International Journal of Electrical and Computer Engineering (IJECE) Vol.
No.
June 2016, pp.
ISSN: 2088-8708.
DOI: 10.
11591/ijece.
Permanent Magnet Synchronous Motor Design Using Grey Wolf Optimizer Algorithm Yannis L.
Karnavas.
Ioannis D.
Chasiotis.
Emmanouil L.
Peponakis Laboratory of Electrical Machines.
Department of Electrical and Computer Engineering.
Democritus University of Thrace.
Xanthi.
Hellas (GR) Article Info
ABSTRACT
Article history:
Common high-torque low-speed motor drive schemes combine an induction motor coupled to the load by a mechanical subsystem which consists of gears, belt/pulleys or camshafts.
Consequently, these setups present an inherent drawback regarding to maintenance needs, high costs and overall system deficiency.
Thus, the replacement of such a conventional drive with a properly designed low speed permanent magnet synchronous motor (PMSM) directly coupled to the load, provides an attractive alternative.
In this context, the paper deals with the design evaluation of a 5kW/50rpm radial flux PMSM with surface-mounted permanent magnets and inner rotor topology.
Since the main goal is the minimization of the machine's total losses and therefore the maximization of its efficiency, the design is conducted by solving an optimization problem.
For this purpose, the application of a new meta-heuristic optimization method called AuGrey Wolf OptimizerAy is studied.
The effectiveness of the method in finding appropriate PMSM designs is then The obtained results of the applied method reveal satisfactorily enhanced design solutions and performance when compared with those of other optimization techniques.
Received Dec 22, 2015 Revised Mar 21, 2016 Accepted Apr 16, 2016 Keyword:
PM synchronous motor Electrical machines design High efficiency motors Meta-heuristic optimization FEM analysis Copyright A 2016 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Yannis L.
Karnavas.
Lab.
of Electrical Machines.
Department of Electrical and Computer Engineering.
Democritus University of Thrace.
Office 0.
Build.
Campus.
Kimmeria.
GR-671 00.
Xanthi.
Hellas (GR).
Email: karnavas@ee.
INTRODUCTION
Three phase induction motors used to be .
nd mainly still ar.
the most popular motors used in consumer and industrial applications.
Their installed capacity is impressive.
it accounts for approximately 70% of the worldwide industrial energy consumption .
Nowadays however, the environmental concern increases continously and at the same time global economy plays an important role as it is getting more and more strict.
Thus, electrical drives with higher efficiency and lower costs are strongly desirable .
In this context, the replacement of induction machines with permanent magnet synchronous motors (PMSM) seems to be a very attractive alternative.
This alternative can be justified by several reasons: .
PMSM are capable of producing very high starting torque, .
current drawn by PMSM is directly proportional to torque so they can be controlled directly from current reading .
his is not possible with induction moto.
, .
they provide higher flux density than comparable induction motors, .
they ehxibit a wider speed range than induction motors, .
the gearbox can be eliminated in low speeds i.
lower than 500rpm, .
PMSM dont have rotor winding resulting in at least 20% lower copper losses and consequently higher efficiency, .
considerable savings in power consumption and lower noise pollution and .
relatively low price of the magnetic materials used .
This study focuses on low-speed drives without a gearbox, which are also called "direct drives".
Typical applications of low-speed direct drives to which permanent magnet machines may be used, include wind turbines, elevators, trams, boat propulsion, and waste water treatment plants.
Direct drives for wind Journal homepage: http://iaesjournal.
com/online/index.
php/IJECE A ISSN: 2088-8708 turbine generators have been studied i.
A boat propulsion system consisting of an 100kW PMSM is designed and analyzed in .
An axial flux PMSM for an elevator system is proposed in .
and an application of PMSM in traction-tram application can be found in .
Also, a PMSM for waste water direct-drive mixer motor is thoroughly discussed in .
From the above, it is apparent the the PMSM design problem has received great attention the last years.
On the other hand, the application of meta-heuristic optimization techniques have become very popular over the last two decades, especially those based on swarm intelligence i.
genetic algorithms (GA), particle swarm optimization (PSO) and ant colony optimization (ACO).
Variants of them have been applied in many fields including the solving of electrical machines design problems .
There are four main reasons that meta-heuristic have become remarkably wide-spread .
Firstly they are quite simple because they have been inspired by simple concepts with respect to physical phenomena, animalsAo behaviors, or evolutionary concepts.
Secontly they are flexible.
Their flexibility refers to their applicability to different problems without any special changes in the structure of the algorithm.
Thirdly, most of them provide derivation-free mechanisms and thus optimize problems stochastically, in contrast to gradient-based optimization approaches.
Finally, they present local optima avoidance capability compared to conventional optimization techniques.
This is due to their stochastic natures which allow them to avoid stagnation in local solutions and search the entire search space extensively.
One of the most recent meta-heuristic has been proposed by Mirjalili et al and is called AuGrey Wolf OptimizerAy (GWO) .
This paper aims to investigate and present the effectiveness of GWO algorithm in the complex problem of finding a high-efficient suitable radial flux surface PMSM design to be used as a replacement of a traditional induction motor-gearbox low speed drive system.
The motivation lies mainly in .
the fact that, to the authors knowledge extend, similar works havenAot found yet in literature and .
the concept that a particular meta-heuristic may show very high promising results on certain problems but at the same time may show poor performance on different set of problems .
The paper is organized as follows: The problem formulation .
eometrical, magnetic and electrical propertie.
of the PMSM design procedure is presented first at Section 2.
Once the optimization problem has been set, the brief preliminaries of the adopted method are shown in Section 3.
Continuing, the design variables are decided, the constants and the relevant constraints are defined, the proposed cost function is presented and the corresponding results are shown and compared next in Section 4.
The results are commented and finally the work is concluded in Section 5.
PMSM MODEL AND PROBLEM FORMULATION
Radial fluxAos PMSM are quite conventional PM machines and they are widely used for direct-driven In these machines, the flux flows radially inside the machines while the current flows in the axial direction.
The permanent magnets are placed on the rotor surface .
uter circumference of the roto.
Figure 1a shows a general cross section of the configuration under study here.
The geometrical representation refers to the outer diameter of the machine (D.
, the inner stator diameter (D.
, the rotor diameter (D.
, the magnet length .
, the pole angle .
, and the airgap length ().
Figure 1b depicts a more detailed geometry of the PMSM with inner rotor regarding slots.
Here, it is seen the slot pitch (E.
, the stator slot height .
, the stator yoke height .
, the stator teeth height .
, the stator slot opening .
, the stator slot width at start .
, the stator slot width at end .
and the stator teeth width .
We denote as Qs the number of stator slots.
Then the slot pitch Es can be defined as Es=ADs/Qs.
From this point on, the following calculations can be applied.
Ds A Dr A 2lm A 2A bss1 A A Ds A 2hsw A bst hsy A A Do A Ds A 2hss A bss 2 A A Ds A 2hss A bst As A A bss1 A bss 2 AA hss A hsw A .
where As is the slot area.
Also, since the inner stator diameter Ds is very large compared to the slot pitch Es, the relevant slot widths .
hich are actually arcs of circle.
are approximated as straight lines in Eqs.
can be seen that the 2-dimensional geometrical structure of the stator can be described entirely if the following parameters: Do.
Dr, bst, bss0/bss1 ratio, hsw, hss, lm, , and Qs are obtained.
By adding the number of poles p, the half pole angle and the active length L, the whole 3-dimensional geometry can be described.
IJECE Vol.
No.
June 2016 : 1353 Ae 1362
IJECE
ISSN: 2088-8708
Magnetic properties The PMSM design procedure relies on the amplitude of the fundamental airgap flux density e and thus its b ss0 .
Figure 1.
Representation of the design problem considered: .
inner rotor PMSM cross section, .
detailed geometry of slots, airgap and permanent magnets .
ne pol.
typical PMSM drive system.
Figure 2.
Magnetic considerations of PMSM design: .
airgap flux density variation with and without considering magnetic leakage, .
approximation of magnetic leakage as a function of poles number Calculation should be conducted.
For surface mounted permanent magnet motor designs the airgap flux density is assumed to have a rectangular shape .
s wide as the magnet widt.
and a maximum value of Bm as depicted in Figure 2a.
This value can be calculated as:
E E A A b2 E E Bm A Br kleak E A kC Ar EE lm E s ss 0 E EE A 1E E E bss 0 A 5A E E where Br is the remanence flux density of the magnet, kleak is the leakage factor .
he percentage of the flux lines which pass through the airga.
, r is the relative magnet permeability and kC is the Carter factor.
The Carter factor can be derived from .
and is shown in Eq.
, along with the leakage factor which can be obtained through magneto-static FEM simulations.
A linear approximation of kleak dependency on the poles Permanent Magnet Synchronous Motor Design Using Grey Wolf Optimizer Algorithm (Yannis L.
Karnava.
A ISSN: 2088-8708 number p has been found in .
as illustrated in Figure 2b and it is also used here:
bss2 0 E kC A A s EA s A E bss 0 A 5A E kleak A 1 A A 7 p / 59.
5A 100
Electrical and thermal properties The design problem formulation is finalized by defining the electrical parameters of the PMSM i.
inductances, stator's resistance, supply and induced voltages, ampere-turns and current density.
This kind of motors is of a non salient type and thus the d-axis and q-axis synchronous inductances are equal.
Moreover.
Ld=Lq=Ll Lmd=Ll Lmq, and the leakage inductance (L.
and the magnetizing reactances (Lmd.
Lm.
Ll A pqn LA0A1 Lmd A A qns kw1 A A0 E A kC A m E Ar E A Ds A A A L where q is the number of slots per pole per phase, nc is conductors per slot number, 1 is the specific permeance coefficient of the slot opening depending on the slot geometry, kw1 is the fundamental winding factor .
qual to unity for q=.
and L is the active machine length.
Furthermore, the induced voltage rms value along with the correponding stator winding per phase resistance can be evaluated by A A E A 1 2 Akw1qns Bm L A Ds A A A R A ACu ns2 q A pL A kewA A Ds A hss A A A f s As A A1 where fs is the slot fill factor equals to 0.
45 for distributed windings and kew is the end-winding factor .
Finally, the ampere-turns are derived from the peak current loading.
A ns I A 4TA s A A Ds A A A LBA kw1ksl where ksl is a slot leakage loss compensating factor which equals to 0.
95 for inner rotor configuration.
Now, the current density and the supply voltage across one phase can be derived using the corresponding vector diagram (Iq=I), and are therefore given by.
J A ns I A As f s A A1
V A
A E A R I A A AL A I A
Each value should not exceed the maximum allowable slot current density (Jma.
It is known that, for air-cooled machines, this value is about 800A/cm2, while for forced cooling Jmax is allowed to be more than 1kA/cm2.
It should be noted also that voltage V in Eq.
, is actually the inverter's output voltage according to Figure 1c.
Even for a "bad" scenario of 50% modulation ratio, the VLL is about 30% of the rectified dc voltage across the capacitor.
Thus.
V is approximately taken as 86.
55Volts.
OPTIMIZATION AND APPLIED METHODS
The problem stated above is actually a constrained optimization problem.
In a mathematical notation and without loss of generality can be described as .
, f A x A , x A A x1 , x2 ,.
, xn A Ea Rn where xEa W Es S .
The objective function f is defined on the search space S Es R n.
The sets W Es S and Y = S A W , define the feasible and infeasible search spaces respectively.
Usually, the search space S is defined as an n-dimensional rectangle in R n .
omains of variables defined by their lower and upper bound.
x Lb , j A 1,2,.
, n j C xj C xj IJECE Vol.
No.
June 2016 : 1353 Ae 1362
IJECE
ISSN: 2088-8708
Whereas the feasible set W Es S is defined by a set of additional POu0 constraints:
E x gk A x A C 0 .
hj A x A A 0 .
W=E
Ek A 1,2,.
L .
j A L A 1,.
P E
Any point xEa W is called a feasible solution.
otherwise, x is an infeasible solution.
Gray wolf optimization (GWO) As it is claimed by the authors in .
, the main advantage of GWO algorithm over most of the wellknown meta-heuristic algorithms is that the GWO algorithm operation requires no specific input parameters and, additionally, it is straightforward and free from computational complexity.
Further, its advantages include ease of transformation of such concept to the programming language and ease of comprehensibility.
But, as it was stated in the Introduction, there is no meta-heuristic optimization technique well suited for all optimization problems.
For this reason, we examine here the GWO suitability for the optimization problem desribed previously.
The GWO algorithm mimics the leadership hierarchy and hunting mechanism of gray wolves pack in nature.
Four types of grey wolves such as alpha (), beta (), delta (), and omega (O) are employed for simulating the pack leadership hierarchy.
Alpha wolves represent the top-level i.
the absolute leaders .
hich also give the best solution to the optimization proble.
Beta wolves represent a second-level group in hierarchy which actually help alphas in decision making while at the same time command the other lower-level wolves.
Omega wolfs refer to the lowest hirarchy level, thus the wolves belong to this group have to submit to all the other dominant wolves.
Finally, if a wolf can not be characterized as , or O, he/she is called delta.
Delta wolves have to submit to alphas and betas, but they dominate the omega ones.
In a naturecontext environment, the delta wolves may be scouts, elders, sentinels or caretakers.
In addition to this social hierarchy, group hunting is the other basic characteristic of GWO, because it provides interesting social behaviour of grey wolves.
Initialize the grey wolf population Xi .
=1,2,.
Initialize a.
A1 and A2 Calculate the fitness of each search agent X is the best search agent X is the second best search agent X is the third best search agent while .
ount< max no.
of iteration.
for each search agent Update the position of the current search agent by Equation .
end for Update a.
A1 and A2 Calculate the fitness of all search agents Update X .
X , and X count=count 1 end while return X .
Figure 3.
Simplfied illustration of Grey Wolf Optimizer algorithm, .
flowchart, .
pseudo-code There are three main steps of hunting, searching for prey, encircling prey, and attacking prey.
The mathematical implementation of these steps to GWO .
r modifications of the.
is actual the mean of performing optimization in a complex problem.
An illustration of the GWO algorithm in flowchart and pseudo-code form is depicted in Figure 3.
The basic mathematical expressions used by GWO .
rey's encircling behavio.
is formulated by.
U A A1 X prey A t A A Xwolf A t A Xwolf A t A 1A A X prey A t A A A 2 U A1 A 2r1 A 2 A 2ar2a Xprey is the prey's position vector.
Xwolf is the grey wolves position vector.
A1.
A2 are coefficient vectors, r1, r2 are arbitrary vectors in gap .
, a is a vector linearly reduced from 2 to 0 during iterations and Permanent Magnet Synchronous Motor Design Using Grey Wolf Optimizer Algorithm (Yannis L.
Karnava.
A ISSN: 2088-8708 t is any current iteration.
Especially for vector A2, it can take values in the gap [-2a, 2.
, so as when |A.
<1 the wolves are forced to attack the prey.
This is translated as the method's exploitation ability while searching for prey is the exloration ability.
When |A.
>1 the wolves pack is enforced to diverge from the prey.
In any iteration, the wolvesAo position update is performed by.
E Xa _ wolf A t A A A 2(A _ wolf ) A1(A _ wolf ) XA _ wolf A t A A X wolf A t A A E X wolf A t A 1A A A1/ 3A E X A _ wolf A t A A A 2( A _ wolf ) A1( A _ wolf ) X A _ wolf A t A A X wolf A t A A E E XA _ wolf A t A A A 2(A _ wolf ) A1(A _ wolf ) XA _ wolf A t A A X wolf A t A E Genetic Algorithms and Particle Swarm Optimization For fair comparison purposes, two other meta-heuristic optimization techniques are also applied to the same problem, genetic algorithm (GA) and particle swarm optimization (PSO).
These techniques have been proved robust and efficient enough to most engineering problems in the past.
Moreover, since they are well-known, their principles of operation have been extensively presented in the literature and therefore no relative information is given here.
The reader can refer to .
for further implementation details.
Objective functions and case studies The desired PMSM characteristics are shown in Table A1 in the Appendix.
Due to the large number of variable names, symbols etc.
, all the relevant information as well as the results obtained here will be presented in a tabularized form.
From Figure 1 and Eqs.
, it is derived that there are twelve design variables that are needed to be optimized by the applied algorithms.
All the other quantities can be calculated Table A2 in the Appendix shows these design variables, all of them related to the motorAos In the same Table the upper and lower bounds of the variables are given.
Additionally.
Table A3 summarizes the constant values used in the calculations, while Table A4 presents the required problem A simple formulation for the objectives/cost functions (CF) to be minimized is proposed here as.
CFn A A k E Qk E CF A Au 0.
15Ay E E M tot E w E CF2 A Au 0.
15Ay E EE M w E CF3 A Au 0.
70Ay E EE M wtot PLtot M mtot EE PLtot M mtot EE PLtot M mtot EE where k is a 1xk row matrix containing the cost function's weight coefficients and Qk is a kx1 column matrix containing any quantities under concern.
It can be seen that this type of formulation can be expanded to more variables easily.
Here, the quantities chosen are three .
: the total machine weight (Mwto.
, the total magnets weight (Mmto.
and the total copper losses (PLto.
Furthermore, although numerous cost functions can be produced that way, three variations .
are examined.
APPLICATION RESULTS AND DISCUSSION
Nine sets of optimization results were derived .
methods x 3 objective function.
for the inner rotor PMSM design problem.
Table 1 shows the twelve design variables through GWO.
GA and PSO respectively and for each one of the cost functions applied (CF1-CF.
From this Table it is initially clear that all algorithms succeeded in converging to a -near- optimum design solution satisfying all of the existing Also, compared to .
, these results, as well as those follow, are much better.
For example, the outer diameter of the motor (D.
is kept within satisfactory limits of 24-38cm, with a 20-50cm constraint, the machine length (L) found between 10-17cm only .
ith a 10-50cm constrain.
, while the magnet height .
varied from 7.
8mm to 12.
ith a 2-15mm constrain.
More compact information can be found in Table 2 which shows the motors total weight, their efficiency and the total magnets weight in each case, through the algorithms applied and for every cost It is seen that the corresponding quantities ranges .
in-ma.
found to be 48.
1kg, 74.
4% and 1.
5kg respectively, by far less than the relative constraint.
This Table reveals also the .
If the machine weight is the primary objective .
ost function CF.
, the GWO solutions present the lighter .
and more efficient .
%) motor, by keeping the magnet weight low enough .
The second "winner" in this case is GA, while for the PSO method present a little worse results than the GA.
IJECE Vol.
No.
June 2016 : 1353 Ae 1362
IJECE
ISSN: 2088-8708
The same conclusions are valid in the case where the magnets weight is the primary objective .
ost function CF.
GWO method presents a solution with low magnet weight .
, satisfactory efficiency .
%), while keeping the total machine weight reletively low .
, as also shown in Table 2.
The second alternative is given again by GA, while PSO method presents quite much magnet weight.
If the PMSM copper losses, thus the efficiency, is the primary objective (CF.
, the GWO method presents the most satisfactory results, by keeping the losses the lowest possible resulting at a high-efficient motor .
43%), whereas PSO takes the lead over GA, proposing a solution with an efficiency of 84.
55% and finally GA method is the worst performing, presenting an efficiency of 82.
The left part of Table 3 shows the computational costs of the adopted and applied methods.
Here, the first place goes to GWO method which provides solutions in the order of magnitude of 2-3 seconds for our problem.
PSO needs almost 1.
5 times more computational time, while GA is the slowest optimization method with a response time of almost 10 seconds.
Figure 4.
Convergence behavior of applied optimization methods to PMSM design problem, .
GWO, .
GA, .
PSO .
Figure 5.
PMSM design obtained (CF.
by GWO application.
flux density distribution, .
heat density plot Permanent Magnet Synchronous Motor Design Using Grey Wolf Optimizer Algorithm (Yannis L.
Karnava.
A ISSN: 2088-8708 Table 1.
PMSM design variables results through GWO.
GA and PSO optimization methods Nspp bss0/bss1 .
CF1
CF2
CF3
CF1
CF2
CF3
CF1
PSO
CF2
CF3
Table 2.
Main objective quantities results through GWO.
GA and PSO optimization methods Mwtot Mmtot (%) .
CF1
GWO
CF2
CF3
CF1
CF2
CF3
CF1
PSO
CF2
CF3
Table 3.
Computational cost of optimization methods used and corresponding PMSM design solutionsAo slot
current density GWO
PSO
Computational cost .
CF1
CF2
CF3
Slot current density (A/cm.
CF1
CF2
CF3
In terms of converging, there are also interesting results.
Figure 4 shows the behavior of GWO.
GA and PSO convergence throughout the iterative process (CF2 cas.
It can be seen that GWO converges to the 12-variable optimal combination in about 500 iterations giving a cost function value of 0.
GA needs approximately twice iterations while trying to explore/exloit the solution search space and converges to 0.
PSO seems to converge prematurely scoring 0.
Our investigation continues with the thermal calculations involving slot current density for each one of the geometries.
These are summarized in the right part of Table It can be seen that all designs develop a current density lower that 800A/cm2 and can be realized as aircooled machines .
s stated in A2.
One of the proposed by GWO algorithm optimized design .
ptimized for efficiency i.
CF.
was implemented in FEM and thermal analysis software and visualization instants of its operation under nominal load are shown in Figure 5, where the relevant geometry, the flux density distribution as well as the heat density plots are depicted.
It can be seen that the maximum magnetic flux density, in any part of the machine is lower than the constraint set, with a maximum of approximately 1.
ith a 1.
6T constrain.
, while the maximum temperature rize is about 39oC .
ithout an outer enclosure/frame structur.
In other words, these results from both FEM and thermal analyses are judged absolutely satisfactory.
CONCLUSION
The implementation of GWO into inner rotor PMSM design problem has been presented in this The results obtained show that GWO have been successfully implemented to anticipate the practical constraints of PMSM design problem and provide very competitive alternative designs to replace traditional low speed induction-motor/gearbox systems, in terms of minimizing total losses and permanent magnet It has been observed that the GWO has the ability to converge to a better quality optimal solutions and possesses better convergence characteristics than other prevailing techniques reported in literature.
The solution provided by GWO was also evaluated through finite element analysis as well as thermal analysis procedures, where an acceptable and satisfactory performance was recorded.
It is also clear from the results obtained, by several trials, that the GWO shows a good balance between exploration and exploitation, a property which result in high local optima avoidance.
Thus, this algorithm may become very promising for solving complex engineering optimization problems such as the electrical machine design one.
IJECE Vol.
No.
June 2016 : 1353 Ae 1362
IJECE
ISSN: 2088-8708
APPENDIX
Quantity Symbol
Unit
Induction
Motor
Gear-box
Total
PMSM
Table A1.
Induction motor/gearbox data corresponding to PMSM Outer Stator Dia.
Machine length
Machine weight
Efficiency Shaft torque Shaft speed
Output Power
Line current
Supply voltage Frequency Mwtot
Pout
VLL
Volt
1:30
A2.
A2.
Table A2.
Universe of discourse of PMSM design problem Quantity No.
of magnet poles No.
of slots / pole per phase Outer Stator Diameter Outer Rotor Diameter Active machine length No.
of parallel conductor paths Magnet thickness Stator tooth width
Stator slot height
Slot wedge height
Stator slot bs0 / bss1 ratio
Airgap length
Symbol
Nspp
Variable Range 20 - 80
001 - 1
20 - 50 cm
| 20 - Do | cm
10 - 50 cm
10 - 40
2 - 15 mm Ou 2.
5 mm Ou 0 mm 1 - hss mm 2 - 0.
3 - 20 mm Table A3.
Constant values involved in the PMSM design Table A4.
PMSM design problem inequality constraints Quantity Maximum flux density (NdFeB) Remanence flux density Area occupied by conductors Operating frequency Relative permeability MotorAos shaft speed Mass density of the copper No.
of phases Mass density of the back iron Copper resistivity Mass density of the magnet Torque on the motorAos shaft No.
of armature paths Description Stator yoke height Slot wedge height Slot opening height Slot width Tooth width Slot opening width Flux density in stator teeth Flux density in stator yoke Flux density in rotor yoke Airgap flux density Copper losses Magnet weight Machine weight Symbol Bmax Pcopper Nph Pbi ACu Set Value 8920kg/m3 7750kg/m3 72x10-8 /m 7500kg/m3 Symbol Bts Bsy Bry B PCu Mmtot Mwtot Constraint >=hss/2 >=1mm >=2mm 15hss<=bss2<=0.
>=0.
3Es >=2mm <=1.
<=1.
<=1.
<=1.
<=700W
<=5.
<=150kg REFERENCES