Sunteți pe pagina 1din 8

Deadbeat-Direct Torque and Flux Control of IPMSM Drives

using a Minimum Time Ramp Trajectory Method


at Voltage and Current Limits
Jae Suk Lee and Robert D. Lorenz
University of Wisconsin Madison, WEMPEC
lorenz@engr.wisc.edu
Abstract This paper presents the voltage and current limited
operation of an interior permanent magnet synchronous
machine (IPMSM) using deadbeat, direct torque and flux
control (DB-DTFC). A commanded air-gap torque and stator
flux can be achieved by the end of each PWM period using DBDTFC. However, it may take several PWM periods to achieve
a desired air-gap torque that is physically infeasible in one step
due to voltage limits. In that case, the torque and flux
command trajectories operating over multiple periods can be
developed to achieve deadbeat torque and flux response for
every PWM period.
The torque and flux command
trajectories can be developed in different shapes depending on
desired objectives. In this paper, a minimum time ramp
trajectory method is proposed to achieve both simple real time
implementation and fast and stable transient dynamics of
IPMSM drives. Simulation and experimental results for the
minimum time ramp trajectory method for an IPMSM drive
are presented.

I.

INTRODUCTION

The groundwork for finite settling time (FST) control in a


digital control system was laid in [1]. Minimum settling
time can be achieved via deadbeat control, which is a
special case of FST control [2,3]. DB-DTFC applies the
concepts of deadbeat control to achieve the goals of direct
torque control (DTC). Volt-second solutions can be
computed via DB-DTFC that cause the air-gap torque and
stator flux linkage to achieve the desired value in one
inverter switching period. As a result, smooth and fast, one
step torque and stator flux response can be achieved via DBDTFC using rather simple methods.
DB-DTC was
introduced for induction motor (IM) drive systems [4] as an
algebraic voltage solution. An alternative, easily
implemented solution, DB-DTFC was developed which
directly and simply utilizes the inherent Volt-second (flux
linkage) source properties of power electronics over each
switching interval [5]. The DB-DTFC methods have been
implemented for IM drives [5] and IPMSM drives [6].
When DB-DTFC is not in voltage limits, a range of

978-1-4799-0336-8/13/$31.00 2013 IEEE

feasible Volt-sec (flux linkage) vectors can all achieve the


desired change in air-gap torque in one PWM period, i.e.
deadbeat response. However, deadbeat response may not
be feasible when the available voltage is insufficient.
Therefore, it may take several PWM time steps when very
large change in torque is commanded especially at voltage
and current limits. Since the dynamic performance of motor
drives is constrained during voltage- and current-limited
operation, many solutions have been proposed. A single
step optimization has been presented for current vector
control (CVC) [7, 8] and DTC [9]. The maximum rate of
torque changes at the next sample time instant can be
achieved via single step optimization. However, the
dynamics of current or flux in motor drives has not been
taken into account in the single step optimization method
which may result in undesirable motor drive dynamics.
Model predictive control (MPC) has also been applied to
optimize current or torque transient dynamics at voltageand current-limited conditions [10, 11]. A cost function that
minimizes current or torque response time is developed and
future dynamics of motor drives are predicted based on
inverse machine and/or inverter models. When using the
predicted dynamics, an optimum switching plan that
minimizes the time to achieve the desired dynamics is
determined from the MPC algorithm. However, direct
control of inverter switching status results in unnecessary
computational work when a motor drive is not operated
under voltage-limited conditions.
Trajectory-based methods have been proposed as an
alternative multiple-step optimization method at voltage and
current limits [12-15]. Time optimal current trajectories
have been developed that minimize the time required to
achieve a desired current [12-13]. However, the fastest
transient current dynamics do not necessarily guarantee the
fastest transient torque dynamics due to the nonlinear
relationship between torque and current in IPMSMs. A time
optimal torque trajectory method was proposed for DB-

1778

II.

red torque line represents the loci of the Volt-sec. solutions


* (k+1), by the
that will achieve the desired air-gap torque, Tem
end of the current sample (switching) period. The hexagon
is the inverter Volt-sec constraint of a motor drive at each
sampling (switching) period. The dashed blue vectors
s (k+1)) represent the feasible stator flux linkage vectors
(dqs
at the next sample time (end of the switching period). As
seen in Fig.1, multiple Volt-sec vectors exist that achieve the
* (k+1), in one PWM period during
desired air-gap torque, Tem
normal operation. Each Volt-sec yields a different stator flux
linkage at the next sample time instance. During voltagelimited operation, no Volt-sec solution exists to achieve the
desired torque in the next sample time instant. Therefore, it
takes several PWM periods to achieve a desired torque when
a very large torque is commanded under voltage-limited
conditions. The infeasible single step torque command can
be shaped as a feasible multi-step trajectory so that deadbeat
torque and flux can be achieved for every PWM sample
instant.
When Vdc=200 [V]
Torque changes over
one switching period [p.u.]

DTFC IPMSM drives using dynamic programming and


Pontryagins maximum principle as its optimization
methods [14]. The fastest transient dynamics and deadbeat
response at every PWM instant can be achieved along the
time optimal trajectory. However, the simplification of the
time optimal algorithm is required for a practical real time
implementation.
A practical, real-time ramp trajectory method was
proposed for DB-DTFC IM drives [15]. The ramp
trajectory is simple to implement and fast and stable
transient dynamics can be achieved.
However, the
previously proposed method for computing the time for the
ramp trajectory was significantly sub-optimal. It was based
on a very simple to calculate voltage linearization between
inverter voltage limit and voltage required to achieve
deadbeat response. While simple to implement, it
consistently results in longer than necessary ramps and lack
of full utilization of the inverter voltage.
In this paper, a minimum time ramp trajectory method is
proposed for DB-DTFC IPMSM drives using the machine
and torque state equations of an IPMSM drive. Practical
methods for real-time calculation of the minimum time for a
ramp trajectory are the key focus. Fast and stable transient
torque dynamics and full use of the inverter voltage can be
achieved via the proposed minimum time ramp trajectory
method. The single step optimization method in [9], the
time optimal trajectory method in [14], the ramp trajectory
method presented in [15] and the proposed minimum time
ramp trajectory method are implemented and compared in
simulation and experiment.
VOLTAGE LIMITED OPERATION OF DB-DTFC
IPMSM DRIVES

Torque [p.u.]

The graphical representation of IPMSM dynamics as


seen by a DB-DTFC controller under both normal and
voltage-limited operation are shown in Fig.1.
Torque changes over

When Vdc=300 [V]


one switching period [p.u.]

d-axis [Volt-msec]

q-axis [Volt-msec]
(i) normal operation

Speed [p.u.]

q-axis [Volt-msec]
(ii) voltage-limited operation

Fig. 1. Graphical representation of IPMSM DB-DTFC Volt-sec. solution


for (i) normal operation and (ii) voltage-limited operation
s (k)) represents the
In Fig.1, the solid blue vector (dqs
stator flux linkage at the present sample instant. The solid

1779

Torque [p.u.]

Speed [p.u.]

Fig. 2. Torque changes, which can be developed over one sampling time
period at a wide operating space when *s= 0.12 [Volt-sec] and *s= 0.11
[Volt-sec] and Ts=100 [-sec]

For further investigation of voltage-limited operation of


IPMSM drives, the maximum torque changes in one inverter
switching period is calculated over a wide operating space
using simulation. The simulation results are shown in Fig. 2.
During the simulation, the DC bus voltage is limited to either
300 [V] or 200 [V] and the current magnitude is limited to
the rated current of the test IPMSM and the one switching
period is 100 [-sec]. Since torque and flux linkage are
control variables in DB-DTFC, current magnitude is
indirectly limited by constraining maximum and minimum
stator flux linkage magnitude as (1) and (2).
2

| idqs | =

ids + iqs and 0 | idqs | ilim

| min | | s | | max |
r

First of all, the torque and flux domain is quantized into


multiple stages and states. The next step is determining the
initial and final stator flux linkage of a trajectory. In this
research, the initial and final stator flux linkage values are
the optimal levels that minimize steady state copper and iron
losses at desired initial and final operating conditions. The
quantization of the torque and stator flux linkage domain
and the ramp trajectory are graphically represented in Fig. 3.

(1)
(2)

Where, ds = Ldids + pm, qs = Lqiqs


and | dqs | =

ds + qs

As shown in Fig. 2, the torque changes in one inverter


switching period are constrained by the voltage limits of a
motor drive and are dependent on the operating conditions.
Therefore, deadbeat torque response can be achieved when
* (k+1), is smaller than the
the desired change in torque, Tem
maximum feasible torque changes at the desired operating
conditions. Overall, larger torque changes can be achieved
at lower speed and higher torque operating conditions.
Although larger torque changes can be obtained at a higher
torque region, a current limit of a motor drive has to be taken
into account at a high torque region. Fig. 2 also shows
torque changes in one inverter switching period when the
different stator flux linkage magnitude is applied. Higher
torque changes can be achieved when a lower stator flux
linkage magnitude is applied. However, a large d-axis
current is required to reduce the stator flux linkage
magnitude and larger d-axis current may result in more
copper losses in an IPMSM.

III.

Fig. 3. Quantization of torque and flux domain to calculate the minimum


time along the ramp trajectory

The minimum time that it takes along the ramp trajectory


can be simply calculated using (3) and (4).
Tstage (k) =

* (k+1)
Tem

(3)

r
r (k), r (k), r (k))
f(vds(k),vqs
qs
ds

Where,
r

r (k), (k), r (k))


f(vds(k),vqs
qs
ds
r

L Lq) ds(k) + pm Lq
3
r (k)( d
= P vqs
4
Ld Lq

r(k)
2
r
r (k)2) - L r (k) )
( (Lq - Ld) (ds(k) - qs
q ds
pm
Ld Lq
r (k)
Rs qs
2 2
Ld Lq

Ld-Lq
Ld Lq

r (k)
((Lq2 - Ld) ds(k) - Lq2 pm)+vds(k)qs
2

n=k+N

MINIMUM TIME RAMP TRAJECTORY METHOD

As investigated in Fig. 2, the DC bus voltage provides a


maximum voltage limit for motor drives. When a large
change in torque is commanded, it takes several PWM
periods to achieve the desired air-gap torque due to the
voltage limit. In this paper, a minimum time ramp
trajectory method is proposed to obtain fast and stable
control of an IPMSM drive and full use of the inverter
voltage. The key part of the proposed method is estimation
of the minimum time while torque and flux change along
the ramp trajectory. The quantization used in dynamic
programming [16] is applied to calculate the minimum time.

Ttotal =

n=k

Tstage (n)

(4)

* (k+1) represents the desired torque change


In (3), Tem
over one PWM period. The total transient time along the
ramp trajectory can be estimated using (4), which is the sum
of the times between each stage along the ramp trajectory.
The number of stages in the torque and flux domain is
determined by the torque changes, which affects the
computation time of the total transient time along the ramp
trajectory. In this paper, the average of torque changes at the
desired operating condition is calculated using the simulation
results in Fig. 2 and is applied as the torque value between

1780

Torque [Nm]

(a) Time [msec]

Stator flux linkage [V-sec]

(g) Time [msec]

Current mag. [A]

required to achieve a deadbeat response (Vdqs ) using (5).


Ki is a coefficient that adds margin to allow the ramp
method to function well over a wide operating range.

Current mag. [A]

In the ramp trajectory algorithm, the sample time steps


(Nstep) for the ramp trajectory are determined by the ratio
( mi ) between the voltage limit ( Vlim ) and the voltage

(f) Time [msec]

(c) Time [msec]

Fig. 4. Block diagram of time optimal DB-DTFC

mi = | Vdqs | / Vlim and Nstep = Ki mi

(e) Time [msec]

(b) Time [msec]

Voltage mag. [V]

The single step optimization method in [9], the time


optimal trajectory method in [14], the ramp trajectory
method presented in [15] and the proposed minimum time
ramp trajectory method are implemented and compared via
simulation. The entire system level block diagram of DBDTFC for IPMSM drives is shown in Fig. 4.
When a deadbeat control is developed, it is assumed that
the sampling and computation take place in at the same
instant in time. However, a finite amount of time delay
exists due to the computation in a digital control system [17].
The delay yields oscillation and even instability in all
control systems including deadbeat control [5, 18]. The
next sample time value can be estimated by a properly
formed observer and the delay can also be compensated
using the observer. As seen in Fig. 4, a stator current
observer and a stator flux linkage observer are integrated
into a DB-DTFC system and are important parts to
implement DB-DTFC.

r=700[rad/s]=1[pu]

Voltage mag. [V]

EVALUATION VIA SIMULATION

r=200[rad/s]=0.286[pu]

Torque [Nm]

IV.

corresponds to 0.286 [pu] and 1 [pu] of rated speed for the


test IPMSM.

Stator flux linkage [V-sec]

each stage. Once the IPMSM achieves the desired torque,


the IPMSM control algorithm is returned from the minimum
time ramp trajectory method to the loss minimizing, single
step DB-DTFC. Therefore, the computational effort and
memory capacity can be saved when the IPMSM is not
operated in a voltage-limited transient condition.

(5)

During the simulation, a step torque command is changed


from 0.5 [Nm] to 1.5 [Nm] ( 0.448 [p.u.] - medium torque
changes) and 0 [Nm] to 2 [Nm] ( 0.896 [p.u.] - large torque
changes). The speed of the load motor is set at 200 [rad/s]
(medium speeds) and 700 [rad/s] (high speeds), which

1781

(d) Time [msec]

(h) Time [msec]

Fig. 5. Simulation results of transient torque and flux linkage dynamics


using the min. time ramp method (MTR), the ramp solution (RS), the
single step opt. method (1step) and dynamic programming (DP)
at r=200[rad/s] and r=700[rad/s], fsw = 10 [kHz] and Tem= 1 [Nm]

r=700[rad/s]=1[pu]

Torque [Nm]

Torque [Nm]

r=200[rad/s]=0.286[pu]

(e) Time [msec]

Stator flux linkage [V-sec]

Stator flux linkage [V-sec]

(a) Time [msec]

(f) Time [msec]

Voltage mag. [V]

Voltage mag. [V]

(b) Time [msec]

(c) Time [msec]

(g) Time [msec]

Current mag. [A]

Current mag. [A]

Since voltage is not sufficient to achieve the torque in one


PWM period, it takes several PWM time steps to achieve
the desired torque.The initial and final stator flux linkage
values of the stator flux ramp trajectory are chosen from the
look-up tables shown in Fig. 2 to minimize steady state
copper and iron loss at the operating conditions and reduce
the computation load. The simulation results of transient
torque response are shown in Fig. 5 and Fig. 6.
When the command torque is changed from 0.5 [Nm] to
1.5 [Nm], the torque and flux ramp trajectories can be
developed via the minimum time ramp trajectory method
(MTR), the ramp trajectory solution (RS), the single step
optimization (1step) and the time optimal trajectory
developed via dynamic programming (DP) as seen in Fig. 5
(a), (b), (e) and (f). Faster transient torque response and full
use of the inverter voltage can be achieved via the minimum
time ramp trajectory method compared to the ramp
trajectory method. As seen in Fig. 5, each trajectory method
besides the ramp solution shows very similar torque and
stator flux linkage dynamics regardless of operating speeds
when the torque change is not large.
Fig. 6 shows the transient torque and flux linkage
dynamics when a larger torque change is applied. Among
the trajectories, the time optimal flux trajectories in Fig. 6
shows significant changes compared to the time optimal
flux trajectories developed in Fig. 5. The time optimal
method yields faster transient torque dynamics than other
trajectory methods at a larger torque change and higher
speed operating conditions.
The time estimated using (5) is dependent on operating
conditions. Therefore, it sometimes results in a
comparatively slower torque response as demonstrated in
Fig. 6 (a). By applying the estimated minimum time along
the ramp trajectory, torque dynamics close to the time
optimal method can be achieved.
Based on the simulation results in Fig. 5 and Fig. 6, it is
concluded that optimization yields a more significant benefit
for larger torque changes and higher speed operating
conditions. The ramp trajectory method can be enhanced
and full inverter voltage utilization can be achieved by using
the estimated minimum transient time along the ramp
trajectory.

(d) Time [msec]

(h) Time [msec]

Fig. 6. Simulation results of transient torque and flux linkage dynamics


using the min. time ramp method (MTR), the ramp solution (RS), the
single step opt. method (1step) and dynamic programming (DP)
at r=200[rad/s] and r=700[rad/s], fsw = 10 [kHz] and Tem= 2 [Nm]

1782

EXPERIMENTAL EVALUATION OF VOLTAGE AND


CURRENT LIMITED OPERATION

The trajectory-based method DB-DTFC is implemented in


an inverter with a 10 kHz of PWM switching frequency.
Fig. 7 shows the experimental test set up with a SPMSM
load machine and the IPMSM test machine. A torque
detector is mechanically coupled to both motors to measure
the shaft torque of the motor drive testbed. The IPMSM
parameters and mechanical characteristics are shown in
Table.1.

r=700[rad/s]=1[pu]

Torque [Nm]

r=200[rad/s]=0.286[pu]

Torque [Nm]

V.

(e) Time [msec]

Stator flux linkage [V-sec]

Stator flux linkage [V-sec]

(a) Time [msec]

(b) Time [msec]

(f) Time [msec]

Voltage mag. [V]

Table 1. IPMSM parameters


Rs
1.4 []
Ld
8 [mH]
Lq
20 [mH]
pm
0.125 [Volt-sec]
Jp
1.0 x10-4 [kgm2]
Ts
100 [-sec]
Poles
4
Power
1.5 [kW]
Rated Torque
2.26 [N.m]
Rated Speed
6200 [rpm]

Voltage mag. [V]

Fig. 7. Experimental test set up, left-SPMSM load and right IPMSM test

1783

(g) Time [msec]

Current mag. [A]

During the experiment, the total transition time and the


initial and flux stator flux linkage are predefined based on
the test trajectory. This reduces the computational intensity
for a real time implementation of dynamic programming.
The single step optimization solution and the ramp solution
are also implemented experimentally as alternative solutions
for the improvement of voltage- and current- limited
operation of IPMSM drives. The experimental results of
voltage- and current- limited operation of the test IPMSM
drive are shown in Fig. 8 and Fig. 9.

Current mag. [A]

(c) Time [msec]

(d) Time [msec]

(h) Time [msec]

Fig. 8. Experimental results of transient torque and flux linkage dynamic


response using the min. time ramp method (MTR), the ramp solution
(RS), the single step opt. method (1step) and dynamic programming (DP)
at r=200[rad/s] and r=700[rad/s], fsw = 10 [kHz] and Tem= 1 [Nm]

r=700[rad/s]=1[pu]

Torque [Nm]

Torque [Nm]

r=200[rad/s]=0.286[pu]

(e) Time [msec]

Stator flux linkage [V-sec]

Stator flux linkage [V-sec]

(a) Time [msec]

(f) Time [msec]

Voltage mag. [V]

Voltage mag. [V]

(b) Time [msec]

(g) Time [msec]

Current mag. [A]

Current mag. [A]

(c) Time [msec]

(d) Time [msec]

(h) Time [msec]

Fig. 9. Experimental results of transient torque and flux linkage dynamic


response using the min. time ramp method (MTR), the ramp solution
(RS), the single step opt. method (1step) and dynamic programming (DP)
at r=200[rad/s] and r=700[rad/s], fsw = 10 [kHz] and Tem= 2 [Nm]

In Fig. 8, a step torque command is changed from 0.5


[Nm] to 1.5 [Nm] (0.448 [p.u.], i.e. a medium torque
change) when the load motor is set at 200 [rad/s] (0.286
[pu]) and 700 [rad/s] (1 [pu]). The torque and flux ramp
trajectories are developed via the minimum time ramp
trajectory method (MTR), the ramp trajectory solution (RS),
the single step optimization (1step) and the time optimal
trajectory developed via dynamic programming (DP) as
seen in Fig. 8 (a), (b), (e) and (f).
As verified via simulation and experiment, faster transient
torque response and full use of the inverter voltage can be
achieved via the minimum time ramp trajectory method in
comparison to the ramp trajectory method. The torque
response time via each trajectory method is almost identical
regardless of operating speeds when torque change is not
large.
Fig. 9 shows the transient torque and flux linkage
dynamic response when a larger torque change ( Tem
=0.896 [pu]) is applied and the load motor is set at 200
[rad/s] (0.286 [pu]) and 700 [rad/s] (1 [pu]). As seen in Fig.
9, the time optimal method yields faster transient torque
dynamics than other trajectory methods during a larger
torque change and higher speed operating conditions. The
experimental results closely match the simulation results
shown in Fig. 5 and Fig. 6. Based on the simulation and the
experimental results, a more significant benefit can be
obtained by applying optimization methods during larger
torque changes and higher speed operation. The ramp
trajectory method can be improved and inverter voltage can
be more fully utilized by using the estimated minimum
transient time along the ramp trajectory.
During the experiment, the limitation of the minimum
time ramp trajectory method is found. The number of
quantized stages and states are determined by the maximum
torque changes at a specific operating condition (shown in
Fig. 2) and the commanded torque change in one inverter
switching period. When the number of quantized stages and
states is too large, the minimum time ramp trajectory
algorithm implementation is less attractive due to the
intense computational load. In that case, the approximate
minimum time along the ramp trajectory is estimated via a
linear scaling method. For example, five stages are the
maximum number of stages that enable real time
implementation of the minimum time ramp trajectory
method. When the number of quantized stages is ten, the
minimum time which the first five stages take is calculated
using (3) and (4) and thus the total minimum time along the
entire trajectory can be estimated by multiplying by two.
Even though the accuracy of the minimum time calculation
along a ramp trajectory is degraded as the number of
quantized stages increases, the minimum time ramp
trajectory method still achieves fast and stable torque and
flux control as well as a full utilization of the inverter
voltage during voltage- and current- limited operation.

1784

VI.

CONCLUSIONS

REFERENCES

The new contribution of this work is development of a


method to compute real time minimum time torque and flux
ramp trajectories during voltage-limited operation. The key
conclusions are summarized as follows.

The proposed minimum time ramp trajectory method is


implemented in real time via simulation and experiment
and shown to yield fast and stable transient torque and
flux dynamic response with trajectories that are close to
those formed for time optimal control but with less
intense computational load than time optimal control.

Methods were developed to calculate the minimum


time along the torque and flux ramp trajectories during
voltage-limited operation by applying the quantization
method used in dynamic programming.

The maximum torque changes in one inverter switching


period can be used to determine the number of
quantized stages for the minimum time ramp trajectory
method.

Fast transient torque dynamics response can be


achieved via the single step optimization method at
voltage- and current- limits. However, single step
optimization method yields undesirable flux dynamics.

By applying optimization methods, faster transient


torque dynamic response can be obtained for larger
torque changes and higher speed operation.

The inverter voltage can be more fully utilized via the


minimum time ramp trajectory method than the ramp
trajectory method.

The feasibility of deadbeat torque response can be


evaluated based on the calculated maximum torque
change data.

[1]

A.R. Bergen and J.R. Ragazzini, Sampled-data processing


techniques for feedback control systems, Trans. AIEE, Pt. II, vol.73,
pp.236-247, Nov. 1954

[2]

J.R. Ragazzini and G.F. Franklin, Sampled-data control systems,


McGraw-Hill, 1958

[3]

G.C. Goodwin and K.S. Sin, Adaptive filtering prediction and


control, Prentice-Hall, 1984

[4]

G. Griva, T.G. Habetler, F. Profumo and M. Pastorelli, Performance


evaluation of a direct torque controlled drive in the continuous PWMsquare wave transition region, IEEE Trans. Power Electr., vol. 10,
pp. 464471, Jul. 1995.

[5]

B.H. Kenny and R.D. Lorenz, "Stator and rotor flux based deadbeat
direct torque control of an induction machine", IEEE Trans. on Ind.
Appl., vol 39, no. 4, pp. 1093-1101, July/Aug. 2003

[6]

J.S. Lee, C. Choi, J. Seok and R.D. Lorenz, Deadbeat-direct torque


and flux control of interior permanent magnet synchronous machines
with discrete time stator current and stator flux linkage observer,
IEEE Trans. on Ind. Appl., vol. 47, No.45,Jul. /Aug. 2011.

[7]

S. Lerdudomsak, M. Kadota, S. Doki, S. Okuma, Novel techniques


for fast torque response of IPMSM based on space-vector control
method in voltage saturation region, 33rd IEEE IECON International
Conf., Nov. 2007

[8]

S. Lerdudomsak, S. Doki, S. Okuma, Voltage limiter calculation


method for fast torque response of IPMSM in overmodulation range,
Proc. of 35th IEEE IECON Intl. Conf., Porto, Portugal, Feb. 2009
H. Ghassemi, S. Vaez-Zadeh, A very fast direct torque control for
interior permanent magnet synchronous motors start up, Energy
Conversion and Management, May 2004
Y. Hozumi, S. Doki and S. Okuma, Fast torque control system of
PMSM based on model predictive control, The 35th Annual conf. on
the IEEE Ind. Society, pp.1151-1155, Nov. 2009
J. Ishida, S. Doki and S. Okuma, Fast torque control system of
PMSM based on model predictive control considering
overmodulation region, International Power Elect. Conf., pp.14031406,Jun. 2010
J. Choi and S. Sul, Generalized solution of minimum time current
control in three-phase balanced systems, IEEE Trans. on Ind. Electr.,
vol.45, pp.738-744, Oct. 1998
L. Xu and S. Li, A fast response torque control for interior
permanent-magnet synchronous motors in extended flux-weakening
operation regime, IEEE Electr. Machines and Drives Conf., pp.3336, June 2001
J.S.Lee, R.D. Lorenz and M.A. Valenzuela, Time optimal and loss
minimizing deadbeat-direct torque and flux control for interior
permanent magnet synchronous machines, IEEE Energy Conversion
Congress and Exposition., pp.2568-2575, Raleigh, U.S.A, Sept. 2012.
Z.D. Hurst, Evalutation of dynamic trajectories for deadbeat direct
torque and flux control during voltage limited operation. M.S. thesis,
University of Wisconsin Madison, 2011
B. Richard, "Dynamic Programming," Princeton University Press,
1957
N. West and R.D. Lorenz, Digital implementation of stator and rotor
flux linkage observers and a stator current observer for deadbeat
direct torque control of induction machines, IEEE Trans. on Ind.
Appl., vol.45, no.2, Mar./Apr. 2009
J. Lee, C. Kim and M. Youn, Deadbeat type digital controller for the
direct torque control of an induction machine, IEEE Trans. on
Power Electr., vol.17, no.5, Sept.2002

[9]

[10]

[11]

[12]

[13]

[14]

ACKNOWLEDGMENT
The authors wish to acknowledge the Wisconsin Electric
Machines and Power Electronics Consortium (WEMPEC) at
the University of Wisconsin Madison for motivation and
support of this project.

[15]

[16]
[17]

[18]

1785

S-ar putea să vă placă și