World Applied Sciences Journal 7 (3): 393-398, 2009

ISSN 1818-4952
© IDOSI Publications, 2009

Power-to-volume of the High Speed Generators

1
M. Sadeghierad, 2 A. Darabi, 1 H. Lesani and 1 H. Monsef

1
Department of ECE, University of Tehran, Iran
2
Faculty of Electronics and Robotics Engineering, Shahrood University of Technology, Iran

Abstract: The high-speed generators offer superior power-to-volume ratio and highly compact structure
comparing with their lower speed counterparts. These advantages make high-speed machines eligible for
distributed generating application with microturbine that one of the distributed power generation that
produced electric in the remote and disastrous zones where grid power is unavailable. To achieve higher
power-to-volume ratio, it is offer to use optimization methods. In this paper, two powerful optimization
algorithms, i.e. genetic algorithm and Particle Swarm Optimization (PSO) are considered. The model of
machine with loss calculations is prepared for this optimization. It will be shown the power-to-volume ratio
of the High-speed Axial Flux Generator (HSAFG) can be increased two times by using PSO.

Key word: High-speed machine axial-flux generator power-to-volume ratio optimization genetic algorithm
• • • •

PSO
•

INTRODUCTION Rotation losses that the power to overcome drag

resistance of rotating disc can be obtained by this
Lower cost, compactness and high efficiency formula:
features of axial flux machines [1] make it with high-
speed rotor [2, 3], suitable for distributed generating 1
Pmech = c f ρ (2πN m) 3 (R o5 − R sh 5 ) (1)
application. But, designing of such a high-speed 2
generator, in mechanical and electromagnetic aspects,
is quite different to conventional ones with low speed Parameter cf (the coefficient of drag for turbulent
and low frequency [4, 5]. The modeling and optimization flow) can be found in the reference [8].
of high-speed axial-flux generators (HSAFG) for Because of high frequency, it is suggested to use a
microturbine is introduced in this paper. Particle swarm coreless stator [9-11]. So there is not core losses, but
optimization (PSO) and genetic algorithm are presented only high frequency causes additional losses in winding
for optimizing the power-to-volume ratio. It will be (eddy current losses) [3]:
shown PSO proposed better results.
( Bg × 2 × π × f × Dstrand ×10 −3 )2
P eddy_cu = × Vcu (2)
MODEL OF MACHINE 32 × ρcu

Figure 1 shows a schematic view of intended So total losses in HSAFG is equal to:
machines with three rotors and two stators.
For optimization, the magnetic circuit model of the
P = Pmech + Pcu + Pe d d y _ u (3)
HSAEG comprising all reluctances that described in loss

other papers of authors [6, 7] is used. In this modeling,

the nonlinear model of PM and back iron (iron OPTIMIZATION ALGORITHMS
saturation), the PMs leakages and the air gap fringing
phenomena considerations are taking into account for To compare and verify this optimization two
achieving the most accuracy of the model. For efficiency different optimization algorithms i.e. genetic algorithm
calculation, power losses have to be intended. and the intelligent method (PSO) are used.
Corresponding Author: Dr. M. Sadeghierad, Faculty of Electrics and Computer Engineering, Univeristy of Tehran,
North AmirAbad, Tehran, Iran
393
 World Appl. Sci. J., 7 (3): 393-398, 2009
Fig. 1: View of a high-speed axial flux generator
First method, genetic algorithm generates a
population of points at each iteration. The population
approaches an optimal solution. The next population
by computations that involve random choices is
selected. [12]
Second, PSO is a global optimization algorithm for
dealing with problems in which a best solution can be
represented as a point or surface in an n-dimensional
space. Hypotheses are plotted in this space and seeded
with an initial velocity, as well as a communication
channel between the particles. Particles then move
through the solution space and are evaluated according
to some fitness criterion after each time step. Over
time, particles are accelerated towards those particles
within their communication grouping which have
better fitness values [13].
To transform constrained optimization problem into
unconstrained ones, the penalty function is used. The
penalty method replaces a constraint optimization
problem by a series of unconstrained problems whose
solutions must converge to the solution of the original
constrained problem. The penalty term grows when the
constraints are violated and is 0 in the region where
constraints are not violated. The penalty term is usually
a product of a positive penalty coefficient and a penalty
function.
The resultant objective function is:
u
r r r
F(x, ω ) = f(x) + ∑ ωi εi (x)
i =1
r genetic algorithm is 89.6MW/m3, with PSO is 117MW/m3
where ε i (x) are functions that penalize increased
constraint violation.
RESULT OF EFFICIENCY OPTIMIZATION
By described model of machine, the design
algorithm of HSAFG is obtained (Appendix 1).
Design and optimization of the generator is done
according to this algorithm. These are nine optimization
parameters: “Do” outer diameter, “λ “ ratio of Do/Di, “g”
air gap, “Ls” stator width, “Lpm ” PM thickness, “Lbi”
Back iron thickness, “d”pole arc to pole pitch ratio,
“Dstrand “ diameter of litz wire, “Br” denotes remanence of
PM. The power-to-volume ratio of machine is selected
as the optimization function:
Pout
f(x) = (5)
Vgenl
For the generator volume (Vgen) calculation, the
outer diameter and total length (included back-iron
length, PM length, stator length and air-gap length) is
considered.
If neither constraints nor boundaries are used for
the optimization, the result vector may contain “non-
physical” values, such as negative values for
resistances. Consequently, some boundaries are defined
i.e. 2mm<g, λ<0.95, etc. Also the value of output power
and output voltage are the constrains and considered
by penalty function.
The genetic algorithm and PSO are separately
executed. Table 1 shows the results of a power-to-
(4) volume ratio optimization of a 140kW, 50,000-rpm
generator. The power-to-volume that obtained with
and with designed efficiency is 52.1MW/m3. It is shown
in Fig. 2.
394
 World Appl. Sci. J., 7 (3): 393-398, 2009

Table 1: Results (optimal design) obtained by optimization algorithms

Optimization parameter
------------------------------------------------------------------------------------------------------------------------------------------------------------
Parameters Genetic algorithm PSO Initial Design
Do (cm) 22.9 20.1 30.0
λ 0.42 0.45 0.58
g (mm) 4.30 4.20 1.95
Ls (mm) 2.20 5.00 4.50
Lpm (mm) 1.50 3.20 1.50
Lbi (mm) 2.50 6.00 10.00
δ 0.80 0.91 0.90
Dstrand (mm) 0.12 0.17 0.10
Br (T) 1.31 1.19 1.20

power-to-volume Power losses (W)

25,298
Ploss_total 8,066
8,804
52.1
24,731
P_mech Designed
Designed 4,951
8,432
117.0 Swarm Swarm
98
genetic Peddy_cu 3,077 genetic
122
89.6 469
38
Pcu
251

0.0 50.0 100.0 150.0 0 5,000 10,000 15,000 20,000 25,000 30,000

MW/m3 W

Fig. 2: Power-to-volume (kW/m3) of the HSAFG Fig. 4: Power losses (W) obtained by optimization
methods
Efficiency
Output power

84.87%
141.90
Designed
Designed
94.65% Swarm
142.77 Swarm
genetic
genetic
94.17% 142.34

75% 80% 85% 90% 95% 100% 141.0 141.5 142.0 142.5 143.0
kW

Fig. 3: Efficiency of the HSAFG

According to Fig. 3, the efficiencies of HSAFG are
94.17, 94.65 and 84.87% with genetic algorithm, PSO and
designed algorithms. The large numbers of members
that make up the PSO (particle swarm) make the
technique impressively resilient to the problem of local
minima such as this optimization, so it can confer higher
power-to-volume ratio.
Fig. 5: Output power (kW) obtained by optimization
methods
Other main parameters i.e. power losses and output
power are presented in Fig. 4 and 5. According to
presence of permanent magnet in this machine, the
power losses have to be considered. Figure 4
shows the power losses are reduced by both suggested

395
 World Appl. Sci. J., 7 (3): 393-398, 2009
Appendix 1

396
 World Appl. Sci. J., 7 (3): 393-398, 2009

optimization methods (from 17.71% to 5.65% and 6.18%). REFERENCES

The minimum power losses are obtained by PSO
optimization. According Fig. 5, the output power is near 1. Marignetti, F., V. Delli Colli and Y. Coia, 2008.
140kW in all three designs. Design of Axial Flux PM Synchronous
Machines Through 3-D Coupled Electromagnetic
CONCLUSION Thermal and Fluid-Dynamical Finite-Element
Analysis. IEEE Trans. on Industrial Electronics,
In this article, regarding the high power-to- 55 (10): 3591-3601.
volume of HSAFG, the optimization methods were 2. El-Hassan, T. and P.C. Luk, 2003. Magnet topology
used for improving this characteristic of machine. Two optimization to reduce harmonics in high speed
different optimization algorithms, genetic algorithm axial flux generators. INTERMAG IEEE Conference,
and PSO, were presented in this paper for the GS-03.
unconstrained design optimization of the HSAFG. The 3. Park Jae-Do, C. Kale and H.F. Hofmann, 2008.
PSO made the better result than genetic algorithm. Control of High-Speed Solid-Rotor Synchronous
They were shown by using the methods the power-to- Reluctance Motor/Generator for Flywheel-Based
volume ratio of this machine could be increased Uninterruptible Power Supplies. IEEE Trans. on
about two times. Industrial Electronics, 55 (8): 3038-3046.
4. Luk, P.C.K. and T.S. El-Hasan, 2005. Integrated
List of symbols design for a high speed permanent magnet axial
η: Efficiency flux generator. INTERMAG IEEE Conference,
Pout: Output power (kW) pp: 1083-1084.
PLoss: Power loss (kW) 5. Zwyssig, C., S.D. Round and J.W. Kolar, 2008. An
Pmech : Mechanical (rotation) power loss (kW) Ultrahigh-Speed, Low Power Electrical Drive
Pcu: Copper power loss (kW) System. IEEE Trans. on Industrial Electronics,
55 (2): 577-585.
Peddy_cu: Copper power loss (kW)
6. Sadeghierad, M., H. Lesani, H. Monsef and A.
Nm : Rotor speed (rad/s)
Darabi, 2008. Leakage Flux Consideration in
ρ: Air density (kg/m3)
modeling of High Speed Axial Flux PM Generator.
Ro: Outer radius (m)
International Conference on Industrial Technology
Rsh: Shaft radius (m) (ICIT IEEE International), pp: 1-6.
Bg: Flux density at air gap (T) 7. Sadeghierad, M., H. Lesani, H. Monsef and A.
f: Frequency of rotor (Hz) Darabi, 2008. Air gap sensitivity in High-Speed
Dstarnd: Diameter of each wire of conductor (m2) Axial-Flux PM Machine. International Conference
ρ cu: Conductor resistivity (O.m) on Electrical Machines (ICEM IEEE International),
r
F(x, ω) : The objective function to be minimized pp: 1-4.
r
f(x) : The function to be minimized 8. Gieras, J.F., R.J. Wang and M.J. Kamper, 2005. Axial
ωi: Weighting factor Flux Permanent Magnet Brushless machine.
r Kluwer Academic Publisher.
ε i (x) : The function that penalize increased constraint
9. Rong-Jie Wang, Kamper, M.J., K. Van der
violation Westhuizen and J.F. Gieras, 2005. Optimal design of
λ: Ratio of Do/Di a coreless stator axial flux permanent-magnet
g: Air gap length (m) generator. IEEE Transactions on Magnetics,
Ls: Stator width (m) 41 (1-1): 55-64.
Lpm : PM thickness (m) 10. Kamper, M.J., Wang Rong-Jie and F.G. Rossouw,
Lbi: Back iron thickness (m) 2008. Analysis and Performance of Axial Flux
d: Pole arc to pole pitch ratio Permanent-Magnet Machine with Air-Cored
Br: Denotes remanence of PM (T) Nonoverlapping Concentrated Stator Windings.
Vcu: Volume of copper (Winding) (m3) IEEE Transactions on Industry Applications,
Vgen: Volume of the generator (m3) 44 (5): 1495-1504.

397
 World Appl. Sci. J., 7 (3): 393-398, 2009

11. Rong-Jie Wang and M.J. Kamper, 2004. Calculation 13. Engelbrecht, A.P., 2005. Fundamentals of
of eddy current loss in axial field permanent-magnet Computational Swarm Intelligence. Wiley.
machine with coreless stator. IEEE Transactions on
Energy Conversion, 19 (3): 532-538.
12. Bahmani Firouzi, B., E. Jamshidpour and T. Niknam,
2008. A new Method for Estimation of Large
Synchronous Generator Parameters by Genetic
Algorithm. World Applied Sciences Journal,
4 (3): 326-331.

398