AN OBSERVER-BASED DTC OF INDUCTION MOTORS DRIVEN BY 3-LEVEL INVERTER FOR IMPROVING LOW SPEED OPERATION

K. B. Lee*, J. H. Song**, I. Choy**, and J. Y. Choi**

DTC FOR THREE LEVEL INVERTER

In this section, the principle of DTC is briefly discussed and a torque control for 3-level inverter system is studied. Figure 1 shows space vector representation of 3-level inverter output voltages, in which subscripts z, h, i, f denote zero, half, intermediate, and full voltages, respectively. Since 27 possible choices for switching voltage vector selection exist in the 3-level inverters, appropriate selection of the inverter switching voltage vector is more complex than in the 2-level inverters.

* Korea Univ., KOREA ** ISCRC, KIST, KOREA

ABSTRACT Direct torque control algorithm for 3-level inverter-fed induction motors is proposed. Conventional selection method of the stator voltage vector shows problems of stator flux drooping phenomenon and undesirable torque control appeared especially at the low speed operation. To overcome these problems, a proposed method uses intermediate voltage vectors, which are inherently generated in 3-level inverters. An adaptive observer is also employed to estimate some state-variables and motor parameters, which takes a deep effect on the performance of the low speed operation. Simulation and experiment results verify effectiveness of the proposed algorithm. INTRODUTION
Direct torque controlled (DTC) induction motors have been widely used in many industrial application areas. The DTC method gives attractive performances in terms of fast torque response, simple control scheme, and robustness against the motor parameter variation. The i t h no speed sensor equipped DTC can be implemented w to induction motors. The DTC algorithm has been extended into the high power motor drives from small power ratings [ 1-21. As for 2-level inverter to drive induction motors, several types of respective DTC methods can be found in numerous literatures. On the contrary, research results on DTC schemes for 3-level inverter are hardly reported. One of distinct features, which are especially shown in high-power 3-level inverter applications, is that the inverter switching frequency is generally limited below at most IkHz, considering that higher switching frequency causes heavier cooling apparatus for switching semiconductors. This paper pays major attention on the low speed operation characteristics of induction motors driven by 3level inverters. A simple hysteresis control scheme for 3level DTC inverters is studied and a new method to cope with the stator flux-drooping problem appeared in the low speed operation region is proposed. In order to resolve this flux-drooping problem in case of 2-level inverter system, several methods such as a variable switching sector and an improved look-up table are used [3,4]. Also, an adaptive full-order observer [5] is employed to obtain good DTC performance based on the estimation of the stator flux, the rotor speed, and the stator resistance over the wide operation region.

Figure 1: Output voltage vectors of 3-level inverter From induction motor equations, neglecting the voltage drop across the stator resistance, relationship between the inverter output voltage vector and the variation of the stator flux can be expressed as A A , = (V, -

Cs R, ) tsp = v , tsp ~ ,

where, Y, : the inverter output voltage vector, t,s,, : sampling period. Equation (1) shows that an applied stator voltage vector produces a stator flux change. The amplitude of the stator flux change is proportional to the product of the applied voltage vector and the sampling period. The vectorial direction of the stator flux change keeps the same as that of the selected voltage vector. The stator and rotor fluxes are written as the following relation

Power Electronics and Variable Speed Drives,18-19 September 2000, ConferencePublication No. 475 Q IEE 2000
170

where,

(T=

L' . L,'L,

TABLE 1- Switching look-up table Torque in sector k Flux

The electromagnetic torque can be expressed as

L
70 60 50 IO
9.985
0.986

'k+l.h

'k+I,J

v:
VI

Vk+.?,h

vk+2. f

Equation. (3) shows that torque is determined by the stator flux magnitude and the stator flux phase angle with respect to the rotor flux. Assumed that the stator flux vector is in the k-th sector, selection of the respective stator voltage vector is described in Figure 1. The selection of yk+I,f,h and vk+,, f is able to increase the angle between the stator flux and the rotor flux. As a result, developed torque can be increased by the application of ?k+?,/h or ? k + ] , f h . It can be seen in Figure 1 that the stator flux is increased by the selection of yk+],f,h, and decreased by c k + r , f , h . If half voltage vector is selected, the lower slope of torque can be obtained.
Upper Torque Band

:.

i-

0.987

0.988

0.989

0.99

E-12.51 %98

,
0.988

,
0.996
TIne

1.004

1.012

1.02

(a)

Forward direction operation

. ?

-30 -10

, " -50 ,O -60

7 & ! 9

. ?

y -15 2-42.5
-7 8.98 0.388 0.396
rlne 1.001

40 12.5

1.012

1.02

Loner Torque Band

(b)
Rdcrc~cvduc

Reverse direction operation Figure 3: Torque control

-Loner Torque Band

-Upper Torque Band

Figure 2: Torque slope pattern of 3-level inverter The double torque hysteresis band method as shown in Figure 2 is applied for 3-level inverters. When torque comes down to the negative ,upper hystersis band, appropriate full voltage vector is chosen to increase the torque developed. When the controlled torque reaches the positive lower hysteresis band, the full voltage vector is changed into the respective half voltage. If torque increases over the positive upper torque band, zero voltage vector is applied to decrease torque value. The same rules for voltage vector selection can be applied for the reverse direction operation. The resulted switching look-up table is shown in Table I . Assumed that ds is located in the k-th sector, possible voltage vectors to increase torque are .;(+I,, , i i k + ] , h , c k + l J , and ?k+I,h. Among these voltage vectors, a certain voltage vector should be selected, considering the double torque hysteresis band and the stator flux condition. Meanwhile, only one voltage vector, ?, is selected to reduce the developed torque. The simulation results of the forward and reverse operation are shown in figure 3.

Z-LewI Torque Flippk

%Lod Torque Ripplc

To show effectiveness of 3-level DTC algorithm described above, some simulations have been performed. The parameters of the induction motor used are listed in Table 3. In this case, speed estimation is carried out by an equation-based estimator, which is described in the following equation,

It can be seen in Figure 4 that the DTC switching method above-mentioned causes the stator flux demagnetization phenomenon at the low speed operation. As seen in Figure 1, around the boundary between two sectors, there is no effective voltage vector that can assure an increased stator flux, in which the rotating stator flux vector moves its position from one sector to another sector. Another problem that deteriorates performance of the low speed operation is selection of the zero-voltage vector. At low speed region, the zerovoltage can not effectively assure control of torque reduction because the resultant phase angle reduction is not easily obtained due to the low value of the stator flux frequency. The basic switching method of 3-level DTC has to be modified to overcome these problems at the low speed operation.

171

voltage vector, vk,;, is selected instead of Vk+,,, , the demagnetization problem is resolved. To obtain effectively magnetizing effect around the boundary between two sectors, especially at the low speed region, modified look-up table as shown in Table 2 is used. The transition from conventional look-up table to modified one is taken by detecting a certain level of the drooping magnitude of the stator flux. When the magnitude of the stator flux is drooped below minimum fixed level, the basic look-up table, Table 1, is replaced with the modified look-up table, Table 2.

Figure 4: Low speed performance of the DTC; (from top to bottom)speed, flux magnitide, d-axis flux, and phase current

SWITCHING METHOD FOR LOW SPEED OPERATION

As mentioned in previous section, problems of the basic DTC algorithm are stator flux drooping and the application of the zero-voltage vector, especially at the low speed region. The demagnetization effect is chiefly caused by the non-linearity of the induction machine. This problem is appeared just after the position of the stator flux vector changes from one sector to another sector. The application of the zero-voltage vector also deteriorates performance at the low speed operation. This phenomenon is easily comprehended from equations (2),(3). To overcome this problem, the reverse voltage vector is applied to decrease torque instead of zerovoltage vector. The rapid reduction of torque at the low speed region can be obtained by this method, but it may cause the increase of the switching frequency and the torque ripple. The basic switching look-up table can not solve these problems that are occurred in the low speed operation. Even if the demagnetization problems can be worked out by a rotation of reference frame or the switching sector that are similar to methods for 2-level inverters, the phenomenon caused by the application of the zerovoltage vector can not be easily answered. Therefore, to solve these problems, a newly modified switching lookup table has to be applied in the low speed operation. In 3-level inverters, intermediate voltage vectors can give possibility to resolve the demagnetization problem, without considering the reference frame rotation, or the switching sector rotation adopted in the 2-level. As seen in Figure 5 , the switching sectors are divided into 12 sectors to choose the stator voltage vector precisely. The k-th sector is subdivided into the lower sub-sector and the upper sub-sector, each of which has the width of 30". In the lower sub-sector, the stator voltage vector, i$(+, f , which is determined according to Table I , becomes to be ineffective to boost the stator flux. If the intermediate

k-th ower subsector

k-th sector

Figure 5 : Switching method for low speed operation

TABLE 2 - Modified Look-up Table

Toraue

$ 0

: :
9

12.5 6.25

00.5
0.45
0.15

0. 6

1.1

1.4

1.7

7 0.3

q ~
8.5
0.8
1.1
T2nd.l

, l.+

,
1.7

1
2

Figure 6: Improved performance in low speed ; (from top to bottom)speed, flux magnitude, torque, and switching frequency

I72

Advanced control scheme

i 1."
1

...v...j

i
:

:
>

...........................................

I;

1
V4.e

..................... Double torque hysteresis comparator and nux comparator

rhllCOIIIp8mIDI

"---1l

v.

eakulatiai

.............................. Adaptive observer

Figure 7: Schematic diagram of the proposed control strategy

If the stator flux moves into the upper sub-sector, the appropriate voltage vector is determined according to Table 2. This selection is basically same as Table 1, except the reverse voltage vector is chosen instead of the zero voltage vector for torque control. Figure 6 shows the low speed operation of the DTC using the advanced look-up table in the speed region about 1 -%. rated speed. It is shown in this figure that problems of the demagnetization and the zero-voltage vector work out effectively even without the increase of the switching frequency. The performance of the proposed control scheme indicates that it is feasible for the low speed operation.

ADAPTIVE OBSERVER-BASED CONTROL

The overall control scheme is shown in Figure 7, which consists of speed controller, torque and flux comparators, advanced look-up table, adaptive observer, and 3-level inverter. Figure 8 shows the DTC operation with the modified look-up table applied in the low speed region when the variation of the stator resistance( 150%) is considered. Estimation performance makes worse owing to the stator resistance variations, which may deteriorate the drive performance by consequently introducing errors in estimating the stator flux, speed, and electromagnetic torque. This is critical at the low speed region in particular, because the voltage drop across the stator resistance constitutes significant proportion of the applied voltages. To get robust performance against the stator resistance variation, especially in the low-speed operation, adaptive observer-based DTC is used as shown in Figure 7.

6.25 1 h
8
0.2 1.2 2.2

3.2

4.2

5.2

;0.15/
0.2

,
1.2

,
2.2

j
5.2

rI

$1

3.2

4.2

SIMULATION AND EXPERIMENT RESULTS

To confirm validity of the proposed control algorithm, simulation and experiment have been carried out. The performance of the low speed operation is focused on the simulation and experiment. The following simulation results have obtained by using Boland C". All simulation results are executed with no load condition and same sampling frequency (12O[p]). Figure 9 shows several responses of observer-based DTC in the low speed region using the advanced look-up table even though the stator resistance increases up to 150ph] rated value. The proposed DTC scheme shows good responses against the parameter

Lb.2

1.2

2.2

1,nd SI

3.2

1.2

5.2

Figure 8: Equation-based DTC against stator resistance change( 150[%]); (from top to bottom)speed, flux magnitude, torque, and stator resistance

173

variation, whereas the equation-based estimator is easily deteriorated by stator resistance variation. The experimental set-up of the proposed control system is shown in Figure 10. It consists of 7.4[kW]induction motor, power circuit, main controller board DS 1003 containing TMS320C40, I/O board DS400 1, and A/D board DS2001. Experiment is executed with no s 3). load condition and same sampling frequency (120[ p Figure 11 shows speed estimation and phase current responses of the proposed DTC scheme in the forward and reverse operation. Figure 12 shows speed and torque estimation in the low speed operation. Good estimation of the speed is achieved down to about 1% rated speed. Figure I3 shows the stator resistance estimation value in using adaptive scheme.

TABLE 3 -Induction motor parameters

Rated Dower Rated voltage Rated velocity

7.5 [kWl 220 [VI 1740 [rpm]

Rated torque

40 " m l

Mutual inductance Rotor inductance

0.02456 [HI 0.02277 [HI

: :

1 2;q 8.2

i
1.2

2.2

3.2

4.2

5.2

;
-2

a151
8.2

,
1.2

,
2.2 1l"PISl

,
3.2

1
5.2
I

4.2

50 I

2.2

3 . 2 r ~ r s i

4.2

I 5.2

z
20.075 8.2

1.2

2.2
1I"dIl

3.2

4.2

5.2

Figure 9: Observer-based DTC against stator resistance change (150[%]); (from top to bottom) speed, flux magnitude, torque estimation, stator resistance variation

Figure 11: Observer-based DTC, forward and reverse operation, experimental results; (from top to bottem) speed, phase current

os

15

Figure 10: Experimental set-up

Figure 12: Observer-based DTC, experimental results ; (from top to bottom) estimated speed, estimated torque

174

demagnetization effect and the application of the zerovoltage vector deteriorate the performance at the low speed operation. To overcome these problems, the advanced look-up table, in which intermediate voltage vectors are used, is proposed. Adaptive observer is employed to obtain good performance even against the machine parameter variation.

REFERENCE
1. G. Buja, D. Casadei, and G. Serra, 1998, IECON98, T50-T64 2. James N. Nash, 1997, IEEE Trans. Ind. Appli., Vol. 33, NO.2, pp. 333-341 3. CG Mei, SK Panda, JX Xu, and KW Lim, 1999, PEDS99, pp. 80-85 4. D. Alfonso, G. Gianluca, M. Ignazio, and P. Aldo, 1999, EPE99 5.H. Kuboda, K. Matsuse, and T. Nagano, 1993, m s .Ind. Appli., Vol. 29, No. 2, pp. 344-348

Figure 13: Stator resistance estimation

CONCLUSION
Advanced DTC algorithms for high power 3-level inverters are presented. The torque and flux control method for 3- level inverter is studied. The

IEEE

175