You are on page 1of 6

Model Predictive Duty Based Torque and Flux Ripples Minimization of

Induction Motor Drive


Md. Habibullah∗ , Dylan Dah-Chuan Lu†
School of Electrical & Information Engineering, The University of Sydney, NSW, Australia
Email: ∗ md.habibullah@sydney.edu.au, † dylan.lu@sydney.edu.au

Abstract. This paper presents a duty cycle prediction scheme to be used in model predictive control (MPC) to minimize
both torque and flux ripples of an induction motor (IM) drive. The duty cycle is predicted using the model of the IM,
whereas analytical calculations are used in all of the prior duty cycle control methods. To minimize both torque and flux
ripples, two different cost functions are used: one for voltage vector selection and another one for duty cycle prediction.
Since the selection of the cost function in MPC is not a straightforward task, the proposed control strategy reduces the
dependency on the selection of cost function while minimizing torque and flux ripples significantly. The commutation
frequency of inverter is also reduced. Simulation results have revealed that, the proposed controller reduces the torque
ripple and average switching frequency by 56% and 21%, respectively while keeping the flux performance.

1. Introduction Duty cycle control is an effective approach to minimizing


torque and flux ripples. Several different approaches have
Direct torque control (DTC) has become a powerful and popu- been published in the literature to control this duty consider-
lar control method due to its simplicity since it was firstly intro- ing torque ripple minimization [18]-[19], fuzzy logic adoption
duced in the middle of the 1980s [1]. In a DTC based induc- [20], and both of torque and flux ripples minimization [21]-
tion motor (IM) drive fed by a two-level voltage source inverter [22]. The methods in [18]-[20] only considered the torque
(VSI), both of electromagnetic torque and stator flux are simul- ripple reduction and the flux ripple is increased. Both torque
taneously controlled by the application of one of the six active and flux ripples are considered in [21]-[22] but the robustness
voltage vectors and two zero voltage vectors. Inaccurate selec- of the system is deteriorated and the complexity is increased.
tion of the voltage vector and its duty cycle causes unexpected Moreover, the performance of the controller is degraded due
ripples in torque and flux which are the major concerns with to reduction in switching frequency.
a basic DTC drive. To minimize these ripples, several control Introducing duty cycle control in MPTF control can mini-
strategies have been proposed [2]-[8]. Among these strategies, mize torque and flux ripples significantly. Conventional MPTF
increasing attention has been dedicated to the use of model pre- controller employs the optimized voltage vector during the
dictive control (MPC) [9]-[12] due to its intuitive concepts and whole sampling period and this is one of the causes of steady
easy inclusion of nonlinearities and multivariable case in the state ripples. Hence, by applying the selected voltage vector in
model. The resulting controller is also easy to implement due a fraction part of the control period, the torque and flux ripples
to the present development of powerful control platforms. can be reduced. Recently, duty control based MPC is proposed
Model predictive torque and flux (MPTF) control is a very in [23]-[24] where analytical calculations are used to mini-
simple and effective alternative to conventional DTC. It is mize steady state errors/ripples. In [24], only torque ripple is
similar to DTC only in that they both directly manipulate the considered to calculate the optimal duty; hence the flux ripple
optimum voltage vector but their vector selection principles is increased.
are different. In MPTF control, the effect of each possible In this paper, the duty cycle of the active vector is predicted
voltage vector is evaluated using a cost function and the one using MPC and there are no need of analytical calculations.
which minimizes torque and flux ripples most is selected as Both of the torque and flux ripples are minimized simultane-
the optimum voltage vector. The performance of the MPTF ously using two different cost functions. Since the performance
controllers mostly depends on the selection of the cost func- of the MPC is mostly dependent on the selection of a suitable
tion. Several targets, variables, and constraints with weighting cost function and this is not an easy task, the proposed control
factors can be included in the cost function and simultane- strategy reduces the dependency on the selection of cost func-
ously controlled. Thus far, the research has revealed that the tion. An endeavor is also made to reduce average commutation
selection of weighting factors used in the cost function is a frequency. The effectiveness of the proposed controller has
key issue for MPTF control and not an easy task at all [13]. been compared with a controller without duty cycle control.
Different strategies to select accurate weighting factors have
2. Induction motor model
been proposed in [14]-[16] which require a comprehensive
mathematical analysis. However, vector selection for MPTF The state space model of induction motor in a stationary refer-
control is more accurate than conventional DTC [17]. ence frame (α, β ) is used and can be expressed by the follow-

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.
Md. Habibullah and Dylan Dah-Chuan Lu

 ܶ‫כ‬ ഥ௦

+- PI Duty Cycle d Pulse 2-level
߱ ‫כ‬ IM
ԡ<‫כ‬௦ ԡ Prediction Gen. VSI

෡ ୱ ሺ݇ሻ
< Ȧሺ݇ሻ ‫ݒ‬௢௣௧
෡ ୰ ሺ݇ሻ ܑ ሺ݇ሻ
< ୱ
߱
Cost Function
Optimization

ܶ ௣ ሺ݇ ൅ ʹሻ <୮ୱ ሺ݇ ൅ ʹሻ
࢏ୱୟ
Ȧሺ݇ሻ Torque/Flux Estimation ࢏ୱୠ
ഥ௦ 
࢜ and Prediction

Figure 1. Proposed control scheme.

ing Equations (1)-(6): The stator flux is estimated based on the stator voltage Equa-
tion (1). Using forward Euler approximation to discrete Equa-
dψs  r (k)
 s (k) and the rotor flux ψ̂
vs = Rsis + (1) tions (1) and (7), the stator flux ψ̂
dt estimations can be obtained as
dψ r
0 = Rrir + − jω ψr (2)
dt  s (k) = ψ̂
ψ̂  s (k − 1) + Tsv¯s (k) − Ts Rsis (k) (8)
ψ s = Lsis + Lmir (3)
 r (k) = ψ̂
ψ̂  r (k − 1)+
 r = Lmis + Lrir
ψ (4)    
  Lm  Rr 
(9)
T = 1.5pℑm ψ  s∗ .is (5) Ts Rr is (k) − − jω(k) ψ̂r (k − 1)
Lr Lr
dω where d is the predicted duty cycle of the voltage vector vs and
J = T − TL (6) v¯s = dvs . In conventional MPTF control, d is considered as 1.
dt
where vs is the stator voltage vector, is is the stator current The prediction technique of the duty cycle will be shown in
vector, ir is the rotor current vector, ψ
 s is the stator flux vector, section 3.4.
 r is the rotor flux vector, T is the electromagnetic torque, TL
ψ
is the load torque, ω is the rotor angular speed, p is the pole 3.2 Prediction of stator flux and torque
pairs, and the others are the system parameters.
The electromagnetic torque T and the stator flux ψ  s are pre-
3. Proposed control scheme dicted for the next sampling instant k+1. The stator flux
 sp (k + 1) is obtained by using forward-Euler dis-
prediction ψ
In conventional duty based torque and flux control, the duty
cretization of the same stator voltage equation presented in
cycle is calculated in analytical procedure and the calculations
Equation (1).
are very complex when both of torque and flux ripples are
considered [21], [22], [24]. Moreover, the performance of the  sp (k) + Tsv¯s (k) − Ts Rsis (k)
 sp (k + 1) = ψ
ψ (10)
system is deteriorated when only torque ripple [24] or switch-
ing frequency reduction [22] is considered. The proposed The prediction of the stator current is obtained as [25]
MPTF control in this paper considers model predictive method
 
to find out the optimal duty while minimizing both of torque  v¯
is + τσ d is = kr 1
− jω ψr + s (11)
and flux ripples and switching frequency. Fig. 1 presents the dt Rσ τr Rσ
overall control scheme of the proposed MPTF control which
includes estimation/prediction of the flux and torque, voltage where τr = Lr /Rr is the rotor time constant, σ =1 − Lm 2 /L L is
s r
vector selection, duty cycle prediction, pulse generation, and the total leakage factor, kr = Lm /Lr is the rotor coupling factor,
switching frequency reduction. Rσ = Rs + Rr kr2 is the equivalent resistance, and τσ = σ Ls /Rσ
is the transient time stator constant. Now the prediction of the
3.1 Estimation of flux stator current isp (k + 1) is obtained using forward-Euler dis-
cretization of Equation (11) as
In MPTF control, estimations of stator and rotor flux are
required. The rotor flux is estimated based on stator current  
model and can be expressed as isp (k + 1) = 1 + Ts is (k) + Ts
τσ (τσ + Ts )
      (12)
dψr Lm Rr 1 kr
= Rr is − − jω ψ  r. (7) − kr jω ψ r (k) + v¯s (k) .
dt Lr Lr Rσ τr

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.
Model Predictive Duty Based Torque and Flux Ripples Minimization of Induction Motor Drive

Finally, the electromagnetic torque is predicted using the pre-




∗

g =
T ∗ − T p (k + 2)
+ λψ
|ψs | − |
ψsp (k + 2)|
. (20)
 sp (k + 1) and stator currentisp (k + 1) accord-
dicted stator flux ψ
ing to Equation (5) In this study, λ is considered as 0.85 and λψ = Tnom /|ψsnom | .
 
T p (k + 1) = 1.5pℑm ψ  sp (k + 1)∗ .isp (k + 1) . (13) 3.4 Duty cycle prediction

3.3 Voltage vector selection After obtaining optimal voltage vector, the duty cycle of the
voltage vector is predicted using the same MPC. Firstly, duty
Optimal voltage vector is selected based on a conventional cycle is discretized in the range of 0.5 to 1.0 maintaining a step
MPC which includes three steps: estimation, prediction and of 0.1 in order to use the principle of finite state MPC. Then,
actuation. State-space model of IM for estimation and predic- the discrete values of duty cycles are used to actuate a classical
tion of the states and a priority based cost function for actua- cost function [Equation (20)]. The initial value of duty cycle is
tion are used to select the optimal voltage vector. In the cost assumed of 0.5 because there is no duty of less than 0.5 for an
function, priority is given on the stator flux in relation to the optimal active voltage vector at steady state condition and it is
torque. The cost function can be expressed as [10] seen by a heuristic procedure. The possible duties between two




steps, for example 0.55 or 0.65, have not been considered to



∗

g =
T ∗ − T p (k + 1)
+ λ p
|ψs | − |
ψsp (k + 1)|
(14) avoid the computational burden. During voltage vector selec-
tion in section 3.3, a priority based cost function [Equation
where T ∗ is the reference torque and T p (k + 1) is the pre- (18)] was used. Now, the duty cycle that minimizes the cost
dicted torque for a given switching state, ψ  s∗ is the reference function (20) is used as the duty of the optimal voltage vector
p

stator flux and ψs (k + 1) is the predicted stator flux, and λ p for the next sampling instant. The first cost function during
is the weighting factor which sets the relative importance voltage vector selection produces more torque ripple and the
between torque and flux. In a two-level VSI, there are eight second cost function during duty cycle prediction minimizes
possible switching states and the state which minimizes g that ripple while keeping the flux performance. The reason
most is selected to produce the voltage vector to be applied to is that torque slope is more sensitive to zero vector than flux
the IM drives. In a real time implementation, calculation time slope. The algorithm of duty cycle prediction is shown in Fig.
of control algorithm introduces a time delay which must be 2.
compensated [25]. It is done by extrapolating the variables at
instant k + 1 and used as an initial condition for the prediction ௞
Measure ሺ࢏௞௦ ǡ ܸௗ௖ ǡ Z௞ ሻ

at instant k + 2. In the proposed model, only the stator current


is extrapolated at instant k + 1; stator and rotor flux at instant Estimate/predict (<s, <r)
k and k + 1 are assumed to be constant. Hence, to implement
the delay compensation scheme [26], the predicted flux and ௞
Select ሺ‫ݒ‬௢௣௧ ሻ
torque at instant k + 2 are obtained by

ψ  s (k) + Tsv¯s (k + 1) − Ts Rsis (k + 1)


 s (k + 2) = ψ (15) for d = 0.5:0.1:1
 
isp (k + 2) = 1 + Ts is (k + 1) + Ts
Predict {<s (d), T(d)}
τσ (τσ + Ts )
    (16)
1 kr
− kr jω ψ  r (k) + v¯s (k + 1) Evaluate g(d)
Rσ τr
 
T p (k + 2) = 1.5pℑm ψ  sp (k + 2)∗ .isp (k + 2) . (17) d 1?

As the sampling time is very small, T ∗ (k) can be considered as


Optimize dk+1
T ∗ (k + 2). Now, considering the calculation delay in real time
implementation, the cost function to minimize is






∗
Figure 2. Algorithm of duty cycle prediction.
g =
T ∗ − T p (k + 2)
+ λ p
|ψs | − |
ψsp (k + 2)|
. (18)

Here, λ p is considered as 5 for this simulation. To show the 3.5 Pulse generation
dependency on the cost function of the proposed controller,
Switching pulses applied to two-level VSI are generated using
another two cost functions [25], i.e. normalized [Equation
the optimal voltage vector vopt and the predicted duty d. After
(19)] and classical [Equation (20)], are also used with the
the predicted duration of the active vector, appropriate null
following structures:
vector “000” or “111” will be generated for the rest of the dura-
2
p
2 tion within the control period to minimize switching frequency
T ∗ − T p (k + 2)  s∗ | −
ψ
|ψ  s (k + 2)

and losses as well [22]. For example, if the present optimal


g= +λ (19)
Tn2 ψsn2 voltage vector is “110”, then “111” will be selected as a null

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.
Md. Habibullah and Dylan Dah-Chuan Lu

vector rather than “000”. As a result, only one switching state between them at rated speed with rated load in terms of torque
changes during a sampling interval. This technique of switch- ripple, flux ripple, and total harmonic distortion (THD) of
ing frequency reduction is well known and widely applied. stator current is listed in Table 2. There is an average ripple
reduction of 56% for the torque in the proposed predictive
3.6 Switching frequency reduction duty based control compared to the conventional control. It is
also obvious from THD values that the stator current of the
After accompanying appropriate null vector, switching fre- proposed control system is more sinusoidal.
quency can be minimized further by incorporating a switching
state function in the cost function. The resulting cost function

Stator Current, ia[A]


1
is expressed as






∗
0
g =
T ∗ − T p (k + 2)
+ λP
|ψs | − |
ψ p
s (k + 2)|
+ λn .n (21)
-1
where n is the number of switches that change when the
1.5
switching state is applied. It can be expressed as [25]

Torque [N-m]
1

N



0.5
n = ∑
Sx (k) − Sx (k − 1)

0
x=1
Stator Flux [Wb]
where Sx is the state of a switch. The weighting factor λn is 0.6
assumed of 0.009. The selection of λn is very straight-forward
0.3
because the desired torque and flux ripples are known.
Table 1. Induction motor parameters. 0
1
Rs = 47.7Ω Lm = 2.94H ψsnom = 0.6W b
Duty

Rr = 45.3Ω Np = 2 Tnom = 1.0Nm 0.5


Ls = 3.132H J = 0.0067Kg.m2 P = 175W
Lr = 3.132H ωr = 1395rpm isr = 0.46A 0
1 1.1 1.2 1.3 1.4 1.5
Time [s]
Figure 3. Dynamic responses of stator current, torque, stator flux, and duty
4. Simulation results at rated speed with sudden load change for the conventional control system
without switching frequency reduction.
The effectiveness of the proposed control strategy is evaluated
through simulations using MATLAB/Simulink software. The
parameters of IM used for simulation are listed in Table 1. The
Stator Current, ia[A]

inverter is characterized by a dc link of 415V. For the purpose 1

of prediction and actuation of the objective function, a sam-


pling time of 50μs is used. The gain coefficients of PI speed 0

control loop to generate reference torque are tuned using MAT-


LAB sisotool. Some simulated results are presented below. -1
Fig. 3 presents the waveform responses at rated speed 1.5
Torque [N-m]

(146rad/s) with an external rated load of 1.0N-m applied at 1


t=1.25s for the conventional MPC. From top to bottom, the 0.5
curves are stator current, torque, stator flux, and duty cycle(s), 0
respectively. Same responses have also been shown in Fig. 4
Stator Flux [Wb]

for the proposed control scheme. A priority based cost func- 0.6
tion is used in both cases for voltage vector selection. From
Figs. 3 and 4, it is seen that the torque ripple in the proposed 0.3

control is smaller than that of the conventional MPC while


0
their flux ripples are almost similar. Numerical comparison
1
Table 2. Comparison between predictive models with and without duty cycle
Duty

control. 0.5

Index Proposed Conventional


0
Total harmonic distortion 5.63% 7.23% 1 1.1 1.2 1.3 1.4 1.5
(THD) for ia Time [s]
Figure 4. Dynamic responses of stator current, torque, stator flux, and duty at
Torque ripple 0.028Nm 0.065Nm
rated speed with sudden load change for the proposed control system without
Flux ripple 0.014Wb 0.016Wb switching frequency reduction.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.
Model Predictive Duty Based Torque and Flux Ripples Minimization of Induction Motor Drive

Apart from the better performance obtained from duty based 1.5

Torque [N-m]
MPTF control, it reduces the dependency on the selection of 1

cost function in MPC by reducing the torque and flux ripples 0.5

significantly for any cost function and it is clearly seen by com- 0

Stator Flux[Wb]
paring Figs 5 and 6. Two different cost functions; classical and 0.6

normalized, are used in the voltage vector selection to verify 0.3


the effectiveness of the proposed control system. From Fig. 5, 0
1 1.1 1.2 1.3 1.4 1.5
it is seen that the conventional control system suffers from high Time [s]
torque and flux ripples for both of the cost functions, even for (a)
1.5
priority based cost function. But these ripples are reduced sig-

Torque [N-m]
1
nificantly for every cost function in the proposed control sys- 0.5
tem, as shown in Fig. 6. Hence, cost function selection is an 0
important factor for a conventional MPC system. The quanti-

Stator Flux[Wb]
0.6
tative comparisons among the cost functions including the pri-
ority based cost function at rated speed for the conventional 0.3

and the proposed control system are listed in Tables 3 and 4, 0


1 1.1 1.2 1.3 1.4 1.5
respectively. It is seen that the priority based cost function pro- Time [s]
(b)
vides better performance compared to the other two cost func-
Figure 6. Responses of torque and stator flux at 100rad/s with sudden load
tions. change for the proposed control scheme: (a) classical cost function and (b)
1.5
normalized cost function.
Torque [N-m]

1
Stator Current, i [A]

0.5 1
a

0
0
Stator Flux [Wb]

0.6

0.3 -1
1.5
Torque [N-m]

0
1 1.1 1.2 1.3 1.4 1.5 1
Time [s]
(a) 0.5
1.5
0
Torque [N-m]

1
Stator Flux [Wb]

0.5 0.6
0
0.3
Stator Flux [Wb]

0.6
0
0.3 1
Duty

0
1 1.1 1.2 1.3 1.4 1.5 0.5
Time [s]
(b)
0
Figure 5. Responses of torque and stator flux at 100rad/s with sudden load 1 1.1 1.2 1.3 1.4 1.5
change for a conventional control scheme: (a) classical cost function and (b) Time [s]
normalized cost function. Figure 7. Dynamic responses of stator current, torque, stator flux, and duty at
rated speed for the proposed control scheme with switching frequency reduc-
Fig. 7 illustrates the stator current, torque, flux, and duty tion.
cycle(s) to measure the effect of the control strategy on the
Table 3. Conventional predictive model with different cost functions.
switching frequency. The Fig. is obtained under rated speed
(146rad/s) with rated load (1N-m). It is seen that the average Index Priority Normalized Classical
commutation frequency is decreased by 21%, from 2.53kHz Total harmonic dis- 7.23% 8.79% 8.18%
to 2kHz, while the torque and flux ripples are almost sim- tortion (THD) for ia
ilar. This reduction of commutation frequency is useful for Torque ripple 0.065Nm 0.05Nm 0.067Nm
reducing the switching losses of inverter. It is also seen that Flux ripple 0.016Wb 0.026Wb 0.024Wb
THD of stator current (calculated up to 10kHz harmonics) is
slightly increased, from 5.63% to 6.10%, due to the reduc-
optimal voltage vector using MPC, optimal duration is pre-
tion of switching frequency. The quantitative results of the
dicted using the same MPC. The proposed model is superior
proposed control system with and without commutation fre-
to conventional MPTF control by minimizing the torque ripple
quency reduction are summarized in Table 5.
by 56% while maintaining the flux performance. Apart from
this improvement, the proposed model reduces the depen-
5. Conclusions
dency on the selection of cost function by minimizing torque
This paper proposes a predictive duty cycle control strategy and flux ripples significantly for any types of cost functions.
to reduce both of torque and flux ripples. After obtaining The switching frequency can also be significantly reduced by

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.
Md. Habibullah and Dylan Dah-Chuan Lu

Table 4. Proposed predictive model with different cost functions. [11] Linder, R. Kanchan, R. Kennel, and P. Stolze. “Model-based
predictive-control of electric drives”, Germany: Cuvillier Verlag
Index Priority Normalized Classical Gottingen, (2010).
Total harmonic dis- 5.63% 7.51% 6.89%
[12] S. Kouro, P. Cortes, R. Vargas, U. Ammann, and J. Rodriguez.
tortion (THD) for ia “Model predictive control–a simple and powerful method to
Torque ripple 0.028Nm 0.027Nm 0.035Nm control power converters”, IEEE Trans. Ind. Electron., 56(6),
Flux ripple 0.014Wb 0.024Wb 0.021Wb pp. 1826–1838, (2009).
[13] J. Rodriguez, M. p. Kazmierkowski, J. R. Espinoza, P.
Table 5. Quantitative comparison of the proposed control scheme with and Zanchetta, H. Abu-Rub, H. A. Young and C. A. Rojas. “State
without commutation frequency reduction. of the Art of finite control set model predictive control in power
electronics”, IEEE Trans. Ind. Informatics, 9(2), pp. 1003–1016,
Index without fav with fav (2013).
reduction reduction
[14] Y. Zhang, J. Zhu, and W. Xu. “Predictive torque control of per-
Total harmonic distortion 5.63% 6.10% manent magnet synchronous motor drive with reduced switching
(THD) for ia frequency”, in Proc. Int. Conf. Electr. Mach. Syst., pp. 798–803,
Average switching fre- 2.53kHz 2kHz (2010).
quency ( fav ) [15] S. Davari, D. Arab Khaburi, and K. Ralph. “An improved
Torque ripple 0.028Nm 0.028Nm FCS-MPC algorithm for induction motor with imposed opti-
Flux ripple 0.014Wb 0.015Wb mized weighting factor”, IEEE Trans. Power Electron., 27(3),
pp. 1540–1551, (2011).

incorporating a switching state function in the cost function. [16] T. J. Vyncke, S. Thielemans, T. Dierickx, R. Dewitte, M.
Jacxsens, and J. A. Melkebeek. “Design choices for the predic-
tion and optimization stage of finite-set model based predictive
control”, in Proc. Workshop PRECEDE, pp. 47–54, (2011).
References
[17] T. Geyer, G. Papafotiou, and M. Morari. “Model predictive
direct torque control-part I: Concept, algorithm, and analysis”,
[1] Takahashi and T. Noguchi. “A new quick-response and high- IEEE Trans. Power Electron., 56(6), pp. 1894–1905, (2009).
efficiency control strategy of an induction motor”, IEEE Trans.
Ind. Appl., IA-22(5), pp. 820–827, (1986). [18] J.-K. Kang and S.-K. Sul. “New direct torque control of induc-
tion motor for minimum torque ripple and constant switching fre-
[2] Y. Zhang and J. Zhu. “Direct torque control of permanent magnet quency”, IEEE Trans. Ind. Appl., 35(5), pp. 1076–1082, (1999).
synchronous motor with reduced torque ripple and commutation
frequency”, IEEE Trans. Power Electron., 26(1), pp. 235–248, [19] K.-B. Lee, J.-H. Song, I. Choy, and J.-Y. Yoo. “Torque rip-
(2011). ple reduction in DTC of induction motor driven by three-level
inverter with low switching frequency”, IEEE Trans. Power Elec-
[3] B. Singh, S. Jain, and S. Dwivedi. “Torque ripple reduction tron., 17(2), pp. 255–264, (2002).
technique with improved flux response for a direct torque con-
trol induction motor drive”, IET POWER ELECTRONICS, 6(2), [20] L. Romeral, A. Arias, E. Aldabas, and M. Jayne. “Novel direct
pp. 326–342, (2013). torque control (DTC) scheme with fuzzy adaptive torque-ripple
reduction”, IEEE Trans. Power Electron., 50(3), pp. 487–492,
[4] L. Tang, L. Zhong, M. Rahman, and Y. Hu. “A novel direct torque (2003).
controlled interior permanent magnet synchronous machine drive
with low ripple in flux and torque and fixed switching fre- [21] G. Abad, M. A. Rodriguez, and J. Poza. “Two-level VSC
quency”, IEEE Trans. Power Electron., 19(2), pp. 346–354, based predictive direct torque control of the doubly fed induc-
(2004). tion machine with reduced torque and flux ripples at low con-
stant switching frequency”, IEEE Trans. Power Electron., 23(3),
[5] Y.-S. Lai and J.-H. Chen. “A new approach to direct torque con- pp. 1050–1061, (2008).
trol of induction motor drives for constant inverter switching fre-
quency and torque ripple reduction”, IEEE Trans. Energy Con- [22] Yongchang Zhang and Jianguo Zhu. “A novel duty cycle con-
vers., 16(3), pp. 220–227, (2001). trol strategy to reduce both torque and flux ripples for DTC
of permanent magnet synchronous motor drives with switch-
[6] J. Beerten, J. Verveckken, and J. Driesen. “Predictive direct ing frequency reduction”, IEEE Trans. Power Electron., 26(10),
torque control for flux and torque ripple reduction”, IEEE Trans. pp. 3055–3067, (2011).
Ind. Electron., 57(1), pp. 404–412, (2010).
[23] Yongchang Zhang, Wei Xie, Zhengxi Li, and Yingchao. “Model
[7] T. Geyer. “Computationally efficient model predictive direct predictive direct power control of a PWM rectifier with duty
torque control”, IEEE Trans. Power Electron., 26(10), pp. 2804– cycle optimization”, IEEE Trans. Power Electron., 28(11),
2816, (2011). pp. 5343–5351, (2013).
[8] H. Zhu, X. Xiao, and Y. Li. “Torque ripple reduction of the torque [24] Yongchang Zhang and Haitao Yang. “Torque ripple reduction
predictive control scheme for permanent-magnet synchronous of model predictive torque control of induction motor drives”,
motors”, IEEE Trans. Ind. Electron., 59(2), pp. 871–877, (2012). In Proc. IEEE Energy Conversion Congress and Exposition
[9] J. H. Lee. “Model predictive control: Review of the three decades (ECCE), pp. 1176–1183, (2013).
of development”, Int. J. Autom. Syst., 9(3), pp. 415–424, (2011). [25] J. Rodriguez and P. Cortes. “Predictive control of power con-
[10] P. Cortes, M. P. Kazmierkowski, R. M. Kennel, D. E. Quevedo, verters and electrical drives”, John Wiley & Sons Ltd., (2012).
and J. Rodriguez. “Predictive control in power electronics and [26] P. Cortes, J. Rodriguez. “Delay compensation in model predic-
drives”, IEEE Trans. Ind. Electron., 55(12), pp. 4312–4324, tive current control of a three-phase inverter”, IEEE Trans. Ind.
(2008). Electron., 59(2), pp. 1323–1325, (2012).

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on December 03,2022 at 07:10:46 UTC from IEEE Xplore. Restrictions apply.

You might also like