research article research on adaptive dual-mode switch control...
TRANSCRIPT
Research ArticleResearch on Adaptive Dual-Mode Switch Control Strategy forVehicle Maglev Flywheel Battery
Hui Gao Ying-Jun Wu and Jing-Jin Shen
Nanjing University of Posts and Telecommunications Nanjing 210023 China
Correspondence should be addressed to Hui Gao gaohui2005163com
Received 3 November 2014 Accepted 15 January 2015
Academic Editor Honglei Xu
Copyright copy 2015 Hui Gao et alThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Because of the jamming signal is real-time changeable and control algorithm cannot timely tracking control flywheel rotor thispaper takes vehicle maglev flywheel battery as the research object One kind of dual-model control strategy is developed based onthe analysis of the vibration response impact of the flywheel battery control system In view of the complex foundation vibrationproblems of electric vehicles the nonlinear dynamic simulation model of vehicle maglev flywheel battery is solved Throughanalyzing the nonlinear vibration response characteristics one kind of dual-mode adaptive hybrid control strategy based on 119867infin
control and unbalance displacement feed-forward compensation control is presented and a real-time switch controller is designedThe reliable hybrid control is implemented and the stability in the process of real-time switch is solved The results of this projectcan provide important basic theory support for the research of vehicle maglev flywheel battery control system
1 Introduction
As the future main traffic tools electric vehicle (EV) isrequired in the performance of starting acceleration andclimbing however this performance depends largely on thepower battery performance [1] But the faults of costly shortrange and short service life become the bottleneck to restrictEV development scale Specifically in the process of frequentstart-stop or climbing the chemical battery life is moreshorten because of fast and deep discharging [2 3] How toeliminate these shortcomings becomes the key EV to bequickly developed
Maglev flywheel can be applied in EV electric powersystem aerospace and other fields because of having highspecific energy high power fast charge and discharge longservice life no waste gas pollution environment-friendlyadvantages and so on [4ndash6] In the field of EV the maglevflywheel either can be as an independent power driving EV[3] or can be used as auxiliary power assisting the motivepower batteries work [7 8] However the maglev flywheelcontrol stability will be affected because of existing start-stopacceleration and deceleration steering and road randomvibration in the process of the EV driving and even causeinstability So the efficiency of magnetic suspension flywheelmust be reduced
For the scientific research and practical application ofmaglev flywheel a dual-mode adaptive hybrid control strat-egy is studied based on 119867infin control and AILC algorithmsand the state space equation ofmaglev flywheel was analyzedTo improve the robust stability of flywheel control systemand reduce the real-time interference one 119867infin controllerbased on the state space equation was solved To reducemaglev flywheel radial run-out one adaptive iterative learn-ing control theory was deduced and unbalance displacementcompensation was implemented The experimental resultshows the dual-model control strategy has better interfer-ence capability in maglev flywheel start-up process and hasstronger active control ability relative to only PID controlThemaglev flywheel based on the dual-model control can help theEVprimary battery improve its discharge characteristics andhelp to prolong its service life
2 Maglev Flywheel Nonlinear Dynamic Model
The below force and movement differential equation of theflywheel are discussed in order to solve the maglev flywheelnonlinear dynamic model Figure 1 shows the below force ofthe flywheel
In Figure 1 119874-119883119884119885 is the space coordinate 1 and 2 arerespectively the left and right position of the radial magnetic
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2015 Article ID 327347 9 pageshttpdxdoiorg1011552015327347
2 Mathematical Problems in Engineering
ΔFx2
ΔFy2
l2
l
fzfy fx
Og
ΔFz
l1
ΔFx1
ΔFy1
X
Y
Z
O120579 120593
1
2
Figure 1 Flywheel below force analysis
bearing 119874119892 (119909119892 119910119892 119911119892) is the mass center of the flywheel 119897
is the distance between 1 and 2 1198971 is the distance from themass center to 1 1198972 is the distance from themass center to 2 120593and 120579 are respectively the angular displacements around the119909-axis and 119910-axis Δ1198651199091 Δ1198651199092 Δ1198651199101 Δ1198651199102 and Δ119865119911 are theelectromagnetic force of the flywheel in the three axes and119891119909119891119910 and119891119911 are the disturbing force and the unbalance forceof the flywheel in the three axes The flywheel movementdifferential equations are deduced according to the belowforce analysis in Figure 1 and the motion laws of particles [9]
119898119892 = Δ1198651199091 + Δ1198651199092 minus 119891119909
119898 119910119892 = Δ1198651199101 + Δ1198651199102 minus 119891119910
119868119903 minus 120596119868119886120579 = Δ1198651199101 sdot 1198971 minus Δ1198651199102 sdot 1198972
119868119903120579 + 120596119868119886 = minusΔ1198651199091 sdot 1198971 + Δ1198651199092 sdot 1198972
119898119892 = Δ119865119911 minus 119891119911
(1)
where 119898 is the flywheel quality 119868119903 and 119868119886 are respectivelythe radial inertia moment and axial inertia moment and 120596 isthe angular velocity of the flywheel The motion differentialequation of matrix form can be written by [10]
(((((((
(
1198972
119897119898
1198971
119897119898 0 0 0
0 01198972
119897119898
1198971
119897119898 0
0 0 minus119868119903
119897
119868119903
1198970
minus119868119903
119897
119868119903
1198970 0 0
0 0 0 0 119898
)))))))
)
(
1
2
1199101
1199102
)
+
(((((((
(
0 0 0 0 0
0 0 0 0 0
120596119868119886
119897minus
120596119868119886
1198970 0 0
0 0 minus120596119868119886
119897
120596119868119886
1198970
0 0 0 0 0
)))))))
)
(
1
2
1199101
1199102
)
=
(((((
(
1198961199091199091 1198961199091199092 0 0 0
0 0 1198961199091199101 1198961199091199102 0
0 0 11989711198961199091199101 minus11989721198961199091199102 0
minus11989711198961199091199091 11989721198961199091199092 0 0 0
0 0 0 0 119896119909119911
)))))
)
(
1199091
1199092
1199101
1199102
119911
)
+
(((((
(
1198961198941199091 1198961198941199092 0 0 0
0 0 1198961198941199101 1198961198941199102 0
0 0 11989711198961198941199101 minus11989721198961198941199102 0
minus11989711198961198941199091 11989721198961198941199092 0 0 0
0 0 0 0 119896119894119911
)))))
)
(
1198941198881199091
1198941198881199092
1198941198881199101
1198941198881199102
119894119888119911
)
+ (
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 1
) (
(
119891119909
119891119910
0
0
119891119911
)
)
(2)
where 1199091 1199092 1199101 1199102 and 119911 are the displacements of theflywheel in five freedomdegrees 1198941198881199091 1198941198881199092 1198941198881199101 1198941198881199101 and 119894119888119911 arerespectively the control current corresponding to each free-dom degree 1198961199091199091 1198961199091199092 1198961199091199101 1198961199091199102 and 119896119909119911 are respectivelythe displacement stiffness of each freedom degree and 11989611989411990911198961198941199092 1198961198941199101 1198961198941199102 and 119896119894119911 are respectively the current stiffness ofeach freedom degree
For the convenience of analysis type (2) can be expressedby type (3)
Mx + Cx + Kx = Bic + If (3)
where M is the mass matrix C is the damping coefficientmatrixK is the displacement stiffnessmatrixB is the currentstiffness matrix I is the unit matrix x is the displacementvector ic is the control current vector and f is the unbalancedinertial force vector Type (3) can be rewritten as
x = minusCM
x minusKM
x +BM
ic +IM
f (4)
Mathematical Problems in Engineering 3
controller
controllerAILC
AMB
Flywheelrotor
Power amplifier
Displacementsensor
Dual-mode switch control strategy
Real-time switching control according to the flywheel speed
DA
AD
Tachometric survey
Reference signal
Hinfin
+minus
Real
-tim
e sw
itchi
ngco
ntro
ller
Reco
gniti
on ju
dgm
ent
Gen
eral
ized
cont
rol e
leph
ant
Figure 2 The dual-mode switch control strategy diagram
The state space of the flywheel system transfer functionmatrix 119866(119904) is
x = Ax + B1f + B2u
y = Cx
(5)
where x = [x x]T is the state vector u = ic is the vector
control y = x is the measured quantity namely the sensoroutput signal vector of the five freedom degrees Moreovertype (5) can also be expressed as
G (119904) = [11986611 (119904) 11986612 (119904)
11986621 (119904) 11986622 (119904)] = [
A B1 B2C 119874 119874
] (6)
where 119874 is the five-order zero matrix it can be solved outthrough types (4) and (5)
A = [
[
119874 IminusKM
minusCM
]
]
B1 = [
[
119874
IM
]
]
B2 = [
[
119874
minusIM
]
]
C = [I 119874]
(7)
So far the nonlinear dynamic model 119866(119904) of the maglevflywheel is solved and the precise model is provided for thedesign of the following dual-mode switch controller
3 Dual-Mode Switch Control Strategy
To improve the control stability and the energy storagedensity of vehicle maglev flywheel a dual-mode switchcontrol strategy is study based on 119867infin control algorithmand adaptive iterative learning control algorithm The con-trol strategy is shown in Figure 2 The control strategy
includes generalized control elephant and dual-mode switchcontroller Among them the generalized control elephantis composed of power amplifiers active magnetic bearing(AMB) flywheel rotor and displacement and the dual-mode switch controller is composed of 119867infin controller andunbalance compensation controller
The basic response is a random signal when EV is drivingon the bumpy road or start-stop acceleration-decelerationand steering To reduce the random impact and improve thecontrol robustness a 119867infin controller is solved Besides tolimit the flywheel radial run-out of the maglev flywheel bat-tery in the charging process and improve the energy storagedensity of the maglev flywheel an unbalance displacementfeed-forward compensation controller based on adaptiveiterative learning control (AILC) algorithm is adopted [10]The switching work between 119867infin control and AILC isimplemented by the real-time switching controller by judgingthe status of the EV and the maglev flywheel
31 119867infin Controller Design A kind of 119867infin control strategywithmixed sensitivity is designed according to the character-istic of the maglev flywheel state space equation it is shownin Figure 3
In Figure 3 KH is the transfer function matrix of the119867infin controller u is the input signal matrix 1198821 1198822 and 1198823
are the weighted functions G is the state space equation ofthe maglev flywheel (it contains the power amplifier transferfunction 119866119860 the sensor transfer function 119866119878 and the controlcurrent ic and so forth) y is the sensor output signal vector ris the system reference signal vector and e is the error signalmatrix
Tc is the closed transfer function in Figure 3 can bewritten as [11]
Tc = 119865119897 (GK119867) = G11 + G12K119867 (I minus G22K119867)minus1G21
(8)
4 Mathematical Problems in Engineering
Generalized controlled object G
Displacement sensor
Flywheel
W1 W2 W3
KH
KH
GAGS
ic
ic
I0
I0
+
+
+
+
minus
minus
uy
r e
Figure 3 The diagram of 119867infin control strategy
where type (8) is the linear fractional transformation of the119867infin controller K119867 The standard of 119867infin control problem isto find a real rational K119867 to make the controlled object Gstable work and to make the minimal 119867infin norm of Tc in thewhole frequency range [12] should be satisfied by
1003817100381710038171003817Tc1003817100381710038171003817infin
= sup120596
120590 (Tc (119895120596)) lt 120574 (9)
where 120590(Tc(119895120596)) is the biggest singular value of Tc ldquosuprdquo isthe supremum of 120590(Tc(119895120596)) in the whole frequency rangeand 120574 is a given positive number
The corresponding sensitivity and complementary sensi-tivity matrix functions S and T combined with Figure 2 canbe deduced
S =er
=1
I + K119867 (119904)G (119904)
T =yr
=K119867 (119904)G (119904)
I + K119867 (119904)G (119904)= I minus S
(10)
1198821(119904) is the weighted function of S its main purpose is tolimit the amplitude of S in a specified frequency range 1198823(119904)is the weighted function of T whose purpose is to limit theamplitude of T In addition the transfer function matrix ofthe 119867infin controller output is
R =ur
=K119867 (119904)
I + K119867 (119904)G (119904) (11)
R also has a weighted function 1198822(119904) its main purpose is tolimit the K119867 controller output 1198822(119904) should be chosen as arelatively small value in order to reduce the compensators andshorten operation cycle
The structural parameters of the maglev flywheel arecalculated and got as 119898 = 108 kg 1198961198941199091 = 1198961198941199092 = 1198961198941199101 = 1198961198941199102 =
9546NA 119896119894119911 = 15438NA 1198961199091199091 = 1198961199091199092 = 1198961199091199101 = 1198961199091199102 =
1517 times 106Nm and 119870119883119885 = 3216 times 10
6Nm Based on 119867infin
controller problem solving limit andMATLAB robust controlinstructions the solved parameters are involved in the G(119904)
solution procedure and the weighting functions 1198821(119904) and1198823(119904) are determined respectively by
1198821 (119904) =98
119904 + 2
1198823 (119904) =15119904
119904 + 3200
(12)
In addition 1198822(119904) is selected as 20 times 107 through the
simulation analysis Then the robust performance index(04232) is calculated based on the controlled object transferfunction matrix G and the three weighted functions if theindex greater than 1 the three weighted functions need to beselected Finally the four radial and one axial discrete transferfunctions of the 119867infin controller are concluded
1198701198671199091 = 1198701198671199092 = 1198701198671199101 = 1198701198671199102
=minus218119911
3+ 153119911
2minus 256119911 + 2132
1199113 minus 10611199112 + 869119911 minus 173
119870119867119911 =minus1013119911
3+ 805119911
2+ 468
1199113 minus 5871199112 + 232119911 minus 021
(13)
Because the maglev flywheel is axially symmetric distri-bution in four radial freedom degrees the four radial discretetransfer functions are the same In order to make the controlcycle consistent with the actual flywheel control system thesimulation sampling frequency is selected as 20000Hz
32 Feed-Forward Compensation Control Analysis To reducethe radial run-out of the flywheel and solve the problemof variable speed the AILC algorithm is adopted as thefeed-forward controller to implement vibratory displacementcompensation [10] In Figure 4 the control scheme consists ofPID feedback control system AILC feed-forward compensa-tion controller and generalized plantThe PID controller cansteady the whole system and improve the anti-interferenceability and the action of AILC is to make the learning gainaccurately track the expectant orbit The generalized plantincludes power amplifier electromagnetic coils and rotorsystem
To improve control performance and enhance the con-vergence rate of the learning law there are two modificationsin AILC The first one is enhancing the error informationaction of previous control period and the second one isproposing a novel impacting factor 120573 as the coefficient of thelearning gain V119896 120573 can reduce the effects of learning gain tothe control system when rotor speed does not coincide withthe learning cycle of AILC AILC can implement vibratorydisplacement compensation without any information of thegeneralized plant and it will not increase the interference ofthe feedback controller Only the expectant signal 119910119889 and theoutput of the sensor 119910119896 are needed in AILC here 119910119889 is 25 Vwitch is defined as the balance position of rotor during staticsuspension The error signal between 119910119889 and 119910119896 is iterativelylearned then the perfect and unknown control signal 119906119896 isobtained as the input signal of the power amplifier
Mathematical Problems in Engineering 5
Sensor
120573
yd
yk
= 25V+
++
+ +
+
+
minus
Q1
ek ekminus1
c uk
P
k+1
k
QMemory II Memory I
PID controller Power amplifier AMB
Flywheel
Feedback control system
AILC feed-forward controller
Generalized plant
Figure 4 AILC compensation principle of AMB system
The functions of AILC can be introduced by the iterativeformulas in discrete domain The error formula is given
The update learning law of AILC is summarized as
V119896+1 (119899) = 120573119875V119896 (119899) + 119876119890119896 (119899) + 1198761119890119896 (119899 minus 1) (14)
To understand the action ofAILCbetter the control inputsignal of generalized plant is written as
119906119896 (119899) = 119888 (119899) + 120573V119896 (119899) (15)
where 119906119896(119899) is the controller input 119888(119899) is PID controlleroutput V119896(119899) is the learning gain of AILC and 120573 is theimpacting factor of V119896(119899) The last objective is to obtainperfect controller signal 119906119889 when having infinitely iterativelearning and then make the error signal become zero Itshould be satisfied as [13]
lim119896rarrinfin
119906119896 (119899) = 119906119889 lim119896rarrinfin
119890119896 (119899) = 0 (16)
The discrete transfer function of the learning law of (14)can be calculated by
119881119896+1 (119911) = 120573119875119881119896 (119911) + 119876119864119896 (119911) + 1198761119864119896 (119911) 119911minus1
= 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) 119864119896 (119911)
(17)
where 119911minus1 represents the lag operator in time domain and
it could make the sampled signal lag one period This is thereason why the former error information is added into themodified learning law of AILC
The PID controller with incomplete differential part isgiven in time domain
119888 (119905) = 119866119875 (119905)
= 119870119901 [119890 (119905) +1
119879119894
int
119905
0
119890 (119905) 119889119905 +119879119889
1 + 119879119891
119889 (119890 (119905))
119889119905]
(18)
The discrete function of (18) is deduced as
119862 (119911)
= 119870119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
(19)
The 119911-transform of the control signal function of (15) isdeduced by
119880119896 (119911) = 119862 (119911) + 120573119881119896 (119911)
= 119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
+ 120573119881119896 (119911)
(20)
where defining
119867 (119911)
= 1 (119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)])
minus1
(21)
The discrete function of error signal can be obtained as
119864 (119911) = (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911) (22)
Putting (22) into (17) it has
119881119896+1 (119911) = 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911)
= 120573 [119875 minus (119876 + 1198761119911minus1
) 119867 (119911)] 119881119896 (119911)
+ (119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
(23)
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Mathematical Problems in Engineering
ΔFx2
ΔFy2
l2
l
fzfy fx
Og
ΔFz
l1
ΔFx1
ΔFy1
X
Y
Z
O120579 120593
1
2
Figure 1 Flywheel below force analysis
bearing 119874119892 (119909119892 119910119892 119911119892) is the mass center of the flywheel 119897
is the distance between 1 and 2 1198971 is the distance from themass center to 1 1198972 is the distance from themass center to 2 120593and 120579 are respectively the angular displacements around the119909-axis and 119910-axis Δ1198651199091 Δ1198651199092 Δ1198651199101 Δ1198651199102 and Δ119865119911 are theelectromagnetic force of the flywheel in the three axes and119891119909119891119910 and119891119911 are the disturbing force and the unbalance forceof the flywheel in the three axes The flywheel movementdifferential equations are deduced according to the belowforce analysis in Figure 1 and the motion laws of particles [9]
119898119892 = Δ1198651199091 + Δ1198651199092 minus 119891119909
119898 119910119892 = Δ1198651199101 + Δ1198651199102 minus 119891119910
119868119903 minus 120596119868119886120579 = Δ1198651199101 sdot 1198971 minus Δ1198651199102 sdot 1198972
119868119903120579 + 120596119868119886 = minusΔ1198651199091 sdot 1198971 + Δ1198651199092 sdot 1198972
119898119892 = Δ119865119911 minus 119891119911
(1)
where 119898 is the flywheel quality 119868119903 and 119868119886 are respectivelythe radial inertia moment and axial inertia moment and 120596 isthe angular velocity of the flywheel The motion differentialequation of matrix form can be written by [10]
(((((((
(
1198972
119897119898
1198971
119897119898 0 0 0
0 01198972
119897119898
1198971
119897119898 0
0 0 minus119868119903
119897
119868119903
1198970
minus119868119903
119897
119868119903
1198970 0 0
0 0 0 0 119898
)))))))
)
(
1
2
1199101
1199102
)
+
(((((((
(
0 0 0 0 0
0 0 0 0 0
120596119868119886
119897minus
120596119868119886
1198970 0 0
0 0 minus120596119868119886
119897
120596119868119886
1198970
0 0 0 0 0
)))))))
)
(
1
2
1199101
1199102
)
=
(((((
(
1198961199091199091 1198961199091199092 0 0 0
0 0 1198961199091199101 1198961199091199102 0
0 0 11989711198961199091199101 minus11989721198961199091199102 0
minus11989711198961199091199091 11989721198961199091199092 0 0 0
0 0 0 0 119896119909119911
)))))
)
(
1199091
1199092
1199101
1199102
119911
)
+
(((((
(
1198961198941199091 1198961198941199092 0 0 0
0 0 1198961198941199101 1198961198941199102 0
0 0 11989711198961198941199101 minus11989721198961198941199102 0
minus11989711198961198941199091 11989721198961198941199092 0 0 0
0 0 0 0 119896119894119911
)))))
)
(
1198941198881199091
1198941198881199092
1198941198881199101
1198941198881199102
119894119888119911
)
+ (
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 1
) (
(
119891119909
119891119910
0
0
119891119911
)
)
(2)
where 1199091 1199092 1199101 1199102 and 119911 are the displacements of theflywheel in five freedomdegrees 1198941198881199091 1198941198881199092 1198941198881199101 1198941198881199101 and 119894119888119911 arerespectively the control current corresponding to each free-dom degree 1198961199091199091 1198961199091199092 1198961199091199101 1198961199091199102 and 119896119909119911 are respectivelythe displacement stiffness of each freedom degree and 11989611989411990911198961198941199092 1198961198941199101 1198961198941199102 and 119896119894119911 are respectively the current stiffness ofeach freedom degree
For the convenience of analysis type (2) can be expressedby type (3)
Mx + Cx + Kx = Bic + If (3)
where M is the mass matrix C is the damping coefficientmatrixK is the displacement stiffnessmatrixB is the currentstiffness matrix I is the unit matrix x is the displacementvector ic is the control current vector and f is the unbalancedinertial force vector Type (3) can be rewritten as
x = minusCM
x minusKM
x +BM
ic +IM
f (4)
Mathematical Problems in Engineering 3
controller
controllerAILC
AMB
Flywheelrotor
Power amplifier
Displacementsensor
Dual-mode switch control strategy
Real-time switching control according to the flywheel speed
DA
AD
Tachometric survey
Reference signal
Hinfin
+minus
Real
-tim
e sw
itchi
ngco
ntro
ller
Reco
gniti
on ju
dgm
ent
Gen
eral
ized
cont
rol e
leph
ant
Figure 2 The dual-mode switch control strategy diagram
The state space of the flywheel system transfer functionmatrix 119866(119904) is
x = Ax + B1f + B2u
y = Cx
(5)
where x = [x x]T is the state vector u = ic is the vector
control y = x is the measured quantity namely the sensoroutput signal vector of the five freedom degrees Moreovertype (5) can also be expressed as
G (119904) = [11986611 (119904) 11986612 (119904)
11986621 (119904) 11986622 (119904)] = [
A B1 B2C 119874 119874
] (6)
where 119874 is the five-order zero matrix it can be solved outthrough types (4) and (5)
A = [
[
119874 IminusKM
minusCM
]
]
B1 = [
[
119874
IM
]
]
B2 = [
[
119874
minusIM
]
]
C = [I 119874]
(7)
So far the nonlinear dynamic model 119866(119904) of the maglevflywheel is solved and the precise model is provided for thedesign of the following dual-mode switch controller
3 Dual-Mode Switch Control Strategy
To improve the control stability and the energy storagedensity of vehicle maglev flywheel a dual-mode switchcontrol strategy is study based on 119867infin control algorithmand adaptive iterative learning control algorithm The con-trol strategy is shown in Figure 2 The control strategy
includes generalized control elephant and dual-mode switchcontroller Among them the generalized control elephantis composed of power amplifiers active magnetic bearing(AMB) flywheel rotor and displacement and the dual-mode switch controller is composed of 119867infin controller andunbalance compensation controller
The basic response is a random signal when EV is drivingon the bumpy road or start-stop acceleration-decelerationand steering To reduce the random impact and improve thecontrol robustness a 119867infin controller is solved Besides tolimit the flywheel radial run-out of the maglev flywheel bat-tery in the charging process and improve the energy storagedensity of the maglev flywheel an unbalance displacementfeed-forward compensation controller based on adaptiveiterative learning control (AILC) algorithm is adopted [10]The switching work between 119867infin control and AILC isimplemented by the real-time switching controller by judgingthe status of the EV and the maglev flywheel
31 119867infin Controller Design A kind of 119867infin control strategywithmixed sensitivity is designed according to the character-istic of the maglev flywheel state space equation it is shownin Figure 3
In Figure 3 KH is the transfer function matrix of the119867infin controller u is the input signal matrix 1198821 1198822 and 1198823
are the weighted functions G is the state space equation ofthe maglev flywheel (it contains the power amplifier transferfunction 119866119860 the sensor transfer function 119866119878 and the controlcurrent ic and so forth) y is the sensor output signal vector ris the system reference signal vector and e is the error signalmatrix
Tc is the closed transfer function in Figure 3 can bewritten as [11]
Tc = 119865119897 (GK119867) = G11 + G12K119867 (I minus G22K119867)minus1G21
(8)
4 Mathematical Problems in Engineering
Generalized controlled object G
Displacement sensor
Flywheel
W1 W2 W3
KH
KH
GAGS
ic
ic
I0
I0
+
+
+
+
minus
minus
uy
r e
Figure 3 The diagram of 119867infin control strategy
where type (8) is the linear fractional transformation of the119867infin controller K119867 The standard of 119867infin control problem isto find a real rational K119867 to make the controlled object Gstable work and to make the minimal 119867infin norm of Tc in thewhole frequency range [12] should be satisfied by
1003817100381710038171003817Tc1003817100381710038171003817infin
= sup120596
120590 (Tc (119895120596)) lt 120574 (9)
where 120590(Tc(119895120596)) is the biggest singular value of Tc ldquosuprdquo isthe supremum of 120590(Tc(119895120596)) in the whole frequency rangeand 120574 is a given positive number
The corresponding sensitivity and complementary sensi-tivity matrix functions S and T combined with Figure 2 canbe deduced
S =er
=1
I + K119867 (119904)G (119904)
T =yr
=K119867 (119904)G (119904)
I + K119867 (119904)G (119904)= I minus S
(10)
1198821(119904) is the weighted function of S its main purpose is tolimit the amplitude of S in a specified frequency range 1198823(119904)is the weighted function of T whose purpose is to limit theamplitude of T In addition the transfer function matrix ofthe 119867infin controller output is
R =ur
=K119867 (119904)
I + K119867 (119904)G (119904) (11)
R also has a weighted function 1198822(119904) its main purpose is tolimit the K119867 controller output 1198822(119904) should be chosen as arelatively small value in order to reduce the compensators andshorten operation cycle
The structural parameters of the maglev flywheel arecalculated and got as 119898 = 108 kg 1198961198941199091 = 1198961198941199092 = 1198961198941199101 = 1198961198941199102 =
9546NA 119896119894119911 = 15438NA 1198961199091199091 = 1198961199091199092 = 1198961199091199101 = 1198961199091199102 =
1517 times 106Nm and 119870119883119885 = 3216 times 10
6Nm Based on 119867infin
controller problem solving limit andMATLAB robust controlinstructions the solved parameters are involved in the G(119904)
solution procedure and the weighting functions 1198821(119904) and1198823(119904) are determined respectively by
1198821 (119904) =98
119904 + 2
1198823 (119904) =15119904
119904 + 3200
(12)
In addition 1198822(119904) is selected as 20 times 107 through the
simulation analysis Then the robust performance index(04232) is calculated based on the controlled object transferfunction matrix G and the three weighted functions if theindex greater than 1 the three weighted functions need to beselected Finally the four radial and one axial discrete transferfunctions of the 119867infin controller are concluded
1198701198671199091 = 1198701198671199092 = 1198701198671199101 = 1198701198671199102
=minus218119911
3+ 153119911
2minus 256119911 + 2132
1199113 minus 10611199112 + 869119911 minus 173
119870119867119911 =minus1013119911
3+ 805119911
2+ 468
1199113 minus 5871199112 + 232119911 minus 021
(13)
Because the maglev flywheel is axially symmetric distri-bution in four radial freedom degrees the four radial discretetransfer functions are the same In order to make the controlcycle consistent with the actual flywheel control system thesimulation sampling frequency is selected as 20000Hz
32 Feed-Forward Compensation Control Analysis To reducethe radial run-out of the flywheel and solve the problemof variable speed the AILC algorithm is adopted as thefeed-forward controller to implement vibratory displacementcompensation [10] In Figure 4 the control scheme consists ofPID feedback control system AILC feed-forward compensa-tion controller and generalized plantThe PID controller cansteady the whole system and improve the anti-interferenceability and the action of AILC is to make the learning gainaccurately track the expectant orbit The generalized plantincludes power amplifier electromagnetic coils and rotorsystem
To improve control performance and enhance the con-vergence rate of the learning law there are two modificationsin AILC The first one is enhancing the error informationaction of previous control period and the second one isproposing a novel impacting factor 120573 as the coefficient of thelearning gain V119896 120573 can reduce the effects of learning gain tothe control system when rotor speed does not coincide withthe learning cycle of AILC AILC can implement vibratorydisplacement compensation without any information of thegeneralized plant and it will not increase the interference ofthe feedback controller Only the expectant signal 119910119889 and theoutput of the sensor 119910119896 are needed in AILC here 119910119889 is 25 Vwitch is defined as the balance position of rotor during staticsuspension The error signal between 119910119889 and 119910119896 is iterativelylearned then the perfect and unknown control signal 119906119896 isobtained as the input signal of the power amplifier
Mathematical Problems in Engineering 5
Sensor
120573
yd
yk
= 25V+
++
+ +
+
+
minus
Q1
ek ekminus1
c uk
P
k+1
k
QMemory II Memory I
PID controller Power amplifier AMB
Flywheel
Feedback control system
AILC feed-forward controller
Generalized plant
Figure 4 AILC compensation principle of AMB system
The functions of AILC can be introduced by the iterativeformulas in discrete domain The error formula is given
The update learning law of AILC is summarized as
V119896+1 (119899) = 120573119875V119896 (119899) + 119876119890119896 (119899) + 1198761119890119896 (119899 minus 1) (14)
To understand the action ofAILCbetter the control inputsignal of generalized plant is written as
119906119896 (119899) = 119888 (119899) + 120573V119896 (119899) (15)
where 119906119896(119899) is the controller input 119888(119899) is PID controlleroutput V119896(119899) is the learning gain of AILC and 120573 is theimpacting factor of V119896(119899) The last objective is to obtainperfect controller signal 119906119889 when having infinitely iterativelearning and then make the error signal become zero Itshould be satisfied as [13]
lim119896rarrinfin
119906119896 (119899) = 119906119889 lim119896rarrinfin
119890119896 (119899) = 0 (16)
The discrete transfer function of the learning law of (14)can be calculated by
119881119896+1 (119911) = 120573119875119881119896 (119911) + 119876119864119896 (119911) + 1198761119864119896 (119911) 119911minus1
= 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) 119864119896 (119911)
(17)
where 119911minus1 represents the lag operator in time domain and
it could make the sampled signal lag one period This is thereason why the former error information is added into themodified learning law of AILC
The PID controller with incomplete differential part isgiven in time domain
119888 (119905) = 119866119875 (119905)
= 119870119901 [119890 (119905) +1
119879119894
int
119905
0
119890 (119905) 119889119905 +119879119889
1 + 119879119891
119889 (119890 (119905))
119889119905]
(18)
The discrete function of (18) is deduced as
119862 (119911)
= 119870119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
(19)
The 119911-transform of the control signal function of (15) isdeduced by
119880119896 (119911) = 119862 (119911) + 120573119881119896 (119911)
= 119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
+ 120573119881119896 (119911)
(20)
where defining
119867 (119911)
= 1 (119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)])
minus1
(21)
The discrete function of error signal can be obtained as
119864 (119911) = (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911) (22)
Putting (22) into (17) it has
119881119896+1 (119911) = 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911)
= 120573 [119875 minus (119876 + 1198761119911minus1
) 119867 (119911)] 119881119896 (119911)
+ (119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
(23)
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 3
controller
controllerAILC
AMB
Flywheelrotor
Power amplifier
Displacementsensor
Dual-mode switch control strategy
Real-time switching control according to the flywheel speed
DA
AD
Tachometric survey
Reference signal
Hinfin
+minus
Real
-tim
e sw
itchi
ngco
ntro
ller
Reco
gniti
on ju
dgm
ent
Gen
eral
ized
cont
rol e
leph
ant
Figure 2 The dual-mode switch control strategy diagram
The state space of the flywheel system transfer functionmatrix 119866(119904) is
x = Ax + B1f + B2u
y = Cx
(5)
where x = [x x]T is the state vector u = ic is the vector
control y = x is the measured quantity namely the sensoroutput signal vector of the five freedom degrees Moreovertype (5) can also be expressed as
G (119904) = [11986611 (119904) 11986612 (119904)
11986621 (119904) 11986622 (119904)] = [
A B1 B2C 119874 119874
] (6)
where 119874 is the five-order zero matrix it can be solved outthrough types (4) and (5)
A = [
[
119874 IminusKM
minusCM
]
]
B1 = [
[
119874
IM
]
]
B2 = [
[
119874
minusIM
]
]
C = [I 119874]
(7)
So far the nonlinear dynamic model 119866(119904) of the maglevflywheel is solved and the precise model is provided for thedesign of the following dual-mode switch controller
3 Dual-Mode Switch Control Strategy
To improve the control stability and the energy storagedensity of vehicle maglev flywheel a dual-mode switchcontrol strategy is study based on 119867infin control algorithmand adaptive iterative learning control algorithm The con-trol strategy is shown in Figure 2 The control strategy
includes generalized control elephant and dual-mode switchcontroller Among them the generalized control elephantis composed of power amplifiers active magnetic bearing(AMB) flywheel rotor and displacement and the dual-mode switch controller is composed of 119867infin controller andunbalance compensation controller
The basic response is a random signal when EV is drivingon the bumpy road or start-stop acceleration-decelerationand steering To reduce the random impact and improve thecontrol robustness a 119867infin controller is solved Besides tolimit the flywheel radial run-out of the maglev flywheel bat-tery in the charging process and improve the energy storagedensity of the maglev flywheel an unbalance displacementfeed-forward compensation controller based on adaptiveiterative learning control (AILC) algorithm is adopted [10]The switching work between 119867infin control and AILC isimplemented by the real-time switching controller by judgingthe status of the EV and the maglev flywheel
31 119867infin Controller Design A kind of 119867infin control strategywithmixed sensitivity is designed according to the character-istic of the maglev flywheel state space equation it is shownin Figure 3
In Figure 3 KH is the transfer function matrix of the119867infin controller u is the input signal matrix 1198821 1198822 and 1198823
are the weighted functions G is the state space equation ofthe maglev flywheel (it contains the power amplifier transferfunction 119866119860 the sensor transfer function 119866119878 and the controlcurrent ic and so forth) y is the sensor output signal vector ris the system reference signal vector and e is the error signalmatrix
Tc is the closed transfer function in Figure 3 can bewritten as [11]
Tc = 119865119897 (GK119867) = G11 + G12K119867 (I minus G22K119867)minus1G21
(8)
4 Mathematical Problems in Engineering
Generalized controlled object G
Displacement sensor
Flywheel
W1 W2 W3
KH
KH
GAGS
ic
ic
I0
I0
+
+
+
+
minus
minus
uy
r e
Figure 3 The diagram of 119867infin control strategy
where type (8) is the linear fractional transformation of the119867infin controller K119867 The standard of 119867infin control problem isto find a real rational K119867 to make the controlled object Gstable work and to make the minimal 119867infin norm of Tc in thewhole frequency range [12] should be satisfied by
1003817100381710038171003817Tc1003817100381710038171003817infin
= sup120596
120590 (Tc (119895120596)) lt 120574 (9)
where 120590(Tc(119895120596)) is the biggest singular value of Tc ldquosuprdquo isthe supremum of 120590(Tc(119895120596)) in the whole frequency rangeand 120574 is a given positive number
The corresponding sensitivity and complementary sensi-tivity matrix functions S and T combined with Figure 2 canbe deduced
S =er
=1
I + K119867 (119904)G (119904)
T =yr
=K119867 (119904)G (119904)
I + K119867 (119904)G (119904)= I minus S
(10)
1198821(119904) is the weighted function of S its main purpose is tolimit the amplitude of S in a specified frequency range 1198823(119904)is the weighted function of T whose purpose is to limit theamplitude of T In addition the transfer function matrix ofthe 119867infin controller output is
R =ur
=K119867 (119904)
I + K119867 (119904)G (119904) (11)
R also has a weighted function 1198822(119904) its main purpose is tolimit the K119867 controller output 1198822(119904) should be chosen as arelatively small value in order to reduce the compensators andshorten operation cycle
The structural parameters of the maglev flywheel arecalculated and got as 119898 = 108 kg 1198961198941199091 = 1198961198941199092 = 1198961198941199101 = 1198961198941199102 =
9546NA 119896119894119911 = 15438NA 1198961199091199091 = 1198961199091199092 = 1198961199091199101 = 1198961199091199102 =
1517 times 106Nm and 119870119883119885 = 3216 times 10
6Nm Based on 119867infin
controller problem solving limit andMATLAB robust controlinstructions the solved parameters are involved in the G(119904)
solution procedure and the weighting functions 1198821(119904) and1198823(119904) are determined respectively by
1198821 (119904) =98
119904 + 2
1198823 (119904) =15119904
119904 + 3200
(12)
In addition 1198822(119904) is selected as 20 times 107 through the
simulation analysis Then the robust performance index(04232) is calculated based on the controlled object transferfunction matrix G and the three weighted functions if theindex greater than 1 the three weighted functions need to beselected Finally the four radial and one axial discrete transferfunctions of the 119867infin controller are concluded
1198701198671199091 = 1198701198671199092 = 1198701198671199101 = 1198701198671199102
=minus218119911
3+ 153119911
2minus 256119911 + 2132
1199113 minus 10611199112 + 869119911 minus 173
119870119867119911 =minus1013119911
3+ 805119911
2+ 468
1199113 minus 5871199112 + 232119911 minus 021
(13)
Because the maglev flywheel is axially symmetric distri-bution in four radial freedom degrees the four radial discretetransfer functions are the same In order to make the controlcycle consistent with the actual flywheel control system thesimulation sampling frequency is selected as 20000Hz
32 Feed-Forward Compensation Control Analysis To reducethe radial run-out of the flywheel and solve the problemof variable speed the AILC algorithm is adopted as thefeed-forward controller to implement vibratory displacementcompensation [10] In Figure 4 the control scheme consists ofPID feedback control system AILC feed-forward compensa-tion controller and generalized plantThe PID controller cansteady the whole system and improve the anti-interferenceability and the action of AILC is to make the learning gainaccurately track the expectant orbit The generalized plantincludes power amplifier electromagnetic coils and rotorsystem
To improve control performance and enhance the con-vergence rate of the learning law there are two modificationsin AILC The first one is enhancing the error informationaction of previous control period and the second one isproposing a novel impacting factor 120573 as the coefficient of thelearning gain V119896 120573 can reduce the effects of learning gain tothe control system when rotor speed does not coincide withthe learning cycle of AILC AILC can implement vibratorydisplacement compensation without any information of thegeneralized plant and it will not increase the interference ofthe feedback controller Only the expectant signal 119910119889 and theoutput of the sensor 119910119896 are needed in AILC here 119910119889 is 25 Vwitch is defined as the balance position of rotor during staticsuspension The error signal between 119910119889 and 119910119896 is iterativelylearned then the perfect and unknown control signal 119906119896 isobtained as the input signal of the power amplifier
Mathematical Problems in Engineering 5
Sensor
120573
yd
yk
= 25V+
++
+ +
+
+
minus
Q1
ek ekminus1
c uk
P
k+1
k
QMemory II Memory I
PID controller Power amplifier AMB
Flywheel
Feedback control system
AILC feed-forward controller
Generalized plant
Figure 4 AILC compensation principle of AMB system
The functions of AILC can be introduced by the iterativeformulas in discrete domain The error formula is given
The update learning law of AILC is summarized as
V119896+1 (119899) = 120573119875V119896 (119899) + 119876119890119896 (119899) + 1198761119890119896 (119899 minus 1) (14)
To understand the action ofAILCbetter the control inputsignal of generalized plant is written as
119906119896 (119899) = 119888 (119899) + 120573V119896 (119899) (15)
where 119906119896(119899) is the controller input 119888(119899) is PID controlleroutput V119896(119899) is the learning gain of AILC and 120573 is theimpacting factor of V119896(119899) The last objective is to obtainperfect controller signal 119906119889 when having infinitely iterativelearning and then make the error signal become zero Itshould be satisfied as [13]
lim119896rarrinfin
119906119896 (119899) = 119906119889 lim119896rarrinfin
119890119896 (119899) = 0 (16)
The discrete transfer function of the learning law of (14)can be calculated by
119881119896+1 (119911) = 120573119875119881119896 (119911) + 119876119864119896 (119911) + 1198761119864119896 (119911) 119911minus1
= 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) 119864119896 (119911)
(17)
where 119911minus1 represents the lag operator in time domain and
it could make the sampled signal lag one period This is thereason why the former error information is added into themodified learning law of AILC
The PID controller with incomplete differential part isgiven in time domain
119888 (119905) = 119866119875 (119905)
= 119870119901 [119890 (119905) +1
119879119894
int
119905
0
119890 (119905) 119889119905 +119879119889
1 + 119879119891
119889 (119890 (119905))
119889119905]
(18)
The discrete function of (18) is deduced as
119862 (119911)
= 119870119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
(19)
The 119911-transform of the control signal function of (15) isdeduced by
119880119896 (119911) = 119862 (119911) + 120573119881119896 (119911)
= 119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
+ 120573119881119896 (119911)
(20)
where defining
119867 (119911)
= 1 (119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)])
minus1
(21)
The discrete function of error signal can be obtained as
119864 (119911) = (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911) (22)
Putting (22) into (17) it has
119881119896+1 (119911) = 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911)
= 120573 [119875 minus (119876 + 1198761119911minus1
) 119867 (119911)] 119881119896 (119911)
+ (119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
(23)
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Mathematical Problems in Engineering
Generalized controlled object G
Displacement sensor
Flywheel
W1 W2 W3
KH
KH
GAGS
ic
ic
I0
I0
+
+
+
+
minus
minus
uy
r e
Figure 3 The diagram of 119867infin control strategy
where type (8) is the linear fractional transformation of the119867infin controller K119867 The standard of 119867infin control problem isto find a real rational K119867 to make the controlled object Gstable work and to make the minimal 119867infin norm of Tc in thewhole frequency range [12] should be satisfied by
1003817100381710038171003817Tc1003817100381710038171003817infin
= sup120596
120590 (Tc (119895120596)) lt 120574 (9)
where 120590(Tc(119895120596)) is the biggest singular value of Tc ldquosuprdquo isthe supremum of 120590(Tc(119895120596)) in the whole frequency rangeand 120574 is a given positive number
The corresponding sensitivity and complementary sensi-tivity matrix functions S and T combined with Figure 2 canbe deduced
S =er
=1
I + K119867 (119904)G (119904)
T =yr
=K119867 (119904)G (119904)
I + K119867 (119904)G (119904)= I minus S
(10)
1198821(119904) is the weighted function of S its main purpose is tolimit the amplitude of S in a specified frequency range 1198823(119904)is the weighted function of T whose purpose is to limit theamplitude of T In addition the transfer function matrix ofthe 119867infin controller output is
R =ur
=K119867 (119904)
I + K119867 (119904)G (119904) (11)
R also has a weighted function 1198822(119904) its main purpose is tolimit the K119867 controller output 1198822(119904) should be chosen as arelatively small value in order to reduce the compensators andshorten operation cycle
The structural parameters of the maglev flywheel arecalculated and got as 119898 = 108 kg 1198961198941199091 = 1198961198941199092 = 1198961198941199101 = 1198961198941199102 =
9546NA 119896119894119911 = 15438NA 1198961199091199091 = 1198961199091199092 = 1198961199091199101 = 1198961199091199102 =
1517 times 106Nm and 119870119883119885 = 3216 times 10
6Nm Based on 119867infin
controller problem solving limit andMATLAB robust controlinstructions the solved parameters are involved in the G(119904)
solution procedure and the weighting functions 1198821(119904) and1198823(119904) are determined respectively by
1198821 (119904) =98
119904 + 2
1198823 (119904) =15119904
119904 + 3200
(12)
In addition 1198822(119904) is selected as 20 times 107 through the
simulation analysis Then the robust performance index(04232) is calculated based on the controlled object transferfunction matrix G and the three weighted functions if theindex greater than 1 the three weighted functions need to beselected Finally the four radial and one axial discrete transferfunctions of the 119867infin controller are concluded
1198701198671199091 = 1198701198671199092 = 1198701198671199101 = 1198701198671199102
=minus218119911
3+ 153119911
2minus 256119911 + 2132
1199113 minus 10611199112 + 869119911 minus 173
119870119867119911 =minus1013119911
3+ 805119911
2+ 468
1199113 minus 5871199112 + 232119911 minus 021
(13)
Because the maglev flywheel is axially symmetric distri-bution in four radial freedom degrees the four radial discretetransfer functions are the same In order to make the controlcycle consistent with the actual flywheel control system thesimulation sampling frequency is selected as 20000Hz
32 Feed-Forward Compensation Control Analysis To reducethe radial run-out of the flywheel and solve the problemof variable speed the AILC algorithm is adopted as thefeed-forward controller to implement vibratory displacementcompensation [10] In Figure 4 the control scheme consists ofPID feedback control system AILC feed-forward compensa-tion controller and generalized plantThe PID controller cansteady the whole system and improve the anti-interferenceability and the action of AILC is to make the learning gainaccurately track the expectant orbit The generalized plantincludes power amplifier electromagnetic coils and rotorsystem
To improve control performance and enhance the con-vergence rate of the learning law there are two modificationsin AILC The first one is enhancing the error informationaction of previous control period and the second one isproposing a novel impacting factor 120573 as the coefficient of thelearning gain V119896 120573 can reduce the effects of learning gain tothe control system when rotor speed does not coincide withthe learning cycle of AILC AILC can implement vibratorydisplacement compensation without any information of thegeneralized plant and it will not increase the interference ofthe feedback controller Only the expectant signal 119910119889 and theoutput of the sensor 119910119896 are needed in AILC here 119910119889 is 25 Vwitch is defined as the balance position of rotor during staticsuspension The error signal between 119910119889 and 119910119896 is iterativelylearned then the perfect and unknown control signal 119906119896 isobtained as the input signal of the power amplifier
Mathematical Problems in Engineering 5
Sensor
120573
yd
yk
= 25V+
++
+ +
+
+
minus
Q1
ek ekminus1
c uk
P
k+1
k
QMemory II Memory I
PID controller Power amplifier AMB
Flywheel
Feedback control system
AILC feed-forward controller
Generalized plant
Figure 4 AILC compensation principle of AMB system
The functions of AILC can be introduced by the iterativeformulas in discrete domain The error formula is given
The update learning law of AILC is summarized as
V119896+1 (119899) = 120573119875V119896 (119899) + 119876119890119896 (119899) + 1198761119890119896 (119899 minus 1) (14)
To understand the action ofAILCbetter the control inputsignal of generalized plant is written as
119906119896 (119899) = 119888 (119899) + 120573V119896 (119899) (15)
where 119906119896(119899) is the controller input 119888(119899) is PID controlleroutput V119896(119899) is the learning gain of AILC and 120573 is theimpacting factor of V119896(119899) The last objective is to obtainperfect controller signal 119906119889 when having infinitely iterativelearning and then make the error signal become zero Itshould be satisfied as [13]
lim119896rarrinfin
119906119896 (119899) = 119906119889 lim119896rarrinfin
119890119896 (119899) = 0 (16)
The discrete transfer function of the learning law of (14)can be calculated by
119881119896+1 (119911) = 120573119875119881119896 (119911) + 119876119864119896 (119911) + 1198761119864119896 (119911) 119911minus1
= 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) 119864119896 (119911)
(17)
where 119911minus1 represents the lag operator in time domain and
it could make the sampled signal lag one period This is thereason why the former error information is added into themodified learning law of AILC
The PID controller with incomplete differential part isgiven in time domain
119888 (119905) = 119866119875 (119905)
= 119870119901 [119890 (119905) +1
119879119894
int
119905
0
119890 (119905) 119889119905 +119879119889
1 + 119879119891
119889 (119890 (119905))
119889119905]
(18)
The discrete function of (18) is deduced as
119862 (119911)
= 119870119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
(19)
The 119911-transform of the control signal function of (15) isdeduced by
119880119896 (119911) = 119862 (119911) + 120573119881119896 (119911)
= 119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
+ 120573119881119896 (119911)
(20)
where defining
119867 (119911)
= 1 (119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)])
minus1
(21)
The discrete function of error signal can be obtained as
119864 (119911) = (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911) (22)
Putting (22) into (17) it has
119881119896+1 (119911) = 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911)
= 120573 [119875 minus (119876 + 1198761119911minus1
) 119867 (119911)] 119881119896 (119911)
+ (119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
(23)
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 5
Sensor
120573
yd
yk
= 25V+
++
+ +
+
+
minus
Q1
ek ekminus1
c uk
P
k+1
k
QMemory II Memory I
PID controller Power amplifier AMB
Flywheel
Feedback control system
AILC feed-forward controller
Generalized plant
Figure 4 AILC compensation principle of AMB system
The functions of AILC can be introduced by the iterativeformulas in discrete domain The error formula is given
The update learning law of AILC is summarized as
V119896+1 (119899) = 120573119875V119896 (119899) + 119876119890119896 (119899) + 1198761119890119896 (119899 minus 1) (14)
To understand the action ofAILCbetter the control inputsignal of generalized plant is written as
119906119896 (119899) = 119888 (119899) + 120573V119896 (119899) (15)
where 119906119896(119899) is the controller input 119888(119899) is PID controlleroutput V119896(119899) is the learning gain of AILC and 120573 is theimpacting factor of V119896(119899) The last objective is to obtainperfect controller signal 119906119889 when having infinitely iterativelearning and then make the error signal become zero Itshould be satisfied as [13]
lim119896rarrinfin
119906119896 (119899) = 119906119889 lim119896rarrinfin
119890119896 (119899) = 0 (16)
The discrete transfer function of the learning law of (14)can be calculated by
119881119896+1 (119911) = 120573119875119881119896 (119911) + 119876119864119896 (119911) + 1198761119864119896 (119911) 119911minus1
= 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) 119864119896 (119911)
(17)
where 119911minus1 represents the lag operator in time domain and
it could make the sampled signal lag one period This is thereason why the former error information is added into themodified learning law of AILC
The PID controller with incomplete differential part isgiven in time domain
119888 (119905) = 119866119875 (119905)
= 119870119901 [119890 (119905) +1
119879119894
int
119905
0
119890 (119905) 119889119905 +119879119889
1 + 119879119891
119889 (119890 (119905))
119889119905]
(18)
The discrete function of (18) is deduced as
119862 (119911)
= 119870119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
(19)
The 119911-transform of the control signal function of (15) isdeduced by
119880119896 (119911) = 119862 (119911) + 120573119881119896 (119911)
= 119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)] 119864 (119911)
+ 120573119881119896 (119911)
(20)
where defining
119867 (119911)
= 1 (119896119901 [1 +1198790
119879119894
1
1 minus 119911minus1+
119879119889
(1 + 119879119891) 1198790
(1 minus 119911minus1
)])
minus1
(21)
The discrete function of error signal can be obtained as
119864 (119911) = (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911) (22)
Putting (22) into (17) it has
119881119896+1 (119911) = 120573119875119881119896 (119911) + (119876 + 1198761119911minus1
) (119880119896 (119911) minus 120573119881119896 (119911)) 119867 (119911)
= 120573 [119875 minus (119876 + 1198761119911minus1
) 119867 (119911)] 119881119896 (119911)
+ (119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
(23)
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Mathematical Problems in Engineering
Equation (24) gives the transformation in limit calcula-tion at the two sides of (23)
119881infin (119911) = lim119896rarrinfin
119881119896 (119911)
= lim119896rarrinfin
(119876 + 1198761119911minus1
) 119867 (119911) 119880119896 (119911)
1 minus 120573 [119875 + (119876 + 1198761119911minus1) 119867 (119911)]
(24)
If 119875 and 120573 all are 1 at the same time the followinginequality is satisfied by [14]
10038171003817100381710038171003817120573119875 minus 120573[119876 + 1198761119911
minus1]119867(119911)
10038171003817100381710038171003817infinlt 1 (25)
Equation (24) can be simplified by
119881infin (119911) = lim119896rarrinfin
119880119896 (119911) = 119880119889 (119911) (26)
That is the controller signal 119906119896(119911) is replaced by 120573V119896(119911)
when infinite iteration is operated and the error signal willbecome zero According to (22) and (26) the error signal canbe shown as
lim119896rarrinfin
119864 (119911) = 119867 (119911) [ lim119896rarrinfin
119880119896 (119911) minus 120573 lim119896rarrinfin
119881119896 (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119881infin (119911)]
= 119867 (119911) [119880119889 (119911) minus 120573119880119889 (119911)] = 0
(27)
The convergence of AILC has been demonstrated accord-ing to (26) and (27) and a perfect controller signal 119906119889 hasbeen obtained as power amplifier input Therefore the AILCalgorithm as the feed-forward compensation controller canbe adopted in the application of maglev flywheel unbalancevibratory compensation
According to the start-up time and the variability of therotor frequency from static suspension to one fixed speed theequation of 120573 is given as
120573 = (119891
119891119889
)
119899
(28)
where 119891 is the flywheel frequency 119891119889 is a given frequencyand the action of 119899 is to reduce the value of 120573 when 119891 is faraway from 119891119889 (usually 119899 is greater than 2) In the start-upprocess due to 119891 ≪ 119891119889 the value of 120573 should be very smallit can weaken the influence of 119881119896 on the generalized plantHowever 120573 will be close to 1 if 119891 asymp 119891119889 it can enhance theeffect of repetitive learning and can make the error signal toconverge toward zero
4 Simulation and Experimental
41 Simulation Analysis First the stability of 119867infin controlleris attested through analyzing the singular values relationshipbetween 119878(119904) and 1198821(119904) as well as between 119879(119904) and 1198823(119904)The sensitivity and complementary sensitivity function andthe corresponding weighted function singular value relations
0
20
40
60
80
100
10minus1 100 101 102 103 104 105
Angular frequency 120596 (rads)
minus20
minus40
minus60
S(s)1W1(s)
T(s)
1W3(s)
Sing
ular
val
ue (d
B)Figure 5 Sensitivity complementary sensitivity and the corre-sponding weighted function singular value relations
are shown in Figure 5 The curves in Figure 5 are created inthe ldquoMrdquo file of MATLAB through some command functionsthe transfer functions of Tc K119867 and G(119904) and so on
The choice principle of 1198821(119904) is guaranteeing the anti-interference and tracking ability of the flywheel system andthe smaller the singular value of 119878(119904) is the better the systemtracking ability is The curve of 119878(119904) should be under thecurve of the 11198821(119904) if it does not conform to the demandthe weighted function 1198821(119904) must be chosen again Thechoice principle of 1198823(119904) is guaranteeing the flywheel systemoutput to recurrence the input and the smaller the singularvalue of 119879(119904) is the smaller the system impact by compounddisturbance because of model uncertainty is The curve of119879(119904) should also be under the curve of the 11198823(119904) if it doesnot conform to the demand the weighted function 1198823(119904)
must be chosen again Figure 5 shows that the curve of 119878(119904)
is under the curve of 11198821(119904) and the curve of 119879(119904) is underthe curve of 11198823(119904) which demonstrates that the selection ofweighted function in types (14) and (15) can meet the designrequirements and the solved 119867infin controller is appropriate
Second the simulation parameters of AILC are selected as119891119889 = 600Hz 119875 = 0995 119876 = 075 1198761 = 025 119899 = 0 1 99and the definition of ldquo119899rdquo indicates that the AILC algorithmhas 100 memory points in one control period Figure 6 showsthe rotor simulation orbits and radial run-out displacementswith AILC compensation
The flywheel has a regular circular trajectory in Fig-ure 6(a) and a sine radial run-out displacement in Fig-ure 6(b) without adding AILC compensation When theAILC algorithm starts to work at 005 s the circular trajectoryis gradual convergence to a point and the amplitude ofthe radial run-out displacement is obvious attenuation Thecurves change simulation result in Figure 6 testifies the AILCalgorithm having good displacement compensation effect
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 7
0 10 20
0
5
10
15D
istan
cey1
(120583m
)
Distance x1 (120583m)
minus5
minus10
minus15minus20 minus10
(a) Rotor position orbit
0 002 004 006 008 01
0
5
10
15
Dist
ance
x1
(120583m
)
minus5
minus10
minus15
Time t (s)
(b) Rotor run-out in axis 1199091
Figure 6 Flywheel trajectory with AILC compensation
00
4
3
2
5
100 600500400300200 700
4
2
6
Flywheel radial displacement curve
Control current curve
Time t (ms)
0 100 600500400300200 700Time t (ms)
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Control current curveeee
Flywheel radial displacement curve
Figure 7 Flywheel displacement and control current in start-upprocess by 119867infin control
and ensuring the flywheel to rotate around its collectioncenter
42 Experimental Results In Figure 7 the radial displace-ment voltage curve and the control current curve of theflywheel are shown in start-up process by the solved 119867infin
controller control the displacement voltage quickly drops tothe equilibrium position (25 V) from the sensor calibrationposition (4V) It illustrates that the 119867infin controller has goodrobust stability and can be used in the control system ofvehicle maglev flywheel
Figure 8 includes the radial displacement curve thecontrol current curve and the speed measuring pulse whenflywheel normal is rotating by 119867infin control and flywheel
0 10 50403020
2
4
3
2
1
06
Time t (ms)
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Con
trol c
urre
nti
(A) Ra
dial
disp
lace
men
tx(V
)
2
4
6
C1 flywheel radial displacement (1Vdiv)
C1
C2
C4 C4 control current (2Adiv)
C2 speed measuring signal (0sim33V)
Figure 8 Flywheel displacement and control current in rotatingprocess by 119867infin control and the speed measuring pulse at 600Hz
rotational frequency is 600Hz The displacement curve is asine wave and it can show the flywheel has mass unbalanceTo restrict flywheel radial run-out the control current alsois sine having basic consistent phase with the displacementvoltage The speed measuring pulse has a voltage range from0V to 33 V the purpose is to guarantee the pulse be capturedby DSP acquisition circuit
The compensated effect by AILC algorithm at 600Hzis shown in Figure 9 including the radial displacementcurve and the control current curve after compensationCompared with the radial sine displacement in Figure 8 theradial displacement is balance in the position of 25 V whichindicates the flywheel is limited rotating around its geometriccenter To limit the radial run-out and increase the activecontrol effect the control current amplitude is significantlylarger than without compensation
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Mathematical Problems in Engineering
00
2
1
0
3
10 50403020
4
2
6
Flywheel radial displacement curve
Control current curve
Con
trol c
urre
nti
(A)
Radi
al d
ispla
cem
entx
(V)
Time t (ms)
0 10 50403020Time t (ms)
Control current curve
Figure 9 Flywheel displacement and control current in AILC com-pensation process at 600Hz
0 50 100 150 2000
5
10
15
20
25
30
35
Start Speed upConstant speed
Flywheel not workingOnly PID controlDual-model switch control
Constantspeed
Time t (s)
Out
put p
ower
P(k
W)
Figure 10 EV primary battery output power in different modes
Figure 10 gives the output power curves of EV primarybattery in three different modes If flywheel is not workingthe output power is about 35 kW when EV starts up if theflywheel battery is working and only PID controlling theoutput power of the primary battery is about 25 kW in EVstart-up process if the dual-model switch control strategy isacting on the flywheel the output power reduces to 23 kW
The power change shows the flywheel has better auxiliaryeffect to EV primary battery when it is controlled by the dual-model switch control strategy Similarly the output powerof EV primary battery is the smallest in the accelerationprocess when the dual-model switch control strategy controlsthe flywheel Moreover because the flywheel rotate speedis higher when the dual-model switch control strategy isworking EV needs lower motive power when restartingTherefore when maglev flywheel participates in dischargeand is controlled by dual-model switch control strategyEV primary battery output power is obviously decreasedin the whole driving process the influence on chemicalcharacteristics of primary battery is reduced because theinstantaneous discharge depth is small and the service lifeof EV primary battery can be improved
5 Conclusion
Through analyzing the nonlinear dynamic characteristicsand the application on EV one kind of dual-model controlstrategy based on 119867infin control and AILC algorithms hasbeen studied in this paper and the state space equationof maglev flywheel was analyzed To improve the robuststability of flywheel control system and reduce the real-time interference one 119867infin controller based on the statespace equation was solved To reduce maglev flywheel radialrun-out one adaptive iterative learning control theory wasdeduced and unbalance displacement compensation wasimplementedThe experimental result shows the dual-modelcontrol strategy has better interference capability in maglevflywheel start-up process and has stronger active controlability relative to only PID control The maglev flywheelbased on the dual-model control can help the EV primarybattery improve its discharge characteristics help to prolongits service life and accelerate the development scale of electricvehicles
Conflict of Interests
The authors declare that they have no conflict of interestsregarding this paper
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (51405244) the China PostdoctoralScience Foundation (2014M551634) and the Jiangsu ProvinceNatural Science Foundation of China (BK20140880) Theauthors would like to thank the editor and the reviewers fortheir constructive comments and suggestions to improve thequality of the paper
References
[1] E Sortomme Optimal aggregator bidding strategies for vehicle-to-grid [PhD thesis of Philosophy] University of Washington2011
[2] H-F Dai Z-C Sun and X-Z Wei ldquoTechnologies to relief un-uniformity of power batteries used in electrical vehiclesrdquo Jour-nal of Automotive Safety and Energy vol 2 no 1 pp 62ndash67 2011
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 9
[3] B Bolund H Bernhoff and M Leijon ldquoFlywheel energyand power storage systemsrdquo Renewable and Sustainable EnergyReviews vol 11 no 2 pp 235ndash258 2007
[4] M Jiang J Salmon and A M Knight ldquoDesign of a permanentmagnet synchronous machine for a flywheel energy storagesystem within a hybrid electric vehiclerdquo in Proceedings of theIEEE International Electric Machines and Drives Conference(IEMDC rsquo09) pp 1736ndash1742 May 2009
[5] J Abrahamsson J de Santiago and J G Oliveira ldquoPrototypeof electric driveline with magnetically levitated double woundmotorrdquo in Proceedings of the 19th International Conference onElectrical Machines (ICEM rsquo10) pp 1ndash5 September 2010
[6] M S Raymond M E F Kasarda and P E Allaire ldquoWindagepower lossmodeling of a smooth rotor supported by homopolaractive magnetic bearingsrdquo Journal of Tribology vol 130 no 2pp 1ndash8 2008
[7] Y-L Xu and J-C Fang ldquoDevelopment of low power loss energystorage flywheelrdquo Transactions of China Electrotechnical Societyvol 23 no 12 pp 11ndash16 2008
[8] T A Aanstoos J P Kajs W G Brinkman et al ldquoHigh voltagestator for a flywheel energy storage systemrdquo IEEE Transactionson Magnetics vol 37 no 1 pp 242ndash247 2001
[9] H Qing-kai Nonlinear Vibration Analysis and DiagnosisMethod for Fault Rotor System Science Press Beijing China2010
[10] H Gao L Xu and Y Zhu ldquoUnbalance vibratory displacementcompensation for active magnetic bearingsrdquo Chinese Journal ofMechanical Engineering vol 26 no 1 pp 95ndash103 2013
[11] Z Gosiewski and A Mystkowski ldquoRobust control of active mag-netic suspension analytical and experimental resultsrdquoMechan-ical Systems and Signal Processing vol 22 no 6 pp 1297ndash13032008
[12] J-M Kim C-K Kim M-K Park and K-H Kim ldquoA robustcontrol of a magnetic suspension system with a flexible railrdquo inProceedings of the IEEE International Symposium on IndustrialElectronics (ISIE rsquo01) vol 2 pp 1309ndash1312 Pusan Republic ofKorea June 2001
[13] H G Chiacchiarini and P S Mandolesi ldquoUnbalanced compen-sation for active magnetic bearings using ILCrdquo in Proceedings ofthe IEEE International Conference on Control Applications pp58ndash63 Mexico City Mexico September 2001
[14] C Bi D Wu Q Jiang and Z Liu ldquoAutomatic learning controlfor unbalance compensation in active magnetic bearingsrdquo IEEETransactions on Magnetics vol 41 no 7 pp 2270ndash2280 2005
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of