(1-1)Fully Digital, Vector-Controlled Pwm Vsi-Fed Ac Drives With An Inverter Dead-Time Compensation Strategy.pdf

552

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21, NO. 3, MAYIJUNE 1991

Fully Digital, Vector-Controlled PWM

VSI-Fed ac Drives with an Inverter

Dead-Time Compensation Strategy

Takashi Sukegawa, Member, IEEE, Keno Kamiyama, Senior Member, IEEE, Katsuhiro Mizuno,

Takayuki Matsui, Member, IEEE, and Toshiaki Okuyama

demonstrates higher performance speed and torque control

capability with a squirrel-cage induction motor, without the

brush and commutator needed for a dc motor [ 11, [2].

As power converters frequently applied to adjustable speed

ac drives, there are three kinds, i.e., the voltage-source

inverter (VSI), current-source inverter (CSI) and cyclocon-

verter. For applications below the medium,power range, the

most popular is the pulse-width modulation (PWM) VSI,

which is more competitive in terms of cost and performance

than the others. When performance characteristics with less

torque ripple are required, a fast response ac current con-

troller must be provided in a vector-controlled PWM VSI-fed

ac drive [5].

As more microprocessor-based controllers for ac drives

are being employed, desire has arisen for a fully digital type,

ac drive with higher precision and performance and better

functions. To realize such a digital controller for the ac

current loop control and its sophisticated software, high-speed

A/D converters would increase costs so that a hybrid digital

control system, in which only the ac current control loop is

implemented by an analog controller, has generally been

employed [5], [6].

Now, the authors have developed a fully digital, vector-

controlled PWM VSI-fed ac drive, which has no ac current

control loop. Performance is comparable to that obtained for

an ac drive with the ac current control loop. The new drive

provides a PWM VSI output voltage distortion correction for

both fundamental and harmonic components. This correction

is not affected by the magnitude of the inverter output voltage

or current distortions.

In this paper, the authors describe the principle and method

for the inverter dead-time compensation and results from

simulations and test.

Abstract-A dead-time compensation method in vector-con-

trolled PWM VSI’s is proposed. The method is based on a

feedforward approach that produces compensating signals ob-

tained from the Zd-Zq current and inverter output angular fre-

quency references in the rotating reference (d-q) frame. It

provides excellent inverter output voltage distortion correction

for both fundamental and harmonic components. The correc-

tion is not affected by the magnitude of the inverter output

voltage or current distortions. Since this dead-time compensa-

tion method allows current loop calculations in the d-q frame at

a slower sampling rate, with a conventional microprocessor,

than calculations in the stationary reference frame, a fully

digital, vector-controlled speed regulator with just a component

current control loop is realized for PWM VSI’s.

Simulations and test results obtained for the compensation

method are also described.

I. INTRODUCTION

N ADJUSTABLE-speed drives for advanced rolling mills,

I technology that makes it possible to produce different kinds

of thinner and higher quality strips of various materials, with

the same mill, at both high productivity and low cost is

strongly desired. Such mill drives require higher accuracy

and faster response in speed and torque control over a wide

speed range. In addition, they must provide flexibility in

performance and operations for the various materials.

From the viewpoint of maintenance, energy, and installa-

tion space savings, ac drives have been extensively applied.

This tendency is being accelerated because ac drives that

exceed the drive system performances attainable using dc

drives are available [ 11, [2]. With the evolution of microelec-

tronics and power electronics, the software-based implemen-

tation of sophisticated and precise motor control algorithms

has become possible. Excellent drive system performances

have been obtained by vector control theory [3], [4] in

combination with multivariable control theory using a digital

controller. For example, the vector-controlled ac drive

11. DEAD-TIME

COMPENSATION

A. Description of Problem

Modem control schemes of adjustable-speed ac drives

began with a V/ f control type, followed by a slip frequency

control type. Taking advantage of advances in microelectron-

ics and power electronics, the latter was successfully ex-

panded into the present slip frequency vector control type,

which combines vector control and multivariable control

theories 161. In the slip frequency vector control type, an ac

Paper IPCSD 90-49, approved by the Industrial Drives Committee of the

IEEE Industry Applications Society for presentation at the 1988 Industry

Applications Society Annual Meeting, Pittsburgh, PA, October 2-7.

Manuscript released for publication November 26, 1990.

T. Sukegawa and K. Kamiyama are with Omika Works of Hitachi, Ltd.,

Hitachi-Shi, Ibaraki-Ken, Japan.

K. Mizuno is with the Head Office of Hitachi, Ltd., Tokyo, Japan.

T. Matsui and T. Okuyama are with the Hitachi Research Laboratory of

Hitachi, Ltd., Hitachi-Shi, Ibaraki-Ken, Japan.

IEEE Log Number 9142782.

0093-9994/91/05OO-0552$01 .OO 0 1991 IEEE

553

SUKEGAWA et al.: FULLY DIGITAL VECTOR-CONTROLLED PWMVSI-FED AC DRIVES

TABLE I

INVERTER

DEAD-TIME

COMPENSATION

METHODS

Description

Block Diagram

1) The detected current signal is

tailored through the compensator to

generate a compensating signal.

2) It is summed with the voltage reference

in the stationary reference frame.

1) The detected voltage signal is

compared to the voltage reference

signal.

2) The deviation signal is summed with PWM voltage

reference in the stationary reference frame.

1) The compensating signal is tailored by

a current component reference.

COMPEN-

SATOR

2) It is summed with the voltage

id*

reference by a feed forward in the d-q

frame.

IO*

current control loop, which must be highly responsive, is

generally employed to provide better sinusoidal current

waveforms. This type is widely used in present PWM VSI-fed

ac drives.

As the use of vector-controlled ac drives and the desire for

their complete digital control has increased, a fully digital

regulator for these drives has come to be considered. Before

the fully digital regulator can be developed, however, there

are several problems to solve. The major one is that a very

fast ac current control loop makes it difficult to realize digital

current control, which is not economical, although techni-

cally possible. Therefore, research and development work

have focussed on the fully digital regulator, excluding the ac

current control loop [7].

Then, the next problem to solve is development of the

dead-time compensation method because, otherwise, an in-

verter output voltage distortion arises. Its distortion causes a

torque ripple that prevents a highly responsive speed control

system.

In PWM VSI's, there is an inherent source for the output

waveform distortion in the power stage, which is caused

mainly by the dead time [8], [9]. This is the short time

elapsing between switching one device in an inverter leg off

and switching the other one on, which ensures that both do

not conduct simultaneously. Overcoming the inverter dead

time in the slip frequency vector control type without using

the ac current control loop has been the target of much work

171, 181.

Table I shows dead-time compensation methods previously

employed in the VIF and slip frequency control schemes. In

these schemes, no ac current loop control is used. Therefore,

dead-time compensation must be made either by supplying a

compensating signal produced by the output current to the

voltage reference, which is the current feedback type, or by

supplying a compensating signal directly produced by the

deviation signal between the voltage reference and detected

voltage to the voltage reference, which is the voltage feed-

back type. Compensation for these two schemes is made in

the stationary reference frame, which then leads to the same

problem as for the slip frequency vector control type, includ-

ing the ac current control loop. To overcome this, the authors

propose the dead-time compensation method shown at the

bottom of Table I.

B. Dead-Time Compensation

Fig. 1 shows a speed control system block diagram with

the proposed dead-time compensator. This system is operated

in the same way as a conventional ac drive. Therefore, only

its different parts are explained here. Compensation is made

in the rotating reference (d-q) frame; therefore, the ac

current control loop is eliminated. The dead-time compen-

sator receives the current references 1; and 12 from the

speed controller and the field current generator, respectively.

It also receives the inverter output angular frequency w, from

the input signal to the two-phase oscillator. The dead-time

compensator produces a compensating signal, which is a

554

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21, NO. 3, MAYIJUNE 1991

where ZT and w, are the amplitude of the motor current and

inverter output angular frequency, and 0 * = tan-

( I,* /

) .

The space current vector, therefore, moves continuously as

w,t increases. Its trajectory, from (l), is a circle. The

three-phase instantaneous currents i;, i:, and i*, can be

expressed by using i*, and ig:

r1 01

DETECTOR

Fig. 1.

Fully digital speed control system with current controller using

dead-time compensator.

Based on the above-mentioned conditions, the compensat-

ing signals Vf;, Vfz, and Vf: in the stationary reference

frame, which are shown in Fig. 2, are obtained from i;, i:,

and it. Consequently, the compensating signals Vfz and V'g

in the a-0 frame can be obtained:

space vector to be added to the voltage references V: and

V,* in the d-q frame. Here, the voltage reference is con-

cerned with generating a signal for the fundamental output

voltage, whereas the compensating signal is concerned with

generating a signal mainly for the required harmonic com-

pensation. The reason for the latter is described in Section

111. Therefore, the compensating signal can be produced

corresponding to the region where the motor current may be

found in the stationary frame. The method is performed by a

feedforward in the d-q frame.

The advantages of this method include inverter output

voltage distortion correction for both fundamental and har-

monic components in the d-q frame. Thus, the method

allows control implementation in the d-q frame at a slower

sampling rate with a conventional microprocessor than when

implementation is performed in the stationary reference

frame. This helps to provide a fully digital, vector-controlled

speed regulator with just a component current control loop

for PWM VSI's.

Furthermore, between the two-phase compensating signals

Vf*, and Vf; in the a-0 frame and the proposed compensating

signals f: and f,* in the d-q frame, the following expres-

sion can be obtained:

The process demonstrating how the compensating signals

f: and f,* are produced is shown in Fig. 2. The compensat-

ing space vector moves discontinuously as w t increases, and

its trajectory is a hexagon. This is because the error voltage

depends on the polarity of the motor current but not on its

magnitude or shape. Furthermore, amplitude of the error

voltage depends linearly on the ratio td/ T,, where t, is the

dead time, and T, is the period of switching frequency.

Referring to Fig. 2, a space vector expression of the

current vector i (a circle) and the inverter dead-time compen-

sating vector (a hexagon) on the a-0 frame is made in Fig. 3.

Comparing Fig. 2 with Fig. 3(b) shows that the space vector

of the compensating signals consists of six components a to f,

which are defined in the regions A to F, where the current

space vector exists. The regions A to F in Fig. 3(b) corre-

spond to

AND EVALUATION

A. Principle of the Proposed Compensation Method

The error voltage caused by the inverter dead time con-

tains the fundamental and odd harmonics. However, the

zero-sequence voltages such as third- and ninth-order har-

monics are not present in the line-to-line voltage. This fact

suggests that compensation of the dead time can be made in

the rotating reference (d-q) frame. In other words, compen-

sation signals for the error voltage can be produced in the

d-q frame, which contains all the odd harmonics without

zero-sequence components.

In a vector-controlled ac drive, when the component cur-

rent references I: and I,* in the d-q frame are given, the

two-phase instantaneous currents i*, and ii in the two axis

(a-@ frame can be fixed as follows:

cos w, t

111. RUNCIPLE

[:;I = [sin olt

where T is the period of the inverter output frequency.

Namely, the compensating vector in each region is

- sin w1 t

cos w,r][

oii

0 5 t < -

for

COS (w,t + e*)

sin (w,f + e*)

- 5 t < -

0% for

555

SUKEGAWA et al.: FULLY DIGITAL VECTOR-CONTROLLED PWM VSI-FED AC DRIVES

Component

Current

References (d-q)

Two-Phase

Currents (a+)

Three-phase

Currents and

Compensating

Signals

Stationary

(Rererence Frame)

Compensating Vector

when space current

/vector

is found in\

I /A

/ /

'\,\region

Compensating

Signals (a-P)

- - _- _- -- 4'

(b)

Fig. 3. Relationship between space current vector and inverter dead-time

compensating vector on a-fi frame: (a) Space current vector; (b) compensat-

ing vector corresponding to space current vector.

Compensating

Signals (d-q)

where Vf = V,, t/ T, (V,, = inverter input voltage), and

e* = tan-' (Zt/Z$). In (5) and (6), Trunc (A, B) denotes

the following:

Trunc(A,B) =m, when A =mB+C and C<B

where A, B, and C are real numbers, and m is an integer.

Thus, the proposed method gives a feedforward method,

which can compensate the fundamental and harmonic compo-

nents of the error voltage in the d-q frame.

Fig. 2. Compensating signals of proposed inverter dead-time compensation

method.

02 for

- 5 t < -

02 for

- st<-

B. Evaluation

Accuracy and sensitivity of the control system imple-

mented in the proposed method are evaluated with a simula-

tion model, as shown in Fig. 4. The simulation model is

verified later by comparing computed and measured results

shown in Figs. 8 and 9. In Fig. 4, the error voltage compo-

nents due to the dead-time appear in the distortion generator,

where respective errors are calculated by the phase of the

space current vector. Next, in the dead-time compensator,

the proposed compensating signals f: and f,* in the d-q

frame can be calculated by (5) and (6). The compensating

signals are added to the voltage references from the compo-

nent current controllers I,C and IqC, respectively.

Simulations are carried out with the specification shown in

for

07 for - 5 t < T.

Accordingly, the proposed compensating signals are given by

the following equations.

- Trunc o,t + e* + - -1 - olt]

fd* = 3 Vf cos [ f

(

6'3

556

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS. VOL. 21, NO. 3, MAYIJUNE 1991

p.,

Under

Over

-Compensation

---* Compensation

SPEED Wr

IndNcoter 11s

references

Fig. 4. Simulation model of proposed control system for evaluating torque

ripples caused by inverter dead time.

k ( p U 1

Compensating

Signal

Amplitude

Fig. 6. Characteristics between torque ripple and inverter dead-time com-

pensating signal amplitude.

Roted Load

: No Load

-’

.~.

Time

With

Compensation

Oulpul Frequency f

( Hr 1

Fig. 5.

Characteristics between torque ripple caused by inverter dead time

and output frequency.

-15

-10

-5

the appendix. Fig. 5 shows characteristics between torque

ripple p caused by the dead time and inverter output fre-

quency f,, where p is defined as the ratio of torque ripple

variation to rated torque of the motor. The torque ripple due

to the dead time decreases logarithmically as the inverter

output frequency increases. This is because the ratio of the

error voltage to the inverter output voltage decreases with the

output frequency. In addition, the torque ripple at no load is

larger than that at the rated load without compensation. This

is because almost all waveform distortion at the motor cur-

rent zero-cross point affects the torque current in the d-q

frame when the motor operation is no load. On the other

hand, when the motor is running at the rated load, the

above-mentioned distortion affects the field current in the d-q

frame. However, the motor flux shows hardly any effect on

the current ripple. Therefore, the torque ripple is smaller

than that at no load.

Fig. 6 shows characteristics between the torque ripple p

due to the dead time and the compensating signal amplitude k

at 10% of the rated motor speed. The sensitivity to the

compensating signal amplitude at no load is higher than that

at the rated load. This is because the waveform distortion at

the motor current zero cross point due to under or overcom-

pensation affects the torque current at no load. On the other

hand, at the rated load, this waveform distortion affects the

field current. Therefore, the torque ripple is smaller because

the motor flux has hardly any effect on the current ripple. In

both cases, the compensating signal amplitude plays a critical

role in minimizing torque ripple in the motor.

Delay Angle 6

Compensating

Signal

(deg)

Fig. 7.

Characteristics between torque ripple and compensating signal

delay angle.

Fig. 7 shows characteristics between the torque ripple and

compensating signal delay angle 6 at 10%of the rated motor

speed. The effect of sampling delay time in the microproces-

sor on the compensation accuracy can be evaluated from this

result. The delay angle is obtained as the product of the delay

time and the inverter output frequency. The delay angle also

produces torque ripple in the motor. Thus, compensation for

the signal delay angle is required. The compensation can be

made by correction of the phase f3* of the space current

vector described in (5) and (6) [lo].

Fig. 8 shows simulation results of the inverter output

references and the output currents under the condition that

the motor is running at 10% of the rated speed at no load.

For undercompensation k = 0.5, the motor currents are

trapezoidal shaped. On the other hand, for overcompensation

k = 1.25, the motor currents are convexly shaped. The

ripple due to undesired compensation appears on the torque

and field currents in the d-q frame. The torque current ripple

is larger than the field current ripple at no load. It also

contains six times the output frequency.

Fig. 9 shows experimental results of the inverter output

voltage references and the output currents under the same

conditions as Fig. 8. From comparison of the computed and

measured results in the two figures, the computed motor

current waveform shapes accord very well with measure-

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: