3424 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 7, JULY 2012
Using Full Order and Reduced Order Observers for
Robust Sensorless Predictive Torque Control
of Induction Motors
S. Alireza Davari, Student Member, IEEE, Davood Arab Khaburi, Fengxiang Wang,
and Ralph M. Kennel, Senior Member, IEEE
Abstract—In this paper, two kinds of observer-based sensorless
predictive torque control methods are proposed. The predictive
method is based on examining feasible voltage vectors (VVs) in
a prescribed cost function. The VV that minimizes the cost func-
tion is selected. A novel robust prediction model is presented. The
prediction model includes sliding mode feedbacks. The feedback
gains are assigned by the H-inf method. Two kinds of observers are
applied for flux and speed estimation, i.e., sliding mode full order
observer and reduced order observer. In order to verify the pro-
posed method, simulation and experimental results are presented
in wide speed range. A comparison of the two methods is performed
based on the results.
Index Terms—Induction motor, observer, predictive torque con-
trol, sensorless drive.
I. INTRODUCTION
T
HE PREDICTIVE torque control (PTC) method is of in-
terest as an alternative to the direct torque control (DTC)
method in applications where torque control is more important
than speed control, such as the traction, paper, and steel in-
dustries. The PTC method shows faster dynamic response and
causes less torque ripple compared to the DTC method because
of its characteristic, i.e., predictive control.
Different kinds of PTC methods have been investigated to
date. The direct mean torque control method was introduced to
control the mean value of the torque at a reference value [1]–[3].
The deadbeat control method calculates the voltage reference to
achieve a zero torque error in the next control step [4]. The model
predictive control (MPC) method determines the optimum volt-
age by using the explicit model of the motor and inverter by
minimizing a cost function [5]. In the MPC method, the criteria
for voltage selection is more flexible [6]. As explained in [5], [7],
and [8], if cost function minimization is performed by the trans-
fer function-based controlled auto regressive integrated moving
average (CARIMA) model, the method is called generalized
Manuscript received July 25, 2011; revised October 24, 2011; accepted
November 25, 2011. Date of current version April 3, 2012. Recommended for
publication by Associate Editor J. O. Ojo.
S. A. Davari and D. A. Khaburi are with the Center of Excellence for Power
Systems Automation and Operation, Department of Electrical Engineering,
Iran University of Science and Technology, Tehran 1311416846, Iran (e-mail:
adavari@iust.ac.ir; khaburi@iust.ac.ir).
F. Wang and R. M. Kennel are with the Institute for Electronics Drive
Systems and Power Electronics, Technische Universitaet Munchen, Munich
80333, Germany (e-mail: fengxiang.wang@tum.de; ralph.kennel@tum.de).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPEL.2011.2179812
predictive control. The mathematical process is time consum-
ing in this method. The finite control set model predictive control
(FCS-MPC) method uses another approach to minimizing the
cost function. In this method, the discrete nature of the power
converters is contemplated. In this approach, the feasible volt-
age vectors (VVs) are examined in terms of cost and the one
that minimizes the cost function is selected [9]–[12].
Thus far, the PTC method has not been adopted in many
industrial applications because it is implemented by means of
speed sensor in most cases. Thus, one of the main advantages
of the DTC method is not included in the PTC method. There
have been few investigations into implementing PTC without
using a speed sensor. In [13], a predictive method is used in a
neural network observer in order to estimate the speed. Some
investigations have proposed sensorless methods for predictive
current control in the FOC method [14]–[16]. In these investi-
gations reduced order observer (ROO) is used to avoid heavy
calculations because complex predictive methods are used. Fur-
thermore, these methods are not based on the direct control of
torque and the aforementioned features are not included. The
sensorless PTC method, in which the torque is controlled di-
rectly, is proposed in [17]. ROO can be unstable, especially
at low speeds, because of its open-loop instinct. To overcome
this problem, drift and offset compensation by considering the
accurate model of inverter and adaptive parameter estimation
has been proposed in [17]–[19]. In this method, however, the
accurate parameters of inverter model are needed. A sensor-
less deadbeat DTC is developed in [20]. The deadbeat control
method is not sufficiently robust [21] and this drawback can de-
teriorate the performance of the drive system in sensorless ap-
plications. Applying test voltage signals during PTC can reach
accurate speed estimation in reluctance motor control even at
very low speeds [22], [23]. This method reduces the dynamic
performance of the predictive method and can make audible
noise. In [24], the effect of saliencies without signal injection is
used. In [25], the method is improved in order to increase the
robustness.
The additional difficulty of sensorless PTC method versus
conventional sensorless methods is the combination of estima-
tion and prediction. A measured speed will be used during both
of the aforementioned steps if the PTC method is not applied
as the sensorless method. However, if a sensorless observer is
used for flux estimation and a sensorless prediction model is
adopted, the sensorless PTC will be achieved.
Regarding the observing stage, despite the fact that the
full order observer consists of speed-dependent terms, some
0885-8993/$26.00 © 2011 IEEE
资源评论
