 0 Ω load peak (169.45 V)
Va1, 6 Ω load peak (119.35 V)
0 20 40 60 80 100 120 140 160 180 200
0
5
10
15
20
25
Load : 0 Ω load
Load : 24 Ω load
Load : 6 Ω load
Ia1, 6 Ω load peak (20.06 A)
Ia1, 24 Ω load peak (6.4 A)
Ia1, 0 Ω load
Freq. (Hz)
Ia1 (A)

Energies 2014, 7 7118
Figure 13. Experimental apparatus: (a) schematic diagram; (b) platform photo.
(a)
(b)
It is seen from Figure 13 that the 72-slot, 78-pole PMSG under test is driven via a reducer with speed
ratio of 6.23:1. The wiring of double three-phase and six-phase winding configurations can easily be
accomplished through the proposed scheme shown in Figure 14. Calculation of powers and efficiencies
using simulated results will be given to compare with the experimental evaluation in this section.
Figure 14. External wiring photo.
The measured and simulated induced voltages of the PMSG from no-load test for both winding
configurations under different speeds are given in Figure 15, where linear relationships between the
induced voltage and speed are observed. Besides, Figure 15 also reveals that close agreements for
induced voltages between measured and simulated results are obtained. In fact, the differences for both
wirings are less than 1%.
50 HP induction motor 6.23:1 gear box PMSG
Encoder
Load cell Z-axis lifting
mechanism
50 HP induction motor 6.23:1 gear box PMSG
Encoder

Energies 2014, 7 7119
Figure 15. Measured and simulated no-load induced voltages: (a) double three-phase winding;
(b) six-phase winding.
(a)
(b)
Since the maximum output power of the load box is 5 kW when the generator is operated at rated
speed of 90 rpm, the experimental results can only be obtained and compared with simulations for the
resistive load of 24 Ω. The experimental results under no-load and 24 Ω load shown in Table 8 indicate
that THD of the induced line voltages are within 3% for both winding configurations, complying with
IEEE Standard 519 as in the simulation case. The corresponding waveforms of the induced voltages
given in Figure 16 show near-sinusoidal outputs for no-load case.
Table 8. Measured induced voltages and THD.
Parameter
Vp (V)/THD (%)
Vl (V)/THD (%)
Ip (A)/THD (%)
Po (W)
Double three-phase
winding
No load
209
0.79%
359
0.59%
-
-
Load 24 Ω
200
2.82%
344
0.75%
8.1
1.40%
4860
Six-phase winding
No load
217
2.35%
376
2.34%
-
-
Load 24 Ω
204
3.94%
360
1.92%
8.3
3.45%
5080
Figure 16. Measured no-load induced voltage: (a) double three-phase winding;
(b) six-phase winding.
(a)
(b)
From the frequency domain analysis for Va1, Figure 17 shows that the differences for the peak values
of Va1 between simulation and experimental results are 7.84 V and 5.94 V for double three-phase winding
and six-phase winding, respectively, under 24 Ω load. The corresponding differences for the peak values
of Ia1 between simulation and experimental results are 0.73 A and 0.50 A.
20
40
60
80
100
120
140
160
180
200
220
Measured
Simulated
speed (rpm)
voltage (V)
10 20 30 40 50 60 70 80 90
20
40
60
80
100
120
140
160
180
200
220
speed (rpm)
voltage (V)
10 20 30 40 50 60 70 80 90
Measured
Simulated
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
-250
-200
-150
-100
-50
0
50
100
150
200
250
Time(s)
Voltage(V)
Va1
Vb1
Vc1
Va2
Vb2
Vc2
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
-250
-200
-150
-100
-50
0
50
100
150
200
250
Time(s)
Voltage(V)
Va
Vb
Vc
Vx
Vy
Vz

Energies 2014, 7 7120
Figure 17. Measured induced voltage and current with 24 Ω load: (a) double three-phase
winding; (b) six-phase winding.
(a)
(b)
Finally, a load cell installed at the bottom of the generator will yield torques which, in turn, is used
to calculate the input and output powers of the generator for the two winding connections. The results
are shown in Table 9, which indicates that the output powers of 5.029 and 5.433 kW at 90 rpm for 24 Ω
load are obtained with efficiencies of 96.56% and 98.54%, respectively, for double three-phase and
six-phase winding.
Table 9. Calculated efficiencies from measured results (Prime mover measured).
Vl,rms: the rms of line induced voltage; Il,rms: the rms of line current; ωm: speed; T: torque;
Po: output power; Pi: input power; η: efficiency.
Vl,rms (V)
Il,rms (A)
ωm (rpm)
T (kg·m)
Po (W)
Pi (W)
η (%)
Double
three-phase
winding
110
2.7
40
27.10
1028.84
1112.46
92.48
136
3.4
50
33.30
1601.80
1708.71
93.74
164
4.0
60
39.00
2272.45
2401.43
94.63
190
4.7
70
45.20
3093.44
3247.07
95.27
217
5.3
80
50.60
3984.06
4154.27
95.90
242
6.0
90
56.40
5029.88
5209.26
96.56
Six-phase
winding
115
2.8
40
28.00
1115.44
1149.40
97.05
144
3.5
50
34.80
1745.91
1785.68
97.77
171
4.2
60
41.20
2487.92
2536.90
98.07
198
4.9
70
47.60
3360.87
3419.48
98.29
226
5.5
80
53.30
4305.88
4375.95
98.40
253
6.2
90
59.70
5433.79
5514.06
98.54
020 40 60 80 100 120 140 160 180 200
0
20
40
60
80
100
120
140
160
180
Freq. (Hz)
Va1 (V)
Load : 0 Ω load - SIM
Load : 24 Ω load - SIM
Load : 0 Ω load -EXP
Va1, 24 Ω load peak (159.24 V) - SIM
Va1, 0 Ω load peak (161.3 V) - EXP
Va1, 0 Ω load peak (166.6 V) - SIM
Va1, 24 Ω load peak (151.4 V) - EXP
Load : 24 Ω load - EXP
0 20 40 60 80 100 120 140 160 180 200
0
1
2
3
4
5
6
7
8
Freq. (Hz)
Ia1 (A)
Ia1, 24 Ω load peak - SIM (7.05 A)
Ia1, 24 Ω load peak - EXP (6.32 A)
Ia1, 0 Ω load - SIM
Ia1, 0 Ω load - EXP
Load : 0 Ω load - SIM
Load : 24 Ω load - EXP
Load : 0 Ω load - EXP
Load : 24 Ω load - SIM
020 40 60 80 100 120 140 160 180 200
0
20
40
60
80
100
120
140
160
180
Freq. (Hz)
Va1 (V)
Va1, 24 Ω load peak - SIM (160.48 V)
Va1, 0 Ω load peak - SIM (169.45 V)
Load : 0 Ω load - SIM
Load : 24 Ω load - SIM
Load : 0 Ω load -EXP
Load : 24 Ω load - EXP
Va1, 0 Ω load peak - EXP (165.42 V)
Va1, 24 Ω load peak - EXP (154.54 V)
Freq. (Hz)
0 20 40 60 80 100 120 140 160 180 200
0
Ia1, 24 Ω load peak - SIM (6.4 A)
Ia1, 0 Ω load - SIM
Ia1 (A)
2
4
6
8
Load : 0 Ω load - SIM
Load : 24 Ω load - EXP
Load : 0 Ω load - EXP
Load : 24 Ω load - SIM
Ia1, 0 Ω load - EXP
Ia1, 24 Ω load peak - EXP (5.9 A)

Energies 2014, 7 7121
Table 10 compares the aforementioned experimental result from 24 Ω-load with that from Maxwell-2D
simulation. It is seen from Table 10 that close agreement between measured and simulated results are also
obtained concerning the induced voltages, load currents and output powers.
Table 10. Comprehensive comparison.
Winding
connection
Maxwell-2D software
Oscilloscope measured
Prime mover measured
No Load
Load 24 Ω
No Load
Load 24 Ω
No Load
Load 24 Ω
Vp
(V)
Ip
(A)
Vp
(V)
Ip
(A)
Po
(W)
Vp
(V)
Ip
(A)
Vp
(V)
Ip
(A)
Po
(W)
Vp
(V)
Ip
(A)
Vp
(V)
Ip
(A)
Po
(W)
Double
three-phase
winding
208
0.0
199
8.3
4955
209
0.0
200
8.1
4860
212
0.0
198
8.5
5030
Six-phase
winding
215
0.0
204
8.5
5315
217
0.0
204
8.3
5080
222
0.0
207
8.8
5434
5. Conclusions
This paper presents a systematic and sequential methodology for the design of PMSGs. With the
selected αp−p = 0.800, zero 3rd and smaller 5th harmonics can be obtained by using Y-connected wiring.
As for the Taguchi method, an L16 orthogonal array is applied, and 16 simulations and analyses are
performed to determine the appropriate control parameters. These include slot-opening length, ho, hi, and
tooth width. The selected S/N ratio is based on increasing the induced voltage and having smaller values
of THD and TB. Under the no-load test, the induced voltage differences between experiment and simulation
for both double three-phase winding and six-phase winding are less than 1%. This verifies the reliability
of Maxwell 2-D analysis. In addition, 24 Ω load experiments show the output powers of 5.029 kW and
5.433 kW at 90 rpm for double three-phase winding and six-phase winding, respectively. From the results
in this study, it can be seen that the PMSG design clearly achieved the proposed high performance features.
Author Contributions
The manuscript is a part of the Ph.D. dissertation of Chun-Yu Hsiao, who had been a Ph.D. student
from September 2007 to June 2012, and jointly supervised by Sheng-Nian Yeh, and Jonq-Chin Hwang,
Department of Electrical Engineering, National Taiwan University of Science and Technology.
The design and analyses of 10 kW PMSG, using FEM and Taguchi method, as well as the experiments
were carried out by Hsiao under the supervision and support of both advisors. In addition, the manuscript
was first drafted by Hsiao and then revised by Yeh.
Notation
Am
magnet cross-sectional area
Ag
airgap cross-sectional area
Bg
airgap flux density
Br
remanence
Bk
k-th harmonic flux ratio

Energies 2014, 7 7122
Ip
phase current
Il,rms
rms of line current
Kl
leakage factor
kh
the k-th harmonic
Kr
reluctance factor
lm
magnet thickness
lg
airgap length
Pc
permeance coefficient
Po
output power
Pi
Input power
l

leakage magnetic reluctance
r

rotor steel reluctance
s

stator steel reluctance
g

airgap reluctance
m

magnet reluctance
Rsi
Stator inside radius
Rso
Stator outside radius
Rri
Rotor inside radius
Rro
Rotor outside radius
T
torque
Vp
phase induced voltage
Vl
line induced voltage
Vl,rms
rms of line induced voltage
ϕg
airgap flux
ϕr
Magnet remanent flux
ϕl
Leakage flux
µ
r
relative permeability
µ
0
permeability of free space
αp-p
The ratio between the width of the magnet and the pole-pitch of rotor core
αarc
pole-arc
αpitch
pole-pitch
θe
electrical degree
ωm
speed
η
efficiency
Conflicts of Interest
The authors declare no conflicts of interest.

Energies 2014, 7 7123
References
1. Bumby, J.R.; Stannard, N.; Dominy, J.; McLeod, N. A permanent magnet generator for small scale
wind and water turbines. In Proceedings of the International Conference on Electrical Machines,
Vilamoura, Portugal, 6–9 September 2008; pp. 1–6.
2. Maia, T.A.C.; Faria, O.A.; Cardoso, A.A.R.F.E.; Borges, F.S.; Mendonca, H.G.; Silva, M.A.;
Vasconcelos, J.A.; Silva, S.R.; Lopes, B.M. Electromechanical design for an optimized axial flux
permanent magnet torus machine for 10 kW wind turbine. In Proceedings of the International
Conference on Electrical Machines and Systems, Beijing, China, 20–23 August 2011; pp. 1–6.
3. He, Q.; Wang, Q. Optimal design of low-speed permanent magnet generator for wind turbine
application. In Proceedings of the Asia-Pacific Power and Energy Engineering Conference, Shanghai,
China, 27–29 March 2012; pp. 1–3.
4. Eriksson, S.; Solum, A.; Leijon, M.; Bernhoff, H. Simulations and experiments on a 12 kW direct
driven PM synchronous generator for wind power. Renew. Energy 2008, 33, 674–681.
5. Eriksson, S.; Bernhoff, H.; Bergkvist, M. Design of a unique direct driven PM generator adapted
for a telecom tower wind turbine. Renew. Energy 2012, 44, 453–456.
6. Chalmers, B.J.; Wu, W.; Spooner, E. An axial-flux permanent-magnet generator for a gearless wind
energy system. IEEE Trans. Energy Conv. 1999, 14, 251–257.
7. Sadeghierad, M.; Darabi, A.; Lesani, H.; Monsef, H. Rotor yoke thickness of coreless high-speed
axial-flux permanent magnet generator. IEEE Trans. Magn. 2009, 45, 2032–2037.
8. Chaar, L.E.; Lamont, L.A.; Elzein, N. Wind energy technology-industrial update. In Proceedings of
the IEEE Power and Energy Society General Meeting, Detroit, MI, USA, 24–29 July 2011; pp. 1–5.
9. Chan, T.F.; Lai, L.L.; Xie, S. Field computation for an axial flux permanent-magnet synchronous
generator. IEEE Trans. Energy Convers. 2009, 24, 1–11.
10. Caricchi, F.; Maradei, F.; de Donato, G.; Capponi, F.G. Axial-flux permanent magnet generator for
induction heating gensets. IEEE Trans. Ind. Electron. 2009, 57, 128–137.
11. Hwang, J.C.; Hsiao, C.Y.; Yeh, S.N. Composite Permanent Magnet Synchronous Machine.
U.S. Patent No. 13/452,465, 20 April 2014.
12. Hanselman, D. Brushless Permanent Magnet Motor Design; Magna Physics Pub.: Orono, ME,
USA, 2003.
13. Zhu, Z.Q.; Howe, D. Influence of design parameters on cogging torque in permanent magnet machines.
IEEE Trans. Energy Convers. 2000, 15, 407–412.
14. Phadke, M.S. Quality Engineering Using Robust Design; Prentice Hall: Upper Saddle River, NJ,
USA, 1995.
15. Ross, P.J. Taguchi Techniques for Quality Engineering; McGraw-Hill Book Company: New York,
NY, USA, 1998.
16. Lee, H.H. Taguchi Methods: Principles and Practices of Quality Design; Gau Lih Book Company
Ltd.: Taipei, Taiwan, 2011.
17. Kim, W.H.; Kim, K.C.; Kim, S.J.; Kang, D.W.; Go, S.C.; Lee, H.W.; Chun, Y.D.; Lee, J. A study
on the optimal rotor design of LSPM considering the starting torque and efficiency. IEEE Trans. Magn.
2009, 45, 1808–1811.

Energies 2014, 7 7124
18. Awada, A.; Wegmann, B.; Viering, I.; Klein, A. Optimizing the radio network parameters of the long
term evolution system using Taguchi’s method. IEEE Trans. Veh. Technol. 2011, 60, 3825–3839.
19. Hwang, C.C.; Li, P.L.; Liu, C.T. Optimal design of a permanent magnet linear synchronous motor
with low cogging force. IEEE Trans. Magn. 2012, 48, 1039–1042.
20. Shi, T.; Qiao, Z.; Xia, C.; Li, H.; Song, Z. Modeling, analyzing, and parameter design of the
magnetic field of a segmented Halbach cylinder. IEEE Trans. Magn. 2012, 48, 1890–1898.
21. Gologlu, C.; Sakarya, N. The effects of cutter path strategies on surface roughness of pocket milling
of 1.2738 steel based on Taguchi method. J. Mater. Process. Technol. 2008, 206, 7–15.
22. Esme, U.; Bayramoglu, M.; Kazancoglu, Y.; Ozgun, S. Optimization of weld bead geometry in TIG
welding process using grey relation. Int. J. Eng. Res. Appl. 2009, 43, 143–149.
© 2014 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article
distributed under the terms and conditions of the Creative Commons Attribution license
(http://creativecommons.org/licenses/by/4.0/).