field-weakening control for spmsm

Upload: mgecic

Post on 08-Aug-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/22/2019 Field-Weakening Control for SPMSM

    1/6

    Field-Weakening Control for SPMSM Based on

    Final-State Control Considering Voltage LimitTakayuki Miyajima1, Hiroshi Fujimoto1, and Masami Fujitsuna2

    1 The University of Tokyo, 5-1-5 Kashiwanoha Kashiwa Chiba, Japan2 DENSO CORPORATION, 1-1 Syouwacho Kariya Aichi, Japan

    Abstract SPMSMs (Surface Permanent Magnet SynchronousMotors) are employed for many industrial applications. SPMSMdrive systems should achieve fast torque response and wideoperating range. In this paper, in order to fulfill fast response infield-weakening region, field-weakening control based on final-

    state control (FSC) is proposed. FSC settles state variables in afinal state from an initial state with feedforward input duringfinite time. Previously, in the single-input single-output system,FSC considering input limit with linear matrix inequality wasproposed. It is adapted to SPMSM which is the two-input two-output system in this paper. Finally, simulations and experimentsare performed to show that the proposed method can achieve fastfield-weakening control.

    Index TermsSPMSM, voltage limit, PWM hold model, final-

    state control, linear matrix inequality

    I. INTRODUCTION

    Permanent magnet synchronous motors (PMSMs) are

    widely employed in many industries because of high effi-

    ciency and high power density. Especially, surface PMSMs

    (SPMSMs) are used as machine tool and electric power

    steering. In these applications, SPMSM drive systems are

    demanded high speed operation and fast torque response

    but output voltage is limited. Thus, in order to expand the

    operating range, field-weakening control applies in the linear

    range of inverter.

    Under the voltage limit, the manipulated variable is volt-

    age phase only. However, conventional current vector control

    determines d-axis and q-axis voltage independently but itcannot operate voltage phase directly. Therefore, in field-

    weakening range, fast torque response is not achieved. Authors

    proposed control methods for Interior PMSM (IPMSM) basedperfect tracking control [1] and PWM hold model [2] in

    overmodulation range [3][4]. These methods improve torque

    response with feedforward (FF) controller but these methods

    operate d-axis and q-axis voltage independently too. Thus, ifoutput voltage is limited, it is not desired that FF controllers

    mend response.

    Control method which can operate voltage phase directly is

    necessary. Many methods for fast torque response under the

    voltage limit had been studied in IPMSM. Methods which op-

    erate voltage phase with the voltage limiter [5][6][7], methods

    which control torque and modulation index by torque loop

    [8][9], and methods which operate voltage phase directly byfeedback (FB) controller [10][11] were proposed.

    In this paper, for the fastest response under voltage limit,

    a field-weakening control based on final-state control (FSC)

    [12] is proposed. In [13], FSC considering input limit with

    linear matrix inequality (LMI) was proposed for single-input

    single-output system. However, in the case of vector control,

    SPMSM is the two-input two-output system as dq-axis voltage

    and current. In this paper, field-weakening control is solvedin form of programming program whose evaluation function

    and constraint functions are quadratic function and quadratic

    inequality, respectively. This proposed method determines

    voltage phase directly but the FF input is generated considering

    the voltage limit. Thus, the proposed method takes voltage

    phase in consideration. In addition, the proposed method

    considers the voltage limit circle with transient term because

    it consists of PWM hold model.

    II. MODEL AND DISCRETIZATION

    A. dq Model of SPMSM

    The voltage equation of SPMSM in dq-axis is representedin the form of state equation as y = x = [id iq]

    T and u =[vd vq]T by

    x(t) = Ac(e)x(t)+ Bc

    u(t)

    [0 eKe

    ]T}(1)

    y(t)= Ccx(t) (2)Ac(e) Bc

    Cc 0

    :=

    RL ee RL

    1L 00 1L

    I 0

    (3)

    where vd, q are d-axis and q-axis voltage, R is stator windingresistance, L is inductance, e is electric angular velocity, id, q

    are d-axis and q-axis current, and Ke is back EMF constant.The motor torque T is given by

    T = Kmtiq (4)

    where Kmt = P Ke and P is the number of pole pairs. In thispaper, 2-phase/3-phase transform is absolute transformation.

    B. Discretization Based on PWM Hold

    In order to discretize a plant, a zero-order hold is generally

    applied. However, in the case of single-phase inverter, it

    cannot output arbitrary voltage but only 0 or E (E is dc-bus voltage of single-phase inverter). Therefore, in order to

    control instantaneous values precisely, the zero-order hold isnot suitable. When the pulse is allocated in the center of

    control period Tu, the plant in inverter drive system can bediscretized as follows [2].

    1

  • 8/22/2019 Field-Weakening Control for SPMSM

    2/6

    i

    d[k]

    i

    q[k]

    Td[k]

    Tq[k]

    iu[k]

    iw[k]

    e[k]

    uw

    dq iq[k]

    id[k]SVM

    C[z]

    C[z]

    Decoupling Control

    SPMSM+

    INVT[k]

    Tu

    Vdc

    Tu

    Vdc

    Current

    Reference

    Generator

    +

    +

    +

    +

    +

    +

    e[k]+e[k]

    Fig. 1. Block diagram of conventional method 1.

    daxis

    qaxis

    T

    Kmt

    iqmax

    iqmin

    voltage limit circle

    Fig. 2. Point at the intersection of voltagelimit circle with constant torque line.

    A continuous-time state equation of a plant is given by

    x(t) = Acx(t) + Bcu(t), y(t) = Ccx(t). (5)

    The precise discrete model in which the input is the ON time

    T[k] is obtained as

    x[k + 1] = Asx[k] + BsT[k], y[k] = Csx[k], (6)

    As := eAcTu, Bs := e

    AcTu/2BcE, Cs := Cc. (7)

    In derivation of (6) and (7), the voltage amplitude is consid-

    ered +E only. Therefore, if T[k] is negative, the voltageamplitude is E.

    C. PWM Hold Model of SPMSM

    (1) is discretized based on PWM hold. Here, it is assumed

    that the speed variation during one control period can be

    neglected. Hence, the back EMF eKe can be presumed as thezero-order hold. Therefore, the PWM hold model of SPMSM

    is designed as E = Vdc in (5) as follows:

    x[k + 1] = As(e)x[k] + Bs(e)T[k]

    Bs2(e)[

    0 eKe]T

    , (8)

    y[k] = Csx[k], (9)

    As(e) := eAc(e)Tu , Bs(e) := e

    Ac(e)Tu2 BcVdc,

    Bs2(e) := A1c (e)

    (eAc(e)Tu I

    )Bc, (10)

    where Vdc is dc-bus voltage of three-phase inverter, T =

    [Td Tq]T

    , and Td, q are d-axis and q-axis ON time.As(e), Bs(e), and Bs2(e) are functions of e. Thus,these functions are calculated again when rotor speed is

    changed.

    Here, in SPMSM drive systems with three-phase inverter,

    input of three-phase can be generated based on space vector

    modulation (SVM) [15].

    III. CONTROL SYSTEM DESIGN

    A. Conventional Method 1

    Fig. 1 shows the block diagram of conventional method 1.

    The decoupling control which is given by (11), (12) is applied.

    vrefd [k] = vd

    ref[k] e[k]Lqiq[k] (11)

    vrefq [k] = vq

    ref[k] + e[k](Ldid[k] + Ke) (12)

    The current PI controller is designed with pole-zero cancella-

    tion as follows:

    C(s) =Ls + R

    s, = 10Tu (13)

    By discretizing (13), C[z] is obtained by Tustin transformation

    with control period Tu. Here, if Va :=

    v2d + v2q > Vmax (Va

    is the voltage amplitude and Vmax :=

    32

    Vdc3

    is the maximum

    value of voltage amplitude.), the integrator of C[z] is stopped.Current reference generator generates d-axis and q-axis

    current references id and iq from the point at the intersection

    of voltage limit circle with constant torque line to make current

    amplitude minimum as Fig. 2. Here, T is torque reference.The generation of current references is represented by

    id[k] =2e KeL

    R2 + 2e

    L2+ (iq max iq )(iq iq min), (14)

    iq [k] =

    iq max if T[k] > Kmtiq max

    iq min elseif T[k] < Kmtiq min

    T[k]Kmt

    otherwise, (15)

    iq max := eKeR Va

    R2 + 2e L

    2

    R2 + 2e L2

    , (16)

    iq min := eKeR + Va

    R2 + 2e L

    2

    R2 + 2e L2

    . (17)

    If id[k] strengthens the field flux, id[k] = 0.

    In SVM, e(= 0.5eTu) is added to e of the dq/2-phasetransform for sampling error compensation [16]. The ON time

    limit in SVM is described by

    T[k] =

    {T[k]

    |T[k]|Tmax if |T[k]| > Tmax

    T[k] otherwise, (18)

    where Tmax

    (:=

    32

    Tu3

    )is the maximum value of dq-axis

    ON time vector amplitude and T[k] is the limiter output. Inthis paper, all limiters calculate in the same way as (18).

    B. Conventional Method 2

    The block diagram of conventional method 2 is shown in

    Fig. 3. It consists of the 2-DOF control system.

    The inverse system of (8) is obtained as

    Tff[k] = B1s (e)As(e)x[k]+ B

    1s (e)xd[k + 1]

    + B1s (e)Bs2(e)[

    0 eKe]T

    . (19)

    2

  • 8/22/2019 Field-Weakening Control for SPMSM

    3/6

    i[k]

    C2[z]

    +

    +

    +

    i[k]e[k]

    Tff[k]

    x[k]

    uw

    dq S(Tu)

    (SVM)HPWM

    e[k]

    Tu

    VdcI +e[k]

    e[k]Pn[z]

    Current

    Reference

    Generator

    T[k]B

    1s

    (e)As(e)x[k] + B1s

    (e)i[k]

    Tff[k]

    SPMSM

    INV+

    T[k]C1[z]

    +B1s

    (e)Bs2(e) [0 eKe]T

    Fig. 3. Block diagram of conventional method 2.

    i[k]

    C2[z]

    +

    +

    +

    i[k]e[k]

    Tff[k]

    x[k]

    uw

    dq S(Tu)

    (SVM)HPWM

    e[k]Tu

    VdcI +e[k]

    e[k]Pn[z]

    Current

    Reference

    Generator

    T[k]

    Tff[k]

    SPMSM

    INV+

    T[k]C1[z]

    Fig. 4. Block diagram of proposed method.

    From this stable inverse system, the FF controller C1[z] isdesigned. Therefore, if the plant is nominal and the input is

    not limited, C1[z] assures the perfect tracking at the sampletime. Here, x[k] := [id[k] iq [k]]T in Fig. 3 is the nominal

    output which takes the input limit in consideration. If the inputis limited, x[k] do not equal to the current reference delayedone sample i[k 1].

    The FB controller C2[z] suppresses the error e[k] betweenx[k] and i[k]. Here, C2[z] is same as the FB controller ofconventional method 1. However, in conventional method 2,

    anti-windup control is not necessary because x[k] considersthe input limit.

    C. Proposed Method

    Fig. 4 and 5 show the block diagrams of the proposedmethod and FF controller C1[z], respectively. The proposedmethod has two FF controllers. In steady state and transient

    state which does not cause the voltage limit, the FF controller

    which is represented by (19) applies. In the other state, the FF

    input is generated with FSC trajectory generator. The proposed

    method switches two types of FF controller but the inverse

    system is FSC as prescribed time interval is one. Therefore,

    the stability is assured.

    FSC transits an initial state to a final state by prescribed

    time interval. Here, FSC considering the voltage limit is ex-

    pressed as a programming program whose evaluation function

    and constraint functions are quadratic function and quadratic

    inequality in the form of LMI. The FF input called FSC

    trajectory is generated by solving these LMIs.

    In PWM hold model of SPMSM, a final state x[N] is

    i[k]

    FSC Trajectory Generator

    Tff[k]

    C1[z]

    B1

    s (e)As(e)x[k] + B1

    s (e)i[k]

    Inverse system

    +B1s

    (e)Bs2(e) [0 eKe]T

    Fig. 5. Feedforward controller C1[z] of proposed method.

    represented with an initial state x[0] as follows:

    Y = U (20)

    Y := x[N] ANs x[0] 2UEM F (21)

    :=

    AN1

    s Bs AN2

    s Bs Bs

    (22)

    2 :=

    AN1s Bs2 AN2s Bs2 Bs2

    (23)

    U :=

    TT[0] TT[1] TT[N 1]T

    (24)

    UEM F := eKe[

    0 1 0 1]T

    (25)

    Here, it is assumed that e is constant in this derivation.

    The FF input U which satisfies (20) is not determined

    uniquely. Thereupon, the square sum of the current error is

    minimized. The evaluation function J is represented by

    J =ETQE

    ,Q

    > 0,(26)

    E := [eT[1] eT[2] eT[N]]T, (27)

    e[k] := i x[k] =[

    id iq

    ]T x[k]. (28)

    From

    A :=

    As A2s A

    Ns

    T, (29)

    B :=

    Bs 0 0AsBs Bs 0

    ......

    . . ....

    AN1s Bs AN2s Bs Bs

    , (30)

    B2 :=

    Bs2 0 0

    AsBs2 Bs2 0...

    .... . .

    ...

    AN1s Bs2 AN2s Bs2 Bs2

    , (31)

    I :=

    (i)T (i)T (i)TT

    , (32)

    E is given by

    E = I Ax[0] BU B2UEM F. (33)

    Moreover, (8) is controllable. Thus, is full row rank. There,

    R2N(2N2) and R2N2 that fulfill = 0

    and = I are defined. In addition, U is obtained by

    U := [ ]U. (34)

    3

  • 8/22/2019 Field-Weakening Control for SPMSM

    4/6

    By assigning (34) to (20), Y = [I 0]U is given. Therefore,U is represented by

    U =[

    Y q]

    T , (35)

    where q R(2N2)1 is a free parameter.By substituting (33), (34), and (35) to (26), the evaluation

    function J is transformed as follows:

    J = R(q) + ST(q)QS(q), (36)

    Z := I Ax[0] BY B2UEM F, (37)

    R(q) := ZTQZ 2ZTQS(q), (38)

    S(q) := Bq. (39)

    With LMI, the condition which satisfies J < for provided

    is given by (40) [13]. R(q) S(q)T

    S(q) Q1

    > 0. (40)

    Next, the voltage limit is described with LMI. Voltage limits

    at each sampling points are described as

    TT[k]T[k] = T2d [k] + T2q [k]

    T2max (k = 0, 1, , N 1). (41)

    Here, g(i) R22N (i := 2k + 1) of which (1, i) th and (2,i + 1) th entries are 1 and other entries are 0 is defined. g(i)

    separates input vector T[k] as follows:

    T[k] = g(i)U(q). (42)

    With g(i), (41) is transformed as

    {g(i)U(q)}T {g(i)U(q)} T2max. (43)

    (43) is expressed with N LMIs byT2max U(q)

    Tg(i)T

    g(i)U(q) I

    0, (44)

    where i = 2k + 1, k = 0, 1, , N 1.

    By minimizing under (40) and (44), the FF input whichdrives an initial state x[0] to a final state x[N] within theprescribed time interval N for PWM hold model of SPMSMand minimizes square sum of current error to fulfill voltage

    limit is obtained. Moreover, the proposed method takes volt-

    age limit circle with transient term in consideration exactly

    because PWM hold model treats transient term as difference.

    IV. SIMULATION

    The parameters of SPMSM for the simulation are shown in

    Table I. The dc-bus voltage of the three-phase inverter Vdc is36 [V]. The control period Tu is 0.1 [ms].

    Fig. 6 shows simulation results when step torque referencewas given and the rotor velocity was 1000 [rpm]. Here, the

    voltage amplitude of current reference generator was set to be

    0.98Vmax. The dq-axis ON time vector amplitude Ta and the

    TABLE I

    PARAMETERS OF SPMSM

    Inductance L 1.80 [mH]

    ResistanceR

    0.157 []Pairs of poles P 4Back EMF constant Ke 74.6 [mV/(rad/s)]

    dq-axis ON time vector phase (voltage phase) are representedby

    Ta =

    T2d + T2

    q , (45)

    = tan1TdTq

    . (46)

    Dotted lines in Fig. 6(c), 6(g), and 6(k) are described Tmax.Conventional method 1 cannot operate the voltage phase

    because it consists of current vector control. Therefore, the

    voltage phase variation was slow and fast response was not

    achieved. In addition, when the voltage amplitude was not lim-

    ited, current response still is slow because anti-windup control

    made closed-loop discontinuous and conventional method 1 is

    only composed with FB controllers. Conventional method 2

    uses a FF controller and does not apply anti-windup control

    because nominal output takes voltage limit in consideration.

    Thus, it controlled faster than conventional method 1 but it

    cannot operate voltage phase. Therefore, by operating voltage

    phase, settling time should be able to be shortened.

    In the proposed method, Q = I and the FSC trajectory wasmade to be the shortest trajectory. The prescribed time interval

    was N = 40. During the beginning of response, variation of d-axis current is larger by decreasing q-axis current. This maded-axis current lager than d-axis current reference quickly.Furthermore, excess d-axis current made variation of q-axiscurrent large. Finally, the proposed method achieved the fastest

    response and the proposed method shortened the settling

    time a half as short as conventional method 2. In general,

    this operation is executed on experiential grounds or by FB

    controller but the proposed method optimally operates with

    final-state control.

    V. EXPERIMENT

    Experimental results are shown in Fig. 7. Experiments

    were performed in the same condition as simulations. The

    parameters of SPMSM and the control period were same too.

    From a point of view of velocity control by load motor, Vdcwas set to be 80 [V]. In the controller calculation, the nominal

    three-phase inverter dc-bus voltage was Vdcn = 36 [V] and theinput was multiplied by Vdcn/Vdc to compensate the differencebetween Vdc and Vdcn in SVM. in experiments is 0.9eTuto consider the computational time 0.4Tu.

    Variations of d-axis reference in all experimental results are

    due to speed variation. In conventional method 1, the voltagephase variation was slow and anti-windup control deteriorates

    current response. Conventional method 2 improved the settel-

    ing time a little by FF controller.

    4

  • 8/22/2019 Field-Weakening Control for SPMSM

    5/6

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (a) id (Conventional 1)

    0 10 20 30 400

    2

    4

    6

    8

    10

    12

    14

    Time [ms]

    Current[A]

    iq

    iq

    *

    (b) iq (Conventional 1)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    Amplitudeofinput[ms]

    (c) Ta (Conventional 1)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    Angle

    ofInp

    ut[rad]

    (d) (Conventional1)

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (e) id (Conventional 2)

    0 10 20 30 400

    2

    4

    6

    8

    10

    12

    14

    Time [ms]

    Current[A]

    iqiq

    *

    (f) iq (Conventional 2)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    A

    mplitudeofinput[ms]

    (g) Ta (Conventional 2)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    AngleofInput[rad]

    (h) (Conventional 2)

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (i) id (Proposed)

    0 10 20 30 400

    2

    4

    6

    8

    10

    12

    14

    Time [ms]

    Current[A]

    iq

    iq

    *

    (j) iq (Proposed)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    Amplitudeofinput[ms]

    (k) Ta (Proposed)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    AngleofInput[rad]

    (l) (Proposed)

    Fig. 6. Simulation results.

    In the proposed method, previously, the FSC trajectory

    whose the prescribed time interval is N = 40 was calculatedoff-line. When torque reference was obtained, the FSC trajec-

    tory was output as FF input. If the response is not settled,

    FF input was generated from (19). When torque reference

    was obtained, q-axis current was decreased. Therefore, duringthe beginning of response, q-axis current raised the slowestresponse. However, this made larger variation of d-axis cur-rent. Therefore, q-axis response was fast by excess d-axiscurrent. d-axis and q-axis current were not settled within theprescribed time interval because of speed variation but the

    proposed method achieved the fastest response.

    VI. CONCLUSION

    Field-weakening control for SPMSM based final-state con-

    trol considering voltage limit was proposed in this paper.

    The FF controller of proposed method is composed with

    PWM hold model. Therefore, the proposed method can take

    voltage limit circle with transient term in consideration strictly.Simulations and experiments were performed to show that

    current responses remarkably are deteriorated on voltage limit

    in conventional methods. On the other hand, the proposed

    method can achieve fast response on voltage limit because

    feedforward input operates voltage phase quickly.

    In our future paper, current response of PMSM on voltage

    limit will be analyzed. In this paper, prescribed time interval Nwas provided previously and FF input was calculated off-line.

    In our future work, online calculation will be realized.

    REFERENCES

    [1] H. Fujimoto, Y. Hori, and A. Kawamura: Perfect Tracking Controlbased on Multirate Feedforward Control with Generalized SamplingPeriods, IEEE Trans. Ind. Eletron., Vol.48, No.3, pp.636644, 2001.

    [2] K. P. Gokhale, A. Kawamura, and R. G. Hoft: Deat beat microprocessorcontrol of PWM inverter for sinusoidal output waveform synthesis,IEEE Trans. Ind. Appl., Vol.23, No.3, pp.901910, 1987.

    [3] T. Miyajima, H. Fujimoto, and M. Fujitsuna: Control Method forIPMSM Based on Perfect Tracking Control and PWM Hold Modelin Overmodulation Range, The 2010 International Power ElectronicsConference, pp.593-598, 2010.

    [4] T. Miyajima, H. Fujimoto, and M. Fujitsuna: Control Method forIPMSM Based on PTC and PWM Hold Model in OvermodulationRange -Study on Robustness and Comparison with Anti-Windup

    Control-, The 2010 IEEE Energy Conversion Congress and Exposi-

    tion, pp.28442850, 2010.

    [5] J.-K. Seok, J.-S. Kim, and S.-K. Sul: Overmodulation Strategy forHigh-Performance Torque Control, IEEE Trans. Power Electronics,Vol.13, No.4, pp.786792, 1998.

    5

  • 8/22/2019 Field-Weakening Control for SPMSM

    6/6

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (a) id (Conventional 1)

    0 10 20 30 400

    5

    10

    15

    Time [ms]

    Current[A]

    iq

    iq

    *

    (b) iq (Conventional 1)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    Amplitudeofinput[ms]

    (c) Ta (Conventional 1)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    Phase

    ofinput[rad]

    (d) (Conventional 1)

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (e) id (Conventional 2)

    0 10 20 30 400

    5

    10

    15

    Time [ms]

    Current[A]

    iqiq

    *

    (f) iq (Conventional 2)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    A

    mplitudeofinput[ms]

    (g) Ta (Conventional 2)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    Phase

    ofinput[rad]

    (h) (Conventional 2)

    0 10 20 30 4025

    20

    15

    10

    5

    Time [ms]

    Current[A]

    id

    id

    *

    (i) id (Proposed)

    0 10 20 30 400

    5

    10

    15

    Time [ms]

    Current[A]

    iq

    iq

    *

    (j) iq (Proposed)

    0 10 20 30 400.06

    0.065

    0.07

    0.075

    0.08

    Time [ms]

    Amplitudeofinput[ms]

    (k) Ta (Proposed)

    0 10 20 30 400.4

    0

    0.4

    0.8

    1.2

    Time [ms]

    Phase

    ofinput[rad]

    (l) (Proposed)

    Fig. 7. Experimental results.

    [6] B.-H. Bae and S.-K. Sul: A Novel Dynamic Overmodulation Strategyfor Fast Torque Control of High-Saliency-Ratio AC Motor, IEEE Trans.Ind. Appl., Vol.41, No.4, pp.10131019, 2005.

    [7] S. Lerdudomsak, S. Doki, and S. Okuma: Voltage Limiter Calcula-tion Method for Fast Torque Response of IPMSM in OvermodulationRange, The 35th Annual Conference of the IEEE Industrial ElectronicsSociety, pp.13851390, 2009.

    [8] K. Kondo, K. Matsuoka, Y. Nakazawa, and H. Shimizu: Torque feed-

    back control for salient pole permanent magnet synchronous motor atweakening flux control range, IEEJ Tran. IA , Vol.119, No.10, pp.11551164, 1999 (in Japanese).

    [9] T.-S. Kwon, G.-Y. Choi, M.-S. Kwak, and S.-K Sul : Novel Flux-Weakening Control of an IPMSM for Quasi-Six-Step Operation, IEEETrans. Ind. Appl., Vol.44, NO.6, pp.17221723, 2008.

    [10] H. Nakai, H. Ohtani, E. Satoh, and Y. Inaguma: Development andTesting of the Torque Control for the Permanent-Magnet SynchronousMotor, IEEE Trans. Ind. Electron., Vol.52, No.3, pp.800806, 2005.

    [11] W. Hatsuse, Y. Notohara, K. Ohi, K. Tobari, K. Tamura, C. Unoko,and Y. Iwaji: A Stable Field-Weakening Control Using Voltage PhaseOperations in the High-Power Region, The 2010 International PowerElectronics Conference, pp.599604, 2010.

    [12] T. Totani and H. Nishimura: Final-State Control Using CompensationInput, Trans. of the SICE, Vol.30, No.3, pp.253260, 1994.

    [13] M. Hirata, T. Hasegawa, and K. Nonami: Seek Control of Hard DiskDrives Based on Final-State Control Tracking Account of the FrequencyCompensates and the Magnitude of Control Input, The 7th InternationalWorkshop on Advanced Motion Control, pp.4046, 2002.

    [14] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan: Linear MatrixInequalities in System and Control Theory, Society for Industrial and

    Applied Mathematics, 1994.[15] K. Sakata and H. Fujimoto: Perfect Tracking Control of Servo Motor

    Based on Precise Model with PWM Hold and Current Loop, The ForthPower Conversion Conference, pp.16121617, 2007.

    [16] J. Kudo, T. Noguchi, M. Kawakami, and K. Sano: Mathematical ModelErrors and Their Compensations of IPM Motor Control System, IEEof Japan Technical Meeting Record, IEE Japan, SPC-08-25, pp.25-31,2008 (in Japanese).

    6