Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Accepted Manuscript

A Comprehensive performance assessment of the integration of magnetic


bearings with horizontal axis wind turbine

M. Fekry, Abdelfatah M. Mohamed, Mohamed Fanni, S. Yoshida

PII: S0378-4754(18)30168-X
DOI: https://doi.org/10.1016/j.matcom.2018.06.011
Reference: MATCOM 4606

To appear in: Mathematics and Computers in Simulation

Received date : 6 November 2017


Revised date : 23 June 2018
Accepted date : 23 June 2018

Please cite this article as: M. Fekry, A.M. Mohamed, M. Fanni, S. Yoshida, A Comprehensive
performance assessment of the integration of magnetic bearings with horizontal axis wind turbine,
Math. Comput. Simulation (2018), https://doi.org/10.1016/j.matcom.2018.06.011

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
A Comprehensive Performance Assessment of the Integration of Magnetic Bearings
with Horizontal Axis Wind Turbine

M.Fekrya,b,∗, Abdelfatah M. Mohameda,c , Mohamed Fannia,d , S. Yoshidae


a Dept. of Mechatronics and Robotics Engineering

Egypt-Japan University of Science and Technology


New Borg El-Arab City, Alexandria, 21934, Egypt
b On leave from Dept. of Electrical Power and Machine Engineering, Zagazig University, Zagazig, Egypt
c On leave from Dept. of Electrical and Electronics Engineering, Assiut University, Assiut, Egypt
d On leave from Dept. of Production Engineering and Mechanical Design, Mansoura University, Mansoura, Egypt.
e Research Institute for Applied Mechanics, Kyushu University, Kasuga, Fukuoka 816-8580, Japan

Abstract
This paper presents a comprehensive performance assessment for the use of Magnetic Bearing (MB) with Wind Turbine
(WT) . It is reported that this integration aims to improve the WT performance by eliminating fractional losses,
mitigating vibration, reducing cut-in speed and prolonging the life span. However, there is no thoroughly studies
regarding this integration in the literature to show in what extent the power generation will be affected in comparison
with the use of conventional mechanical bearing (CB). This task constitutes the aim of this work. First, the main shaft
of an existing conventional WT is redesigned to match the assembly with the MBs. Second, the design of two Radial
Homopolar Pole Biased Hybrid Magnetic Bearings (RHPBHMB) and one Axial Active Magnetic Bearing (AAMB) is
introduced. The MB design is analyzed using nonlinear magnetic circuit analysis and FEM. The MSC Marc software is
employed to build 3D and 2D FEM models for the two radial and axial MBs respectively. Thirdly, the mathematical
dynamic equations for the overall system is derived and an elaborated multi-disciplinary dynamic model is built using
Simscape package. In addition, a robust intelligent TSK fuzzy Q-parameterization controller is synthesized to stabilize
the RHPBHMB in order to achieve robust stability, overcome model nonlinearity and reject the step and sinusoidal
imbalance disturbance at any rotational speed. Finally, an extensive comparison between the performance of the WT
supported with Conventional mechanical Bearing (WT-CB) and the WT supported with Magnetic Bearing (WT-MB)
is provided. The results shows the ability of MB to defeat the disadvantages of the Conventional mechanical Bearing
(CB), as well as, enhance the WT performance without decreasing the power generated.
Keywords: Wind Turbine, Magnetic bearing design, Homopolar magnetic bearing, Hybrid magnetic bearing, Magnetic
circuit analysis, FEM, Pole biased magnetic bearing, Parallel distribution control, Intelligent control, Robust control,
Robust stability, Q-parameterization, Wind turbine performance, Extensive comparison, Comprehensive assessment.
2010 MSC: 00-01, 99-00

1. Introduction more, this will result in decreasing cut in speed, reducing


noise, increasing efficiency, rationalizing energy consump-
The MB is a suspension system to levitate a rotating tion, eliminating lubricant contaminations, being mainte-
object freely without any physical contact. It enables a ro- 15 nance - free and prolonging turbine design lifetime [2, 3, 4,
tating shafts to attain very high speed without any friction 5, 6, 7, 8]. Also, the inherent instability of MB is a chal-
5 or mechanical wear [1]. The operation of MB depends on lenging problem which must be overcome via an efficient
harnessing magnetic fields to control the position of a fer- controller. This controller should be able to reject dis-
romagnetic object (typically rotor) precisely. These forces turbances, compensate for vibrations and overcome model
may be attractive or repulsive according to the suitability 20 nonlinearity and uncertainties.
of application. The WT is a multi-disciplinary complex system that
10 The integration between MB and WT helps in eradicat- includes aerodynamics, mechanical and electrical subsys-
ing the frictional losses throughout the turbine. Further- tems. As a matter of fact, to provide comparative study
between using WT-CB and WT-MB, the study should be
∗ Corresponding author 25 multi-disciplinary which incorporates all these subsystems
Email addresses: mohamed.fekry@ejust.edu.eg (M.Fekry), and consider the interactions between them.
Abdelfatah.mohamed@ejust.edu.eg (Abdelfatah M. Mohamed),
Mohamed.fanni@ejust.edu.eg (Mohamed Fanni), The authors in [9] have investigated the possibility of
yoshidas@riam.kyushu-u.ac.jp (S. Yoshida) applying MB to direct drive wind turbine in order to re-
Preprint submitted to Journal of Mathematics and Computers in Simulation June 24, 2018
duce the weight of generator and increase the structure can drive the AC radial-axial HMB in a stable way with
30 flexibility. Firstly, they have designed the MB for a WT excellent performance. However, the authors neither con-
called Harakosan Z72 (1.5MW @ 18.5rpm). Based on the sider the aerodynamic system nor provide any comparison
analytical and numerical design, they estimated that the with CB.
losses in the MB is below 0.5% of the rated power which 90 The authors in [15] have designed a practical testbed
is typically the same magnitude of loss for CB. After that, for a WT with a PMB. The testbed provides a basis for
35 a new concept depending on using the CB to isolate wind the experimental determination of vibration, evaluation of
loads was applied in designing 5MW generator in [10]. As bearing performance and estimation of efficiency of a WT
a result, the rotor weight is reduced by 45%( from 50 tons under various wind conditions.
to 28 tons). In addition, they have introduced three con- 95 The authors in [16] have presented a 5-DOF magnet-
cepts of using MB in direct drive machine in [11]. The first ically levitated WT with two radial PMB and one axial
40 concept is a single MB concept which relies on replacing HMB. Additionally, they designed a fuzzy PID-like con-
the CB with a magnetic one, the second concept is the troller to stabilize the system. However, the PMB has
flexible rotor concept which depends on design magneti- limited ability to compensate aerodynamic disturbance.
cally levitated rotor made of a thin cylindrical ring. The100 A new self-decoupling magnetically levitation genera-
third concept is the hybrid magnetic bearing which de- tor has been proposed for WT in [17, 18]. The generator
45 pends on isolating the wind load by CB and magnetically is able to control 2-DOF in the radial direction while the
levitated stiff rotor ring. The later concept is the best WT is suspended by 3-DOF HMB. Although this configu-
one because it combines the ease of control and structure ration is suitable for short shaft length, a portion of shaft
flexibility. Later in [12] they have introduced a complete105 weight is still supported by bias current of the generator
design for 5 MW direct drive generator with the hybrid bearing.
50 magnetic bearing concept. However, the authors of these Double Stage Cross Feedback Control and Cross-feed-
papers focused only on the MB design process regardless back-based Sliding Mode Control have been designed in
of controller design or aerodynamic model which means [19, 20] to overcome the gyroscopic coupling effect and the
the lack of considering the interactions between the differ-110 imbalance forces generated under the high-speed situation
ent subsystems. Moreover, his comparison didn’t include of the magnetically levitated WT. However, this work con-
55 accurate estimation for the CB losses. centrates only on control process.
The authors in [13] have studied the effect of shaft ec- Table 1 summarizes the different aspects of the Mag-
centricity of a small WT on the generator performance netically Levitated Horizontal Axis Wind Turbine (ML-
levitated by Passive Magnetic Bearing (PMB). They used115 HAWT) researches mentioned in above literature. It is
CB in the generator to stabilize the axial DOF. The proper obvious that, there is a still need for providing a fair com-
60 stiffness to compensate this eccentricity is determined to parison between the use of MB and CB with Horizontal
meet the generator requirements. After that, in [7], they Axis Wind Turbine (HAWT). So, a comprehensive model-
have experimentally introduced a complete design of the ing of HAWT which include the aerodynamic, mechanical
the magnetically levitated WT with PMB and studied120 and electrical model is necessary to consider the different
the radial forces and natural frequencies of the rotor sys- subsystems effects on each other. As well, this compara-
65 tem. Later in [8], they have incorporated the aerodynamic tive study will provide a clear pattern on how much the
loads of the impeller using Computational Fluid Dynam- MB improves the performance of WT. Over the above, it
ics (CFD) method. However, the use of CB at any stage is reported in [21, 22, 23] that the bearings and gearbox
degrades the MB performance to some extent because of125 are responsible for 76% and 17% of wind turbine failure
friction. in 2015 respectively. Besides, the gearbox is replaced af-
70 In [14], a design and implementation of a novel real- ter five years of installation, although the typical design
time controller for Active Magnetic Bearing (AMB) equipped life for its utility is 20 years which results in doubling the
for WT has been presented. This controller is implemented capital cost.
using a floating-point digital signal processor (DSP) and
on a Field-Programmable Gate Array (FPGA). Further-
75 more, the controller has high computing capacity, low sys- Table 1: Research status of magnetically levitated Horizontal axis
tem costs and inherent flexibility which result in an im- wind turbine
Ref. Year Suspension DOF Aerodynmic MB design Control Comp. Exp.
provement of the control performance and a remarkable [9] 2007 HMB 5 - + - + -
[10] 2008 HMB& Mech. B 1 - + - - -
reduction in software development time. However, the au- [11] 2009 HMB & Mech. B 1 - + - - -
[12] 2010 HMB & Mech. B 1 - + - - -
thor focused only on the controlling of AMB regardless of [13] 2012 PMB & Mech. B 2 + + o - +
80 the others subsystems. [7]
[8]
2013
2014
PMB & Mech. B
PMB & Mech. B
2
2
+
+
+
+
o
o
-
-
+
+
The authors in [6] have introduced the design and con- [14] 2010 AMB - - - + - +
[6] 2012 HMB 5 - + + - +
trol for AC radial-axial Hybrid Magnetic Bearing (HMB) [15] 2012 PMB & Mech. B 2 - + + - -
[16] 2012 PMB & HMB 5 - + + - -
which has been used in wind energy generation system. [19] 2014 AMB 4 - - + - -
[20] 2015 AMB 4 - - + - -
The experimental results showed that the presented dis- *
+:included, -:not included o:not rquired

85 placement/current measuring method and control system


2
130 In this paper, a comprehensive comparison between the
performance of WT-CB and WT-MB is conducted. So, a
multi-disciplinary elaborated model for the WT including
aerodynamic, mechanical, and electrical subsystems is de-
veloped using Simscape Simulink package. In section 2,
135 the main shaft of an existing WT is redesigned to match
the assembly with two RHPBHMB and one AAMB. The
MBs are designed and analyzed using magnetic circuit and
FEM. In section 3, the overall mathmatical dynamic equa-
tion for the whole system is derived. Besides, the frictional
140 moment of CB is calculated. In section 4, the TSK fuzzy
Q-parametrization control with PDC is synthesized for the
two RHPBHMB and Q-parametrization is synthesized for
the AAMB. Furthuremore, the controller performance is
Figure 1: Magnetically Levitated Wind Turbine
evaluated to prove that the controller achieves the desired
145 objectives. In section 5, an extensive comparison is evolved
between the performance of WT-CB and WT-MB. The re- Table 3: Mechanical properties of S45C [24].
sults depict the ability of MB to improve the performance Property Unit Value
Density (kg/m3 ) 7700-8030
of WT by gaining the well-known merits of MB such as Youngs Modulus (GPa) 190-210
eliminating the frictional losses, decreasing cut in speed, Tensile Strength (Mpa) 569 (Standard) 686 (Quenching, Tempering)
150 minimizing vibration, prolonging the lifespan and increas- Yield Strength (Mpa) 343 (Standard) 490 (Quenching, Tempering)
ing efficiency without decreasing the extracted power from Poisson’s ratio 0.27-0.30
the wind.

2. Design of Overall System

This section presents the design of main shaft of the


155 WT and the MBs that is necessary to position the WT
along its five axes. The parameter of system under study
and loads are given in Table 2 . Two RHPBHMB and one
AAMB are designed to levitate the WT. Fig 1 shows the
different components of the overall system.

Table 2: Loads of wind turbine components


Load types Componant Load
Blades 3×2.75 kg Figure 2: Free body diagram of shaft
Hub 7.2 kg
Static
Shaft and thrust
65 kg [25]. Fig. 2 and 3 show the free body diagram and the
disk of axial AMB
Generator 85 kg dimensions of the designed shaft respectively. In addition,
Axial forces 1554.49 N the static analysis of the shaft is performed by FEM using
Dynamic 170 CATIA software. The shaft is studied under six loading
Tilt moments 250 N.m
conditions described in Table 4. The stress and deflec-
Parameter Value
tion analysis can be found in Fig.4 & 5 respectively. The
Designed Power 3 kW
max. stress is 65.3 MPa which is five times less than the
Rotor Radius of Wind Turbine (Propeller) 1.25 m
designed material yield strength. Furthermore, the max.
Rated rotational speed 300 rpm
175 deflection is 0.39 mm. Besides, the frequency analysis is
performed to determine the first natural frequency of the
modified shaft and it was found to be 62.596 Hz. This
160 2.1. Shaft Design: value is much higher than the rated speed of wind turbine
The shaft is designed iteratively until reaching the de- which is 300 rpm (5 Hz). In addition, the fatigue strength
sired requirement such as withstanding the static and dy-180 is conducted to determine the minimum shaft diameter,
namic loads and matching the dimensions of the two RH- d, at the most critical location using Goodman equation.
PBHMB and one AAMB. The shaft material is made of The calculated diameter is 50 mm which is less than the
165 carbon steel S45C whose parameter is given in Table 3. designed one (55 mm).
The shaft diameter is obtained using Goodman equation

3
Figure 3: The upper half of shaft dimensions

Table 4: Load cases of the shaft


Load conditions Load case forces Value Max. Stress Max. Deflection
Blade+ Hub 15.45 kg Figure 4: The shaft stress analysis (Load case: 1)
Shaft 65 kg
Generator 85 kg
Mtilt 250 N.m
M B1 rotor 5 kg
1 6.53 × 107 Pa 0.388 mm
M B2 rotor 8.5 kg
Aux. bearing 0.4 kg
Axial Loads 1500 N
R1 960 N
R2 800 N
Blade+ Hub 15.45 kg
Shaft 65 kg
Generator 85 kg
Mtilt -250 N.m
M B1 rotor 5 kg
2 6.36 × 107 Pa 0.36 mm
M B2 rotor 8.5 kg
Aux. bearing 0.4 kg
Axial Loads 1500 N
R1 340 N
R2 1420 N
Blade+ Hub 15.45 kg
Shaft 65 kg
Generator 85 kg
Mtilt 250 N.m
M B1 rotor 5 kg
3 6.41 × 107 Pa 0.39 mm
M B2 rotor 8.5 kg
Aux. bearing 0.4 kg
Axial Loads -1500 N
R1 960 N Figure 5: The shaft deflection analysis (Load case: 3)
R2 800 N
Blade+ Hub 15.45 kg
Shaft 65 kg
Generator 85 kg 2.2. Magnetic Bearing System Analysis:
Mtilt -250 N.m
4
M B1 rotor 5 kg
6.3 × 107 Pa 0.36 mm
185 The MB system is composed of two radial Hybrid Mag-
M B2 rotor 8.5 kg
Aux. bearing 0.4 kg
netic Bearing (HMB) and one axial Active Magnetic Bear-
Axial Loads -1500 N ing (AMB)
R1 340 N
R2 1420 N
Blade+ Hub 15.45 kg 2.2.1. Radial HMB Analysis:
Shaft 65 kg
Generator 85 kg The design of Radial HMB is divided into two parts.
Mtilt 0 190 The first part is the design of permanent magnet (PM) to
M B1 rotor 5 kg
5 6.86 × 106 Pa 0..0233 mm
M B2 rotor 8.5 kg compensate the static loads passively in vertical direction.
Aux. bearing 0.4 kg
Axial Loads 1500 N
The second part is the design of electromagnet to compen-
R1 650 N sate and control the aerodynamic forces along vertical and
R2 1110 N
Blade+ Hub 15.45 kg
horizontal direction. The design of the first and second
Shaft 65 kg 195 parts depend on the calculated reactions which obtained
Generator 85 kg
Mtilt 0 from the shaft free body diagram as in Table 5. The con-
6
M B1 rotor 5 kg
7.33 × 106 Pa 0.0236 mm figurations of the proposed RHPBHMB for both MB1 and
M B2 rotor 8.5 kg
Aux. bearing 0.4 kg MB2 and the forces direction are shown in Figs. 7 and 6
Axial Loads -1500 N respectively. Each MB consists of four identical PM pole
R1 650 N
R2 1110 N 200 pairs and four identical electromagnetic pole pair. Fur-
thermore, each pole pair has its separated magnetic path
and isolated from the others by a nonmagnetic material.

4
Table 5: Bearing design load
Bearing Static Load (N) Dynamic Load (N)
MB1 650 310
MB2 1110 310

So, it is preferable for the magnetic flux to confine the


low reluctance iron path rather than high reluctance path
205 of nonmagnetic material. The PM material is NdfeB-N40
grade 40 and the steel material is Silicon Iron M-19.
The equivalent magnetic circuits of the bias and control
flux are shown in Fig. 8. This analysis assumes that there
is no leakage flux or fringing at air-gap. In addition, The
210 flux density is assumed to be constant around the magnetic
core. (a) MB1

(b) MB2

(c) Legend

Figure 6: Force directions of MB Figure 7: Proposed design of MB1 & MB2 (mm).

Let FP M i is the MMF of the ith PM pole. <P M i Reluctance of ith PM.
2FPM1 <s Reluctance of iron core of stator.
= φP M 1 (2<P M 1 + 2<pg1 + <s ) + φP M t <r
σP
2FPM2 <r Reluctance of iron core of rotor.
= φP M 2 (2<P M 2 + 2<pg2 + <s ) + φP M t <r
σP σP M Leakage coefficient of PM pole.
(1)
2FPM3
= φP M 3 (2<P M 3 + 2<pg3 + <s ) + φP M t <r By ignoring the core reluctance, the eqn. 1 can rewritten
σP
as:
2FPM4
= φP M 4 (2<P M 4 + 2<pg4 + <s ) + φP M t <r 2FPM1
σP = φP M 1 (2<P M 1 + 2<pg1 )
σP
where 2FPM2
= φP M 2 (2<P M 2 + 2<pg2 )
σP
φP M i Flux of ith PM. (2)
2FPM3
= φP M 3 (2<P M 3 + 2<pg3 )
φP M t Summation of all fluxes of all PM poles. σP
2FPM4
215 <pgi Reluctance of air gap related to the ith PM pole. = φP M 4 (2<P M 4 + 2<pg4 )
σP
5
2FC 1
= φP 1 (2<P 1 + 2<cg1 + <s ) + φP t <r
σc
2FC 2
= φP 2 (2<P 2 + 2<cg2 + <s ) + φP t <r
σc
(5)
2FC 3
= φP 3 (2<P 3 + 2<cg3 + <s ) + φP t <r
σc
2FC 4
= φP 4 (2<P 4 + 2<cg4 + <s ) + φP t <r
σc
where
(a) Magnetic circuit of Bias Flux
φP i Flux of ith electromagnetic pole.
φP t Summation of all fluxes of all electromagnetic poles.
230 <cgi Reluctance of air gap related to the ith electromag-
netic pole.
<P i Reluctance of ith electromagnetic pole.
σc Leakage coefficient of electromagnetic pole.
By ignoring the core reluctance, the eqn. 5 can rewritten
as:
2FC 1
(b) Magnetic circuit of Control Flux = φP 1 (2<cg1 )
σc
2FC 2
Figure 8: Equivalent magnetic circuits.
= φP 2 (2<cg2 )
σc
(6)
2FC 3
= φP 3 (2<cg3 )
At balance condition, the PM poles fluxes can be written σc
as follows: 2FC 4
= φP 4 (2<cg4 )
σc
<g =<pg1 = <pg2 = <pg3 = <pg4
At balance condition, the electromagnetic poles fluxes can
    be written as follows:
φP M 1 FP M 1
φP M 2  1 FP M 2 
   
φP M 3  = σP (<P M + <g ) FP M 3  (3) <g =<cg1 = <cg2 = <cg3 = <cg4
φP M 4 FP M 4
and    
220 where <g is the reluctance of air gap at equilibrium. φC1 FC1
Due to the symmetry between all PM poles, the total PM φC2  1 FC2 
   
force can be deduced as follows: φC3  = σc (<g ) FC3  (7)
T otalP M φC4 FC4

FP M = The magnetic force FCi for ith electromagnetic pole can
Z−axis
be calculated according to:
FP M 1 cos(30)+FP M 2 cos(30)+FP M 3 cos(60)+FP M 4 cos(60)
225 ∵ FP M = FP M 1 = FP M 2 = FP M 3 = FP M 4 Bg2 φCi 2 (N i)2 µo Ag
T otalP M FCi = Ag = = (8)
√ 2µo 2µo Ag 2lg2
∴ FP M = (1 + 3)FP M (4)
Z−axis Where
Where:
FP2 M 235 φCi The magnetic flux of ith electromagnetic pole.
FP M =
µo σP2 Ag (<P M + <g )2
Bg The flux density of air gap.
Similarly, Let FCi is the MMF of the ith coil pole.
µo The permeability of free space (4π × 10−7 H/m).
Ag Cross section area of air gap.
6
lg Length of air gap. 2.2.2. Axial AMB Analysis:
The configuration of the AAMB is shown in Fig. 9.
240 N Number of coil’s turns.
250 The inner and outer air gap is 1 mm. It has two sym-
i Coil current. metrical parts to generate a force along the positive and
The position stiffness of the PM poles along z-axis can negative x-axis. The stator and rotor are made of Silicon
be calculated as follows: iron M-19. Both of electromagnetic coils has 100 turns of
√ copper with ampacity Jmax = 6A/mm2 and the bearing
−2(1 + 3)FP2 M is designed to compensate axial force up to 4000 N. The
∴ KP M Z = 2 2 2 (9)255
µo σP Ag (<P M + <g )3 dimension and the equivalent magnetic circuit of the Axial
The forces of electromagnetic poles can be deduced as fol- AMB is shown in Fig. 9.
lows: The fluxes can be obtained as follows:
T otal FA1

FC = FC2 − FC1 φA1 =
σA (<gin + <Svin + <Svout + <gout )
Z−axis
T otal (13)
FA2
FC = FC3 − FC4 φA2 =
245 σA (<gin + <Svin + <Svout + <gout )
T otal Y −axis
FP 2 FP 2 Using equation 8, the force of the axial bearing are given
∴ FC = σ2 µo A g (<g3 )
2 − σ 2 µ A (<
g4 )
2
Y −axis
c c o g by:
The position stiffness of electromagnetic poles is: φ2A1 φ2A1
2 FA1 = +
4FP 2µo Agin 2µo Agout
KCY =
σc2 µ2o A2g (<g )3 φ2A2 φ2A2 (14)
(10) FA2 = +
4FP 2 2µo Agin 2µo Agout
Similarly, KCZ =
σc2 µ2o A2g (<g )3 FAt = FA2 − FA1
The total position stiffness along z-axis is equal to the sum- φAi Flux of ith half of AAMB.
mation of both position stiffness of PM poles and electro-
magnetic poles: FAi MMF of ith half of AAMB.
T otal
FP 2 FP 2 260 FAi Force of ith half of AAMB.
F = 2
− 2
Z−axis σc µo Ag (<g2 ) σc µo Ag (<g1 )
√ FAt Net force of AAMB.
(1 + 3)FPM 2
− (11) <Svin , <Svout Reluctance of inner and outer core respec-
σP µo Ag (<P M + <g1 )2
√ tively.
4FP 2 2(1 + 3)FPM 2
Kz = + <gin , <gout Reluctance of inner and outer air gap respec-
σc µ2o A2g (<g )3 σP µ2o A2g (<P M + <g )3
265 tively.
The current stiffness of electromagnetic poles can be cal-
culated as follows: Agin , Agout Cross section area of inner and outer air gap
4N i 2 respectively.
KCY =
σc2 µo Ag (<g )2 σA Leakage coefficient of AAMB.
(12)
4N 2 i
Similarly, KCZ = 2 Lg Air gap length.
σc µo Ag (<g )2
The parameters of radial HMB is given in Table 6. The position stiffness can be calculates as follows:
2Agin Agout µo FA 2 (Agin +Agout )2
Kx = 2 (A
σA gin Lg +Agout Lg +Agin Agout Rsvin µo +Agin Agout Rsvout µo )
3
Table 6: The parameter of radial MB
Parameter MB1 MB2 (15)
The current stiffness can be calculated as follows:
σP 1.2 1.25
σc 1.11 1.15 2N 2 i(Agin + Agout )
<P M (At/wb) 16293503.59 8961426.97 Ki = 2 (R 2
Agin Agout µo σA svin + Rsvout + Rgin + Rgout )
<g (At/wb) 723431.55 397887.35 (16)
Ag (mm2 ) 1100 2000
lg (mm) 1 1 270 The parameters for the AAMB are given in Table 7.
FPM (AT) 11500 11500
N (turns) 100 100

7
Table 7: The parameter of AAMB
Parameter Value Parameter Value
µr (steel) 8000 σA 1.1
<gin (At/wb) 24616.419 <gout (At/wb) 4180.76
<svin (At/wb) 127.7 <svout (At/wb) 21.7
N (turns) 100 Agin (mm2 ) 32327
Aout (mm2 ) 190342

2.2.3. Finite Element Analysis:


The proposed designs of RHPBHMB and AAMB are
analysed by 3D FEM and 2D FEM analysis respectively.
All these FEM models are constructed using MSC Marc
275 package as shown in Fig. 10. The position and current
stiffness of electromagnetic poles of the MB1, MB2 and
(a) Diamension of the upper half of
axial MB (mm). AAMB are shown in Fig. 11. It is clear that there is a
great consistent between the analytical and FEM results.
A comparison between the analytical and FEM results is
280 given in Table 8

3. Mathematical Dynamic Model of Overall Sys-


tem
(b) Legend
This section presents the mathematical model for each
component of MLWT and describe the interaction between
285 them. The system is composed of three coupled models:
aerodynamic model, mechanical model, permanent mag-
net synchronous generator model. The mechanical model
includes the MB and CB model.

3.1. Modeling of Aerodynamic:


290 This model aims to find the relationship between the
shape of airfoil and its aerodynamic properties. The aero-
dynamic axial forces, spining torque and bending moment
can be determined in terms of the aerodynamic lift (FL )
and drag (FD ) forces. So, the Blade element theory can be
295 used to obtain these forces depending on the two-dimensio-
nal aerofoil characteristics of NACA 0015 as in [26]. The
(c) The direction of flux through the Axial wind turbine aerodynamic parameters can be summarized
MB
in Table 9.

Table 8: Comparison between analytical and FEM results


MB Parameter Analytical FEM Designed
KCY = KCZ (N/mm) 403886.288 396666.66
KCY = KCZ (N/A) 67.314 67.33
T otal
MB1
FP M (N ) 626.84 617.13 650
Z−axis
FC1 = FC2 = FC3 = FC4 (N ) 403.88 394.7 310
Max. Capacity at 6A (N ) 1030.72 1029.9 960
KCY = KCZ (N/mm) 684142.71 675000
KCY = KCZ (N/A) 114.024 110.66
T otal
MB2
FP M (N ) 1139.7 1138.9 1110
Z−axis
FC1 = FC2 = FC3 = FC4 (N ) 475.09 443.1 310
(d) Magnetic circuit Max. Capacity at 5A (N ) 1614.8 1551.3 1420
Kx (N/mm) 4263636.441 4900000
Axial Ki (N/A) 1408 1520
Figure 9: Equivalent magnetic circuit of the axial MB. FA1 = FA2 = (N ) 4236.9 3795.6 4200

8
(a) Force displacement stiffness (b) Current force stiffness of
of MB1 MB1

(a) The FEM model of MB1 with coil current of 6 A

(c) Force displacement stiffness (d) Current force stiffness of


of MB2 MB2

(e) Force displacement stiffness (f) Current force stiffness of


of AAMB AAMB
(b) The FEM model of MB2 with coil current of 5 A
Figure 11: The comparison between FEM and magnetic circuit anal-
ysis

Table 9: Wind turbine aerodynamic parameter


Parameter Value Parameter Value
Rblade 1m β 20o
ρ 1.22 kg/m3 α 7.677o
Rated wind
cl 0.8445 speed 11 m/s
cd 0.008871 Max. wind speed 16 m/s
Lchord 0.208 m

3.2.1. Rotor Motion Equation:


Fig. 12 shows a simplified horizontal shaft suspended
(c) The FEM model of AAMB with coil 305 by MB. In order to fully position the shaft, one has to
current of 5 A apply forces along five axes. Two radial and one thrust
MB are needed in this case. In such a system we have five
Figure 10: The FEM models of the three MB. degrees of freedom (DOF): three linear motion: vertical,
horizontal, and axial, and two rotational motions: pitching
310 and yawing. The remaining DOF, which is rolling motion,
3.2. Modeling of RHPBHMB:
is controlled by load torque. The equations which describe
300 The magnetic bearing is considered to be a rigid float- the 5-DOF motion can be written as follows:
ing body. So, the theory of flight dynamics is proper to
describe its dynamic model. • Axial direction:
9
- Linear motion
X
mx¨o + βxo + 2γ x˙o = Fa (17)

• Radial direction:
315 - Linear motion
1. Horizontal
X X
my¨o − αyo = Fy1 + Fy2 (18)

2. Vertical
X X (a) 3D
mz¨o − αzo = − Fz1 − Fz2 (19)

- Rotational motion
1. Pitching
X X
Jy θ̈ + wr Jx ψ̇ = l1 Fz1 − l2 Fz2 (20)

2. Yawing
X X
Jy ψ̈ − wr Jx θ̇ = l1 Fy1 − l2 Fy2 (21)

3.2.2. Gap Deviation:


(b) 2D
The ith gap deviation, gi0 , between the stator and rotor
of MB can be expressed in terms of yo , zo , θ, ψ as follows:

gi = Do + gi0 i = 11, 13, 21, 23 (22)


 0   0   
g11 g12 zo − l1 θ
g21
0   0   zo + l2 θ 
 0  = − g22  
0 =
 (23)
g13  g14 −yo − l1 ψ 
0 0
g23 g24 −yo + l2 ψ
where
0 0 (c) Gap deviation of MB1 (d) Gap deviation of MB2
g11 , g12 The vertical gap deviation of MB1 in +ve and -ve
320 direction respectively.
Figure 12: Active magnetic bearing system
0 0
g13 , g14 The horizontal gap deviation of MB1 in +ve and
-ve direction respectively.
3.2.3. Magnetic Force:
0 0 The forces in HMB is induced from the biased and
g21 , g22 The vertical gap deviation of MB2 in +ve and -ve
direction respectively. control fluxes. The total forces of biased fluxes has been
0 0
derived in equation 4 and can be rewritten as follows:
325 g23 , g24 The horizontal gap deviation of MB2 in +ve and
T otal √
-ve direction respectively. 4(1 + 3)FPM 2

FP M = 2
zo The rotor centre of mass coordinate along z axis. Z−axis σP µo AP M (2<P M + 2<g + <s + <r )2

yo The rotor centre of mass coordinate along y axis. Furthermore, the force due to control flux for the proposed
homopolar MB arrangement is equal to:
l1 The distance between rotor centre of mass and MB1.
Homo
φ2 (2N i)2
330 l2 The distance between rotor centre of mass and MB2. Fm = = (24)
pole µo Ag µo Ag (2<g + <s + <r )2
θ The pitch angle around y axis
ψ The yaw angle around z axis.
10
3.2.4. Electric circuit equation: 1 N2
x˙15 = (u7 − Re2 x15 ), L23 = o −x4
The electric circuit equation for any electromagnetic L23 <c2 + 2 D
µo Ag2
pole can be specified in terms of current as follows: 1 N2
x˙16 = (u8 − Re2 x16 ), L24 =
di L24 <c2 + 2 Do +x4
µo Ag2
E = Ri + L
dt The summation of forces along the horizontal and vertical
(25)
(2N )2 directions can be expressed as follows:
L=
<c + 2<g X
Fy1 = F13 − F14
335 where <c is total constant reluctance of the magnetic path. X
Fy2 = F23 − F24
3.2.5. State space representation: X (27)
Fz1 = F11 − F12 − F R1 + FP M 1
The state space can be obtained from the equations
X
derived in the previous sections. The definition of states Fz2 = F21 − F22 − F R2 + FP M 2
can be written as follows:
The magnetic forces are expressed as follows:
u1 = e11 u2 = e12 u3 = e21 u4 = e22
 2
u5 = e13 u6 = e14 u7 = e23 u8 = e24 x9
F11 = k1 o −x1
0
x1 = g11 0
x2 = g21 0
x3 = g13 0
x4 = g23 <c1 + 2 D
µo Ag1
 2
x5 = g 0˙
11 x6 = g 0˙
21 x7 = g 0˙
13 x8 = g 0˙
23 x10
F12 = k1
x9 = i11 , x10 = i12 , x11 = i21 , x12 = i22 <c1 + 2 Do +x1
µo Ag1
x13 = i13 , x14 = i14 , x15 = i23 , x16 = i24  2
x11
F21 = k2 o −x2
x˙1 = x5 , x˙2 = x6 , x˙3 = x7 , x˙4 = x8 <c2 + 2 D
µo Ag2
   2
α l2 x1 + l1 x2 w r l1 J x x12
x˙5 = − (x7 − x8 ) F22 = k2
m l1 + l2 Jy (l1 + l2 ) <c2 + 2 Do +x2
µo Ag2
X X
− Hs11 Fz1 − Hd12 Fz2  2
x13
  F13 = k1 o −x3
α l2 x1 + l1 x2 w r l2 J x <c1 + 2 D
µo Ag1
x˙6 = − (x7 − x8 ) (28)
m l1 + l2 Jy (l1 + l2 )  2
X X x14
− Hd12 Fz1 − Hs22 Fz2 F14 = k1
<c1 + 2 Do +x3
µo Ag1
 
α l2 x3 + l1 x4 w r l1 J x  2
x˙7 = − (x5 − x6 ) x15
m l1 + l2 Jy (l1 + l2 ) F23 = k2 o −x4
X X <c2 + 2 D
µo Ag2
− Hs11 Fy1 − Hd12 Fy2  2
  x16
α l2 x3 + l1 x4 w r l2 J x F24 = k2
x˙8 = − (x5 − x6 ) <c2 + 2 D o +x4
µo Ag2
m l1 + l2 Jy (l1 + l2 ) √
X X 4(1 + 3)FP2 M 1
− Hd12 Fy1 − Hs22 Fy2 (26) FP M 1 = 2 o −x1 2
σP 1 µo AP M 1 (<P M 1 |T C + 2 D
µo Ag1 )
2 √
1 N
x˙9 = (u1 − Re1 x9 ), L11 = o −x1
4(1 + 3)FP2 M 2
L11 <c1 + 2 D
µo Ag1 FP M 2 = 2 o −x2 2
σP 2 µo AP M 2 (<P M 2 |T C + 2 D
µo Ag2 )
2
1 N
x˙10 = (u2 − Re1 x10 ), L12 =
L12 <c1 + 2 Do +x1 4N 2 4N 2
µo Ag1 Where k1 = µo Ag1 , k2 = µo Ag2 , FP M = Hm lm , Hs11 =
1 N 2 l12 l2
x˙11 = (u3 − Re2 x11 ), L21 =
1
m + Jy , Hs22 = m1
+ J2y , and Hd12 = m1
− lJ1 ly2 .
o −x2
L21 <c2 + 2 D
µo Ag2
340 The HMB parameters can be summarized in Table 10:
1 N2
x˙12 = (u4 − Re2 x12 ), L22 =
L22 <c2 + 2 Do +x2
µo Ag2 3.3. Modeling of AAMB:
1 N2 The axial equation of motion (Eqnation 17) is indepen-
x˙13 = (u5 − Re1 x13 ), L13 = o −x3
L13 <c1 + 2 D
µo Ag1
dent and simpler than the other radial motion equations.
It can be rewritten as follows:
1 N2
x˙14 = (u6 − Re1 x14 ), L14 =
L14 <c1 + 2 Do +x3
µo Ag1 mx¨o = Fa1 − Fa2 (29)
11
Table 10: The HMB parameters. Table 12: The generator parameters.
Parameter Value Parameter Value Parameter Value Parameter Value
l1 674.5 mm l2 132 mm
m 179.35 kg h 20 mm Rs 2.4Ω Ls 51 mH
Jx 7.66 kg.m2 Jy 39.32 kg.m2 Vrated 220 V Prated 3 kW
Do 1 mm N 100 turns N 375 rpm poles 16
<c1 30625.948 AT/wb <c2 38697.114 AT/wb
<P M 1 | T C 32613111.68 AT/wb <P M 2 |T C 17959064.27 AT/wb
Re2 0.55776 Ω Re1 0.86016 Ω 350 3.5. Analysis of Conventional Mechanical Bearing (CB)
F R1 650 N F R2 1110 N
lm 25 mm Ag2 = AP M 2 2000 mm2 The CB is considered here to provide a fair compar-
Ag1 = AP M 1 1100 mm2
ison between the performance of WT-CB and WT-MB.
The frictional moment for the CB is obtained from the
SKF Rolling bearings catalog [28]. In this work, two con-
assuming β and γ are zero. The state space representation
355 ventional SKF bearings, CB1 and CB2, are located in the
can be written as follows:
same position of MB1 and MB2 respectively. Their desig-
ua1 = ea1 , ua2 = ea2 nation number is 22211EK. The grease is used as lubricant
at operating temperature 50C o . The grade viscosity for
x1 = 0
ga1 , ˙ 0
x2 = ga1
the two selected conventional bearing CB1 and CB2 are
x3 = ia1 , x4 = ia2 360 ISO VG 100 and ISO VG 150 respectively.

3.6. Simulation of Multi-domain Physical Systems:


1 The wind turbine is a multi-disciplinary system. As
x˙1 =x2 x˙2 = (Fa1 − Fa2 )
m aforementioned, it includes aerodynamic, mechanical and
1 N2 electrical systems. Simscape is able to create models con-
x˙3 = (u1 − Raxial x3 ), LA1 = Do −x1 Do −x1
LA1 <ca1 + + 365 sist of multi-physical domain under the Simulink environ-
µo Agin µo Agout
2
ment which helps to provide more realistic simulation and
1 N contributes in studying the interactions between the dif-
x˙4 = (u2 − Raxial x4 ), LA2 = Do +x1 Do +x1
LA2 <ca2 + µo Agin + µo Agout ferent physical systems comprehensively. Fig.13 shows the
(30) sub-systems of the wind turbine and the interaction be-
370 tween them. The aerodynamic model convert the wind
The magnetic forces are expressed as follows: speed into three components: spin torque, bending mo-
 2  2 ment and axial forces. All these components are applied
N2 x3 N2 x3 to the shaft which is magnetically levitated. According
Fa1 = Do −x1
+ Do −x1
2µo Agin µo Agin
2µo Agout µo Agout to the shaft position, the air-gaps are measured by the
 2  2 (31)375 MB1, MB2, and AAMB to generate the required forces
N2 x4 N2 x4
Fa2 = + that compensate the reactions, R1 , R2 , and Faxial . The
2µo Agin Do +x1 2µo Agout Do +x1
µo Agin µo Agout generator whose rotor is fixed on the shaft rotates with
speed w caused by aerodynamic spinning torque. As well,
where <ca1 , <ca2 are the lift and right constant reluctance the electromagnetic torque which is induced due to elec-
of AAMB whose parameters are given in Table 11. 380 trical load on the generator is reflected to the shaft and
subsequently to the aerodynamic subsystem. Fig.14 shows
Table 11: The axial AMB parameters. the visualization of the whole system whose animation can
Parameter Value Parameter Value be seen in [29].
m 179.35 kg <ca1 = <ca2 149.4 At/wb
Agin 32327 mm2 Agout 190342 mm2 4. Controller Design
N 100 turns Do 1 mm
Raxial 0.007389 Ω 385 The design of the TSK fuzzy Q-parameterization con-
troller is discussed in details in [30, 31, 32]. So, This section
3.4. Modeling of Generator only focuses on the linearization process and the applica-
tion of the controller to the previous MB dynamic model
345 The permanent magnet synchronous generator (PMSG) in eqn 26 . The controller objective are: achieve robust
is used to supply an electrical load . The parameter of gen-390 stability, overcome model nonlinearity by increasing the
erator is obtained from [27]. The Simulink PMSM block is dynamic operating range of gap displacement, reject step
operated as generator with the parameters given in Table. and unbalance sinusoidal disturbance at different speeds.
12.

12
and electromagnetic voltage of ith pole and i0i , e0i are the
deviation of them from their nominal values. Then we can
write:
Ii = Ioi + i0i
(32)
ei = Eoi + e0i
for i=11,13,21,23.
Assume:
0 0 0 0 0 0 0 0
I11 = −I12 , I13 = −I14 , I21 = −I22 , I23 = −I24
e011 = −e012 , e013 = −e014 , e021 = −e022 ,
e023 = −e024
(33)
So, the linearized model of each HMB subsystem described
in equation. 26 can be written as follows:
First subsystem (vertical motion of MB1):
x˙1 = x5
x˙5 = Pˆ1 x1 + Ĝ1 x9
(34)
−Re1 2Do
x˙9 = (<c1 + )x9
N2 µo Ag1
where
0
x1 = g11 0˙ , x = i0
, x5 = g11 9 11 √
4A N 2µ 16(1+ 3)A F 2
1 µo

Pˆ1 = −Hs11 (2D+Ag1g1 <c1oµo )3 (Io11
2 2
+Io12 )+ (σ2 (2D+Ag1g1<PPMµ
M1 o )3)
P1
−(2Ag1 Hs11 N 2 µo (Io11 +Io12 ))
Ĝ1 = (2D+Ag1 <c1 µo )2
Figure 13: The Interaction between the different sub-systems
Second subsystem (vertical motion of MB2):
x˙2 = x6
x˙6 = Pˆ2 x2 + Ĝ2 x11
(35)
−Re2 2Do
x˙11 = (<c2 + )x11
N2 µo Ag2
where
0
x2 = g21 0˙ , x
, x6 = g21 0
10 = i21 √
2
16(1+ 3)A F 2 
4A N µ 2 µo
Pˆ2 = −Hs22 (2D+Ag2g2 <c2oµo )3 (Io21
2 2
+Io22 )+ (σ2 (2D+Ag2g2<PPM 3
M 2 µo ) )
P2
−(2Ag2 Hs22 N 2 µo (Io21 +Io22 ))
Ĝ2 = (2D+Ag2 <c2 µo )2

Third subsystem (horizontal motion of MB1):


x˙3 = x7
x˙7 = Pˆ3 x3 + Ĝ3 x13
(36)
Figure 14: The Visualization of the whole system using Simscape. −Re1 2Do
x˙11 = 2
(<c1 + )x11
N µo Ag1
4.1. System Decoupling and Linearization: where
0
x3 = g13 0˙ , x
, x7 = g13 0
In order to simplify the controller design, the following 13 = i13
2 2 2
−(4Ag1 Hs11 N µo (Io13 +Io14 ))
395 assumptions are made to decouple the system into four Pˆ3 = (2D+Ag1 <c1 µo )3

subsystems: −(2Ag1 Hs11 N 2 µo (Io13 +Io14 ))


Ĝ3 = (2D+Ag1 <c1 µo )2
1. The gyroscopic effect is neglected, hence, the vertical
and horizontal motion are decoupled.
Fourth subsystem (horizontal motion of MB2):
2. Assume the left and right bearing are located at their
400 center of percussion, So, the left and right bearing x˙4 = x8
are decoupled (Hd12 u 0)
x˙8 = Pˆ4 x4 + Ĝ4 x15
Then, each subsystem is linearized at different operating (37)
points. Let Ioi , Eoi are the nominal value of the current −Re2 2Do
x˙15 = (<c2 + )x15
N2 µo Ag2
13
and second set of TSK fuzzy rules determines the most
suitable Q-parametrization controller and the steady state
420 input voltage (E1,E2) to drive the HMB system based on
gap deviation and current deviation. The xi represent the
gap deviation for i = 1, 2, 3, 4 and uj represent the voltage
deviation for j = 1, 3, 5, 7. So, the rules can be written as
(a) Gap Deviation (b) Current Devia- (c) Voltage Devia- follows:
of 1st subsystem tion of 1st subsys- tion of 1st subsys-
tem tem 425 Plant Rule 1:
If xi = ZE then ẋ = A1 x + B1 u and y = c1 x.

Plant Rule 2:
If xi = P OS and xj = ZE then ẋ = A2 x+B2 u and y =
c2 x.
430 Plant Rule 3:
If xi = N E and xj = ZE then ẋ = A3 x+B3 u and y =
(d) Gap Deviation (e) Current Devia- (f) Voltage Devia-
of 2nd subsystem tion of 2nd subsys- tion of 2nd subsys- c3 x.
tem tem
Plant Rule 4:
If xi = P OS and xj = N E then ẋ = A4 x+B4 u and y =
435 c4 x.
Plant Rule 5:
If xi = N E and xj = P OS then ẋ = A5 x+B5 u and y =
c5 x.
(g) Gap Deviation (h) Current Devia- (i) Voltage Devia-
Plant Rule 6:
of 3rd subsystem tion of 3rd subsys- tion of 3rd subsys-
tem tem 440 If xi = P OS and xj = P OS then ẋ = A6 x +
B6 u and y = c6 x.

Plant Rule 7:
If xi = N E and xj = N E then ẋ = A7 x+B7 u and y =
c7 x.
Where for the 1st subsystem:
(j) Gap Deviation of (k) Current Devia- (l) Voltage Devia-
   
0 1 0 0 1 0
4th subsystem tion of 4th subsys- tion of 4th subsys-
A1 = 1088.56 0 0.83  A2 = 728149.28 0 40.03 
tem tem
0 0 −20.6 0 0 −4.46
   
Figure 15: Membership functions of HMB system 0 1 0 0 1 0
A3 = 702662.13 0 39.35  A4 =  274229.65 0 −22.33
0 0 −36.74 −65891.74 0 −4.46
   
where 0 1 0 0 1 0
0
x4 = g23 0˙ , x
, x8 = g23 0
15 = i23 A5 = 4861026.12 0 101.71  A6 = 4835090.3 0 101.44
2 2 2
−(4Ag2 Hs22 N µo (Io23 +Io24 ))
405 Pˆ4 = (2D+Ag2 <c2 µo )3

−65891.74 0 −36.74

64880.97 0 −4.46
−(2Ag2 Hs22 N 2 µo (Io23 +Io24 )) 0 1 0
Ĝ4 = (2D+Ag2 <c2 µo )2 A7 = 268071.33 0 −22.06
64880.97 0 −36.74
 
4.2. Application of TSK Fuzzy Q-parameterization Con- 0  
trollers to HMB: Bi =  0  ci = 1 0 0
410 For the HMB system described in equations 26. The 36.94
controller is designed for the four subsystems described in for i=1,2,3,4,5,6,7.
linear equations 34, 35, 36, and 37. The HMB subsys- For the 2nd subsystem:
tems have the same sequence of rules sets to describe the
   
non-linearity of each subsystem but with different mem- 0 1 0 0 1 0
415 bership functions, linearization values and applied input A1 = 714.3 0 0.53  A2 = 331419.16 0 20.41 
voltages. The membership functions of these rules at dif- 0 0 −17.94 0 0 −4.25
ferent operating points are defined in Fig.15. The first

14
   
0 1 0 0 1 0 Plant Rule 3:
A3 = 315952.65 0 19.94  A4 =  111901.72 0 −10.46 If xi = N E and uj = ZE then E1 = em3 and E2 =
0 0 −31.63 −54952.86 0 −4.25
    en3 .
0 1 0 0 1 0
A5 = 2138818.85 0 50.8  A6 = 2127382.3 0 50.67  455 Plant Rule 4:
−54952.86 0 −31.63 53883.33 0 −4.25 If xi = P OS and uj = N E then E1 = em4 and E2 =
 
0 1 0 en4 .
A7 = 109282.09 0 −10.32
53883.33 0 −31.63 Plant Rule 5:
  If xi = N E and uj = P OS then E1 = em5 and E2 =
0   460 en5 .
Bi =  0  ci = 1 0 0
20.86 Plant Rule 6:
for i=1,2,3,4,5,6,7. If xi = P OS and uj = P OS then E1 = em6 and E2 =
For the 3rd subsystem: en6 .
   
0 1 0 0 1 0 Plant Rule 7:
A1 = 500.8 0 0.68  A2 = 702318.92 0 39.33  465 If xi = N E and uj = N E then E1 = em7 and E2 =
0 0 −20.6 0 0 −4.46 en7 .
   
0 1 0 0 1 0
A3 = 702318.92 0 39.33  A4 =  233538.04 0 −20.51
The values of em and en for 1st, 3rd and 2nd, 4th subsys-
0 0 −36.74 −63222.58 0 −4.46 tems are given in Tables 13 and 14 respectively.
   
0 1 0 0 1 0 4.3. Validation of HMB control system
A5 = 4619512.28 0 99.17  A6 = 4619512.28 0 99.17 
470 The proposed TSK fuzzy Q-parameterization controller
−63222.58 0 −36.74 63222.58 0 −4.46
  is designed based on the linear system described in equa-
0 1 0
tions 34, 35, 36, and 37 then applied to a complete magnet-
A7 = 233538.04 0 −20.51
63222.58 0 −36.74
ically levitated system constructed by Simscape package.
  Simulation results are obtained at the rated speed of the
0   475 system (31.41 rad/s = 300 rpm).
Bi =  0  ci = 1 0 0 Fig. 16 shows the step response of g11 0 0
, g13 0
, g21 , and
36.94 0
g23 . The overshoot of these deviations are 14.4%, 14.7%,
for i=1,2,3,4,5,6,7. 13.3% and 13.7% respectively with zero steady state error.
For the 4th subsystem: The settling time for all of them is less than 0.2s.
   
0 1 0 0 1 0
A1 = 295.17 0 0.41  A2 = 315722.39 0 19.93 
0 0 −17.94 0 0 −4.25 Table 13: Values of em and en for first and third subsystem of HMB
    system.
0 1 0 0 1 0 1st subsystem
A3 = 315722.39 0 19.93  A4 =  91962.78 0 −9.4  Rules emi eni
0 0 −31.63 −52220.02 0 −4.25
    1 e11 = 0.6V e12 = 0.42V
0 1 0 0 1 0 2 e11 = 2.27V e12 = 2.23V
A5 = 2009526.92 0 49.26  A6 = 2009526.92 0 49.26 
−52220.02 0 −31.63 52220.02 0 −4.25
3 e11 = 2.27V e12 = 2.23V
  4 e11 = 0.45V e12 = 4.053V
0 1 0
5 e11 = 0.45V e12 = 4.053V
A7 = 91962.78 0 −9.4 
52220.02 0 −31.63 6 e11 = 4.065V e12 = 0.437V
  7 e11 = 4.065V e12 = 0.437V
0   3rd subsystem
Bi =  0  ci = 1 0 0 Rules emi eni
36.94 1 e13 = 0.42V e14 = 0.42V
445 for i=1,2,3,4,5,6,7. 2 e13 = 2.23V e14 = 2.23V
The following is the second set of rules: 3 e13 = 2.23V e14 = 2.23V
4 e13 = 0.48V e14 = 3.98V
Plant Rule 1:
5 e13 = 0.48V e14 = 3.98V
If xi = ZE then E1 = em1 and E2 = en1 .
6 e13 = 3.98V e14 = 0.48V
Plant Rule 2: 7 e13 = 3.98V e14 = 0.48V
450 If xi = P OS and uj = ZE then E1 = em2 and E2 =
en2 .
15
Table 14: Values of em and en for second and forth subsystem of
HMB system.
2nd subsystem
Rules emi eni
1 e21 = 1V e22 = 0.65V
2 e21 = 3.52V e22 = 3.44V
3 e21 = 3.52V e22 = 3.44V
4 e21 = 0.76V e22 = 6.2V
5 e21 = 0.76V e22 = 6.2V
6 e21 = 6.23V e22 = 0.73V
7 e21 = 6.23V e22 = 0.73V 0
(a) g11
4th subsystem
Rules emi eni
1 e23 = 0.65V e24 = 0.65V
2 e23 = 3.44V e24 = 3.44V
3 e23 = 3.44V e24 = 3.44V
4 e23 = 0.82V e24 = 6.07V
5 e23 = 0.82V e24 = 6.07V
6 e23 = 6.07V e24 = 0.82V
7 e23 = 6.07V e24 = 0.82V

480 Fig. 17 shows that the proposed controller can over-


come the non linear dynamics of HMB system by extend- 0
(b) g13
ing the gap deviation upto 800 µm. For brevity, we include
0
only the gap deviation of g11 of HMB1 as all the others
have the same response.
0 0 0 0
485 Fig. 18 shows the response of g11 , g13 , g21 , and g23 due
to step disturbance. The max. deflection of these devi-
ations are 28, 35.5, 18, 31.6 µm due to step disturbance
force of 100 N respectively.
0 0 0
Fig 19 shows the system response of g11 , g13 , g21 , and
0
490 g23 due to different vertical and horizontal initial condi-
tions. These gap displacements (initial conditions) must
be equally divided into small steps (about 50µm) in order
to drive the HMB system to zero position in a stable way.
0
(c) g21
495 The proposed controller can generate a compensating
forces to reject any imbalance sinusoidal disturbance up
to 492 rpm as shown in Fig 20. The imbalance sinusoidal
disturbance is generated by sticking a mass of 1 kg to the
shaft at the location of each MB.

500 4.4. Controller Design of Axial AMB:


The axial motion is independent of the radial motion.
A Q-parameterization controller is synthesized to stabilize
the axial motion described in equation 17 and it is con-
trolled separately. For brevity, it is omitted in this paper.
0
(d) g23
505 5. Performance Comparison between WT-MB and
WT-CB Figure 16: The HMB System response due to a step at reference.

This section presents a comparison between the wind


turbine performance under two different operating cases. case, the wind turbine is supported by magnetic bearing
In the first case, the wind turbine is supported by conven- (WT-MB). The wind turbine performance is evaluated un-
510 tional mechanical bearing (WT-CB). While in the second
16
0
(a) g11

0 Deviation for HMB system


Figure 17: The sinusoidal tracking of g11

der low wind speed and high wind speed assuming that the
operating temperature of CB is 50C o . The steady state
515 values of the WT performance are given in Table 15.

5.1. Wind Turbine Performance at Low Wind Speed:


0
(b) g13
The wind turbine performance is investigated under
different low wind speeds (1, 2, 3, 5 m/s) with no electrical
loads connected to the generator. The comparison includes
520 different aspects such as spinning torque, rotational speed,
consumed power and frictional moment.

1. Spinning Torque: The WT-MB required much low-


er aerodynamic spinning torque than WT-CB as clear
in Fig. 21. The WT-MB can rotates with only 1 m/s
525 wind speed compared with 2 m/s for WT-CB. This
mean that the MB has reduced the starting torque
and cut-in speed of wind turbine

2. Rotational Speeds: The WT-MB are higher than 0


(c) g21
those of WT-CB for the same wind speeds and the
530 difference between them decreases as the wind speed
increases. This is shown in Fig. 22.
3. Consumption Power: The WT-CB records lower
losses than the MB input power below 3 m/s. After
that the situation is reversed and the WT-CB has
535 higher losses for wind speeds over 3 m/s because the
increasing of rotational speeds as shown in Fig. 23.
4. Frictional Moment: The frictional moment of CB
is proportional to the rotational speed and at wind
speed of 1 m/s its value is bigger than aerodynamic
0
(d) g23
540 spinning torque which block the WT-CB from rota-
tion under this low wind speed (see Fig. 24).
Figure 18: The HMB System response due to a step rejection dis-
5.2. Wind Turbine Performance at high wind speed Speed: turbance.

In this section the wind turbine performance is eval-


uated at different high wind speed (9, 10, 11, 12, 13, 14,545 15, 16 m/s) with an electrical load connected to the gen-
erator. The comparison includes different aspects such as
17
(a) g11’ of MB1 (b) Net vertical force of MB1
(a) g11 (b) g12

(c) g13’ of MB1 (d) Net horizontal force of


MB1

(c) g21 (d) g22

Figure 19: The HMB System response due to different initial condi-
tions.

spinning torque, rotational speed, electromagnetic torque,


electrical generated power, consumed power, efficiency and
frictional moment. (e) g21’ of MB2 (f) Net vertical force of MB2

550 1. Spinning Torque: The spinning torque in WT-CB


is higher than that of WT-MB for all mentioned wind
speed because of the presence of frictional moment
in case of WT-CB as shown in Fig. 25
2. Rotational Speeds: The WT-MB achieves a little
555 increase in rotational speed more than that WT-CB
as shown in Fig. 26.
(g) g23’ of MB2 (h) Net horizontal force of
3. Electromagnetic Torque: The electromagnetic torque MB2
of WT-MB, which resulted from connecting electrical
load to the generator, is higher than that of WT-CB Figure 20: The HMB System response due to imbalance sinusoidal
560 as shown in Fig. 27. disturbance rejection.

4. Electrical Generated Power: The electrical net


output power of the generator, after subtracting the 7. Frictional Moment: The frictional moment of CB
MB input power in case of WT-MB, is shown in575 is proportional to the rotational speed as shown in
Fig.28. The WT-MB has higher electrical output Fig. 31).
565 power than WT-CB for the mentioned wind speeds.

6. Conclusions
5. Consumption Power: The MB of WT-MB has
lower input power than the bearing losses of WT-CB In this paper, a comprehensive assessment of the per-
for all mentioned wind speed because the increasing formance of WT-MB compared with WT-CB is provided.
570 of rotational speed. 29. 580 For this purpose, a design of two RHPBHMB and one
AAMB is developed using magnetic circuit and FEM anal-
6. Efficiency: The WT-MB has little higher efficiency ysis. Furthermore, A TSK fuzzy Q-parameterization con-
than that of WT-CB for all mentioned wind speed. troller is synthesised for RHPBHMB. As a result, this
30. study shows that the MB can successfully replace the CB
585 without decreasing the extracted power from the wind.
18
Figure 21: The wind turbine spinning torque under low wind speeds. Figure 23: The bearings power consumption under low wind speeds.

Figure 22: The wind turbine rotational speed under low wind speeds. Figure 24: The frictional moment of CB of wind turbine under low
wind speeds.

Hence, the main benefits of MB can be gained such as


avoiding wind turbine failure, increasing life span and re-600 sity. Authors sincerely acknowledge for this help and for
ducing maintenance. Besides, the obtained results prove providing all necessary equipments to complete this work.
the ability of MB to enhance the wind turbine speed, min-
590 imising starting torque, reducing cut-in speed and com-
References
pensating of wind disturbances. Moreover, a robust gain
scheduled TSK fuzzy Q-parametrization controller is de- [1] H. Bleuler, M. Cole, P. Keogh, R. Larsonneur, E. Maslen,
signed to overcome the nonlinear dynamics of HMB sys- Y. Okada, G. Schweitzer, A. Traxler, G. Schweitzer, E. H.
605 Maslen, et al., Magnetic bearings: theory, design, and applica-
tem, maximise the operating envelope up to 80% of gap
tion to rotating machinery, Springer Science & Business Media,
595 displacement and reject the step and imbalance sinusoidal 2009.
disturbance at any operating speeds. [2] H. Wu, Z. Wang, Y. Hu, Study on magnetic levitation wind
turbine for vertical type and low wind speed, in: 2010 Asia-
610 Pacific Power and Energy Engineering Conference, 2010, pp.
Acknowledgment 1–4. doi:10.1109/APPEEC.2010.5448476.
[3] B. M. Gonzlez, C. G. Garca, H. A. Coyotecatl, S. V. Limn,
Wind Turbine information is provided by RIAM Lab., Characterization of a system suspended by permanent magnets,
Research Institute for Applied Mechanics, Kyushu Univer-
19
Figure 25: The wind turbine spinning torque under high wind speeds. Figure 27: The wind turbine electromagnetic torque under high wind
speeds.

Figure 26: The wind turbine rotational speed under high wind
speeds. Figure 28: The wind turbine electrical net output power under high
wind speeds.

in: 2010 20th International Conference on Electronics Commu-


615 nications and Computers (CONIELECOMP), 2010, pp. 242– [8] M. X. Liu, Y. F. Hu, R. H. Dong, S. M. Peng, H. C. Wu,
246. doi:10.1109/CONIELECOMP.2010.5440761. Load analysis and structure design of small-scale maglev wind
[4] C. Aravind, R. Rajparthiban, R. Rajprasad, Y. Wong, A novel turbine, in: Applied Mechanics and Materials, Vol. 624, Trans
magnetic levitation assisted vertical axis wind turbinedesign635 Tech Publ, 2014, pp. 308–314.
procedure and analysis, in: Signal Processing and its Appli- [9] G. Shrestha, H. Polinder, D. Bang, J. Ferreira, Direct drive
620 cations (CSPA), 2012 IEEE 8th International Colloquium on, wind turbine generator with magnetic bearing, EWEA, 2007.
IEEE, 2012, pp. 93–98. [10] G. Shrestha, H. Polinder, D. Bang, J. Ferreira, A. Mcdonald, A
[5] V. D. D. M. Konagutti, Regenedyne maglev wind power gen- new concept for weight reduction of large direct drive machines,
eration, in: SARC-IRAJ International Conference, 16th June,640 in: Electrical Machines, 2008. ICEM 2008. 18th International
2013. Conference on, IEEE, 2008, pp. 1–6.
625 [6] N. Wei, W. Qinghai, J. Defei, H. Xiaofeng, Z. Tao, Study on [11] G. Shrestha, H. Polinder, D. Bang, A. Jassal, J. Ferreira, In-
measuring and control system of ac radial-axial hybrid magnetic vestigation on the possible use of magnetic bearings in large di-
bearing used in wind energy generation system, in: Proceedings rect drive wind turbines, Europe’s Premier Wind Energy Event,
of the 31st Chinese Control Conference, 2012, pp. 6847–6850. 645 Marseille, France.
[7] N. Wang, Y. Hu, H. Wu, J. Zhang, C. Song, Research on forces [12] G. Shrestha, H. Polinder, D.-J. Bang, J. A. Ferreira, Structural
630 and dynamics of maglev wind turbine generator, Journal of flexibility: A solution for weight reduction of large direct-drive
Magnetics 18 (4) (2013) 443–453. wind-turbine generators, IEEE Transactions on Energy Conver-

20
Figure 29: The bearings power consumption of wind turbine under Figure 31: The frictional moment of CB of wind turbine under high
high wind speeds. wind speeds.

Table 15: Comparison between WT-MB and WT-CB under different


wind speeds.
Wind speed TSpin (N.m) RPM Telecmag (N.m)
(m/s) CB MB CB MB CB MB
1 0.1599 0 0 39
2 0.5344 0 71.1 78
3 0.5513 0 112.3 117
5 0.5778 0 192 195
9 43.09 42.07 227.3 228.4 -42.48 -42.07
10 49.18 48.76 262.9 263.9 -48.55 -48.76
11 55.18 54.73 299.3 300.3 -54.53 -54.73
12 61.02 60.55 336.5 337.5 -60.36 -60.55
13 66.67 66.17 374.4 375.3 -65.99 -66.17
14 72.09 71.56 412.8 413.8 -71.39 -71.56
15 77.23 76.68 451.9 452.8 -76.51 -76.68
16 82.07 81.49 491.4 492.3 -81.34 -81.49
Wind speed MF ric Pelectric (W) PConsumption (W) η
(m/s) (N.m) CB MB CB MB CB MB
1 0.473 0 4.26
2 0.5344 3.99 4.26
Figure 30: The wind turbine efficiency under high wind speeds. 3 0.5513 6.498 4.26
5 0.5778 11.65 4.26
9 0.6121 961.3 966.74 14.61 4.26 93.7 94.69
10 0.6287 1271 1277.74 17.35 4.26 93.9 94.79
sion 25 (3) (2010) 732–740. 11 0.6455 1625 1632.74 20.29 4.26 94 94.87
650 [13] N. X. Wang, J. G. Zhang, G. P. Ding, Influence of magnetic 12 0.6624 2022 2030.74 23.4 4.26 94.04 94.8
bearing stiffness on rotor in wind turbine generator, in: Applied 13 0.6795 2460 2473.74 26.71 4.26 94.1 95.22
Mechanics and Materials, Vol. 150, Trans Tech Publ, 2012, pp. 14 0.6965 2935 2944.74 30.19 4.26 94.15 94.96
57–62. 15 0.7133 3443 3452.74 33.84 4.26 94.22 94.9
[14] H. Wu, L. Xiao, B. Wang, G. Li, P. Li, Digital platform design 16 0.7298 3980 3990 37.66 4.26 94.24 94.99
655 for magnetically suspended bearings equipped for wind turbine
based on dsp28335+fpga, in: Proceedings of 2010 IEEE/ASME
International Conference on Mechatronic and Embedded Sys-
tems and Applications, 2010, pp. 283–287. doi:10.1109/MESA. national Conference on Intelligent Robotics and Applications,
2010.5552053. Springer, 2012, pp. 34–43.
660 [15] M. R. Vorwaller, K. C. Lin, J. H. Gou, C. Ham, Y. H. Joo, [17] Y. Yu, H. Zhu, S. Zeng, A new self-decoupling magnetic levita-
Testbed for a wind turbine with magnetic bearing, in: Advanced670 tion generator for wind turbines, Progress In Electromagnetics
Materials Research, Vol. 512, Trans Tech Publ, 2012, pp. 657– Research M 40 (2014) 111–118.
660. [18] S. Zeng, Y. X. Sun, Y. Du, H. Q. Zhu, X. X. Liu, The structure
[16] G. Zhang, L. Mei, Y. Yuan, Variable universe fuzzy pid control and finite element analysis of a new type of maglev wind gen-
665 strategy of permanent magnet biased axial magnetic bearing erator, in: Applied Mechanics and Materials, Vol. 703, Trans
used in magnetic suspension wind power generator, in: Inter-675 Tech Publ, 2015, pp. 436–439.
[19] H. Ouyang, F. Liu, G. Zhang, L. Mei, X. Deng, D. Wang, Vi-

21
bration suppression in rotor system of magnetic suspended wind
turbines using double stage cross feedback control, in: 2014
IEEE/SICE International Symposium on System Integration,
680 2014, pp. 404–407. doi:10.1109/SII.2014.7028072.
[20] H. Ouyang, F. Liu, G. Zhang, L. Mei, X. Deng, D. Wang, Vibra-
tion suppression for rotor system of magnetic suspended wind
turbines using cross-feedback-based sliding mode control, in:
2015 IEEE/SICE International Symposium on System Integra-
685 tion (SII), 2015, pp. 112–115. doi:10.1109/SII.2015.7404963.
[21] A. Ragheb, M. Ragheb, Wind turbine gearbox technologies, in:
2010 1st International Nuclear Renewable Energy Conference
(INREC), 2010, pp. 1–8. doi:10.1109/INREC.2010.5462549.
[22] S. Sheng, Wind turbine gearbox reliability database, condition
690 monitoring, and operation and maintenance research update,
in: Drivetrain Reliability Collaborative Workshop, Golden, Col-
orado, 2016.
[23] S. Sankar, M. Nataraj, P. V. Raja, Failure analysis of bearing in
wind turbine generator gearbox, Journal of Information Systems
695 and Communication 3 (1) (2012) 302.
[24] Jis s45c - mild steel - an overview.
URL http://www.meadinfo.org/2010/03/
s45c-jis-mechanical-properties.html
[25] R. L. Norton, Design of machinery: an introduction to the
700 synthesis and analysis of mechanisms and machines, Vol. 924,
McGraw-Hill Boston, 1999.
[26] T. Burton, N. Jenkins, D. Sharpe, E. Bossanyi, Wind energy
handbook, John Wiley & Sons, 2011.
[27] M. Chinchilla, S. Arnaltes, J. C. Burgos, Control of permanent-
705 magnet generators applied to variable-speed wind-energy sys-
tems connected to the grid, IEEE Transactions on energy con-
version 21 (1) (2006) 130–135.
[28] S. Group, Rolling bearings catalogue, SKF, 2013.
[29] M. Fekry, Magnetically levitated wind turbine (matlab simula-
710 tion) - youtube, (Accessed on 11/03/2017) (Nov. 2017).
URL https://www.youtube.com/watch?v=
wg62iTlIFUI&feature=youtu.be
[30] M. Fekry, A. M. Mohamed, M. Fanni, An intelligent q-
parameterization control design that captures non-linearity and
715 fuzziness of uncertain magnetic bearing system, in: Conference
on Control Applications (CCA), IEEE, 2015, pp. 1078–1083.
[31] M. Fekry, A. M. Mohamed, M. Fanni, Robust q-parametrisation
control for nonlinear magnetic bearing systems with imbal-
ance based on tsk fuzzy model, International Journal of
720 Modelling, Identification and Control 29 (3) (2018) 195–208.
doi:10.1504/IJMIC.2018.091237.
URL https://www.inderscienceonline.com/doi/pdf/10.
1504/IJMIC.2018.091237
[32] M. Fekry, Design and Control of a Magnetic Bearing System for
725 a Horizontal Axis Wind Turbine, Ph.D. Thesis, Egypt-Japan
University of Science and Technology (E-JUST), Sep. 2017.

22

You might also like