Design and Implementation of a New Three-phase Four-Wire Active Power Filter with Minimum Components

A. Dastfan, D. Platt, V. J. Gosbell

School of Electrical, Computer and Telecommunication Engineering University of Wollongong, Northfields Ave., Wollongong NSW 2500, Australia
Abstract- This paper reports the development of a new threephase four-wire active power filter which has no passive elements and only has four bidirectional switches. A new switching technique based on three-dimensional vector treatment of unbalanced three phase circuits is used for the control of the APF. Experimental results from a prototype APF confirm the suitability of the proposed approach not only for the removal of harmonics, but also for the compensation of fundamental unbalance currents and reactive power consumed by the load. The experimental results show that current Total Harmonic Distortion(THD) in each phase is reduced from 58% to less than 5% which is acceptable by most harmonics standards. The neutral current, including fundamental and harmonics, is reduced by a factor of 15.

11. PROPOSED APF TOPOLOGY A SUMMARY OF ITS AND

SWITCHING STATES OF OPERATIONS

Fig. 1 shows the basic circuit of a newly developed threephase four-wire APF [7]. The four switches are all bidirectional type, which can block or conduct in both directions. These switches can be implemented as shown in Fig. 2. The nonlinear load includes single and three phase diode rectifiers with capacitor filter on the DC side. From Fig. 1, the source voltage can be written as:
d V = L, --I dt

+v

I. INTRODUCTION In three-phase, four-wire systems, significant use of electronic equipment with diode rectifier and capacitor filter front ends, such as Pc's, TV's, high efficiency lighting and air conditioning, gives a high level of harmonic currents in both the three line conductors and more significantly, in the neutral conductor [l, 21. Two major approaches to the implementation of active filtering have been proposed in the past. The first uses three single-phase Active Power Filters (APF) and the second uses one three-phase, four wire inverter [3, 41. All of these proposed APF require substantial energy storage components and in the first case it requires a high number of switches. This paper presents a new APF, not requiring any added large passive elements. The switches alone are sufficient for APF performance whenever there is sufficient upstream inductance (eg from a supply transformer) and downstream diode rectifiers with capacitance, as is the case with the most common type of nonlinear load. A new control algorithm based on three-dimensional vector control has been used to control the APF. Hardware implementation of a benchtop design of the APF is described. This example has been chosen to represent a scaled commercial building installation. Experimental results from the prototype APF for a balanced and unbalanced nonlinear load are presented to illustrate the capability of this system.

where L, is the upstream inductance which is mainly leakage inductance of the distribution transformer and:

In the three-phase four-wire system, the instantaneous a, p, and 0 transformation has been defined by Akagi et a1 [5] as follows:

Fig. 1: Proposed three-phase four-wire APF schematic

b-'

Fig. 2: Implementation of bidirectional switch

0-7803-4943-1/98/$10.000 1998 IEEE

1369

-9

V =TV

where T, the ap0 transformation matrix, is as follows:

zero when the phase is switched to the neutral, v d c when the phase current is positive and the corresponding switch is off, e -Vdc when the phase current is negative and the corresponding switch is off.
e
e

and where Therefore, v,=AV,(dc), v'bn=BVb(dc), V'cn=CVc(dc),, A, B, and C indicate the polarity of the load voltage (van,V'bn, v'~,,)and can be 0, 1, or -1. For the analysis presented here, equation (6) can be simplified by assuming that the DC voltages (Va(dc), Vb(dc)r Vc(dc)) equal. All components and are are also assumed to be ideal. Thus (6) can be re-written as:
a: n] d ; = [ (7)

In balanced three-phase systems, the trajectory of the voltage vector in the ap plane is a circle. Therefore, if the source current in clPO coordinates is forced to follow a set current which is in phase with source voltage defined in (2), the negative and zero sequence components will be removed from the source current [6]. By premultiplying both sides of (1) by the ~ $ 0 transformation matrix, T, the voltage vector can be found as follows: d [TV] = L, -[TI]+ dt Thus:
-A

us introduce the switching vector S as follows:

- A - y -

-+

* =J- (2T- C ) S B
[TV']

(4)

d i V=L,-+v dt

Furthermore it is assumed that the a, p, and 0 components are in the X, Y, and Z axis directions respectively. Therefore switching vector, s can be re-written as:
-j

-3

The voltage vector across the APF is equal to:

V,('+"/ '-V,V)2

-(

6v ' ,
2

-VIm

(v',+v'h+v'm)/Jz

where v ' ~ v't,,,, and vIcn are the instantaneous phase voltages , across the APF and have PWM waveforms. By considering the equivalent single-phase circuit of the system as shown in Fig. 3, it can be seen that these voltages can have only three values:

S =+ i (A--)+ B 2

where i , -J' , and k are unit vectors in the X, Y, and Z directions respectively. Table 1 gives the possible switching states for the four switches shown in Fig. 1. Four other
Table 1

-+

<J3 J -B+ 2
-3

+A+B k-

Jz

(9)

Tt_l! lo* io.

Fig. 3: Equivalent single phase circuit

1370

switching states, where one switch is on and three other switches are off, are equivalent to state 0 (no switching) and have not been shown in this table. The following subsections show a summary of the switching vectors for different switching states.

4y

A. Single Phase to Neutral Switching (States I , 2, and 5 )

There are 12 vectors corresponding to the single phase switching states. These vectors are of length and form two hexagons in the X-Y plane as shown in Fig. 4. The numbers in brackets show the value of components in the Z direction. Fig. 4(a) shows the switching vectors when the load voltage in two other phases have the same polarity. At any time only one vector from this group is available being the closest one to the voltage vector, v . The second group, shown in Fig. 4(b), demonstrates the switching vectors in a single phase switching when the load voltage in two other phases have opposite polarity. From this group, at any time, only the two vectors

/
-1

:
j

v
-

; I

\A2

7(=T) Jr*

__

Fig. 5: Switching vectors for phase to phase switching (Z values given in brackets)

closest to the voltage vector are available. B. Phase to Phase Switching (States 3, 6, and 8) There are six possible switching vectors of length
2

ty

corresponding to these three switching States, forming a hexagon in the X-Y plane as shown in Fig. 5 . At each moment the three switching vectors available are those closest to the voltage vector, v . By examination of equation (9) for State 0, it can be shown that this State will reduce to one of the vectors shown in Fig. 5. Therefore, State 0 is not considered in any further analysis.

-+

C. Double Phase to Neutral Switching (States 4, 7, and 9)

There are six switching vectors corresponding to a double phase to neutral switching. These vectors have the same magnitude of 3 and form a hexagon in the X-Y plane as shown in Fig. 6. At any time, the three vectors closest to the supply voltage vector are available.

&QI@
(0)

(0) .........................................

(0)

x*
_........................

2 ..............

(0)

(b) Fig. 4: Single phase 3D switching vectors diagram with Z

values given In brackets: (a) two other phase voltages have the same polanty (b) two other phase voltages have opposite

polarity

Fig. 6: Switching vectors for double phase to neutral switching (2values given in brackets)

1371

Finally for the State 10, three phases are connected to each other (or to the neutral). Therefore, A= B= C= 0 and S is equal to zero and is located at the origin. 111. CONTROLLABILITY OF THE SOURCE CURRENT Control of the three-phase source currents can be achieved when the current vector follows the supply voltage vector, which in a balanced system is a circle in a@ plane. In order to have a full control, the current vector should be able to move in any direction which may be done by selecting a combination of the available switching vectors. From (5), the current vector changes, df , is:
--f + - + dt d i = ( v - S Vdc)L,

rectifier should be at least equal to the maximum of the lineto-line voltage (eg 586 V in Australia). METHODOLOGY IV. CONTROL The function of the controller is to force the current vector to follow, as closely as possible, the reference current vector. The proposed control algorithm is based on a switching vector selection procedure. After sampling three source currents and by using the as0 transformation, the current vector ( i ) can be calculated. In order to find the voltage vector ( v ), it is assumed that the load is supplied by a three-phase balanced sinusoidal voltage. Thus:
V,"

+ +

+ Vb" + v,,

=0

( 1 1)

where Vdc, dt and L, are average dc voltage, sampling time and supply inductance respectively. Fig. 7 shows an example of ten possible current change vectors, d i , where voltage vector is in the Region 3. These vectors are calculated from (lo), and by applying ten possible switching vectors which have been explained in the previous section. The direction of the d t in the zero axis are given in the brackets. It is clear from this figure that d 1 can be constructed in all directions by selecting a proper combination of switching vectors. Therefore full control of the source current vector is achievable. However, if the DC voltage in nonlinear load is less than the maximum phase voltage, in some part of the cycle d i cannot be constructed in all directions. Thus the condition for full controllability of the source current vector is that the DC voltage of the single-phase diode rectifiers, should be at leat equal to the maximum of the phase voltage (eg 340 V in Australia), and the DC voltage of the three phase diode
--f

By measuring two phase voltages and using (2) the source voltage vector components can be found:

T,

"&

--(va"

+Vb"

+v,,)=O.

The basic structure of the three-phase four-wire APF controller is shown in Fig. 8. The DC voltage on the diode rectifier loads is controlled by the amplitude of reference current vector, i' shown in Fig. 9.

-+

. Detail

of the DC voltage controller is

A traditional integrator controller is used in order to control the DC voltage. The DC voltage is kept constant by adjusting the amplitude of the reference current vector ( i* ). In order to

APF

"C d

Fig. 8: The basic structure of the controller

Fig. 7: Possible current vector changes ( d 1 ) for 10 available switching vectors ( Z direction shown in brackets)

-?

Fig. 9: The DC voltage controller details

1372

have unity power factor, and with the assumption that the source voltages are balanced and sinusoidal, the reference current vector is:

A. Power Circuit

T = p+ ? v 1

(13)

where is Once each -+ Of the DC voltage controller. The desired switching vector ( s ,) can be , calculated from equations (5) as follows:
-+

The power circuit is composed of a three phase four wire power source, three series inductances, the nonlinear load and the APF circuit. The nonlinear load is a combination of single and three phase diode rectifiers with capacitors on the DC side. A 220 Q resistor bank is connected as a DC load for each rectifier, single and three-phase. 1100 pF electrolytic capacitors have been used which limit the maximum ripple of voltage on the DC side to less than 5%. The power source has an impedance dominated by its inductive components. The greatest contribution comes from the distribution transformers leakage inductance which is typically 510% in per unit system for a 50 Hz power system. For the prototype circuit the following parameters have been used:
1 pu VA = 2500 VA

s,J= ( v - L , - ) I V ,
dt

d i

-+

(14)

-+. where Vdc is the measured DC voltage and d i is a small change in the current vector which is desired to be equal to:

(15)
where i is the measured source current vector. The next step is to choose the available switching vector closest to the desired switching vector, S d . Then this switching vector is applied for the time interval dt, after which the process is repeated. The operational principle of the proposed threedimensional vector control scheme is illustrated by the block diagram shown in Fig. 10. V. HARDWARE DESIGN In this section a detailed design of a 2.5 kW laboratory prototype of the proposed three-phase four wire APF is presented. The load is a combination of single and three phase nonlinear loads.
-+

-+

1 pu voltage = 415 V (line to line voltage)

1 pu current = 3.48 A

1 pu impedance = 69 R
1 pu inductance = 220 mH

Due to some software and hardware limitations, the maximum switching frequency which could be achieved for the present set-up is 7.5 kHz. To limit the source current THD to approximately 5% and by referring to [8], the value of the source inductance should be at least 0.125 pu. Therefore the value of the line inductance should be at least: Ls = 12.5% of 220 mH = 27.5 mH Three 30 mH inductances have been used in this experimental set-up.

1
APF
v

1 I
Nonlinear Load
1

B. DSP controller and Associated Hardware

To achieve real time operation of the proposed APF, a Digital Signal Processor (DSP) board has been used. Three phase source currents, two source voltages and a DC voltage are the six analogue signals which are fed into the DSP board.

t l J
current vector
Calculation
Switching State

Voltage Vector Calculation

-3

~,v d c

VI. EXPERIMENTAL RESULTS This section presents the experimental results of the proposed three-phase four-wire APF with different load combinations. All tests have been done with the switching frequency of 7.5 kHz and DC voltage reference of 330 V. In each subsection, experimental and simulation results are compared.

Optimum Cvrrent Vector Setpoint Calculation

A

Fig. 10: Proposed 3-D vector control scheme

1373

A. Balanced Nonlinear Load In this part a load which is a combination of a three-phase and three single-phase diode rectifiers is considered as a balanced nonlinear load. Fig. 11 shows the phase and neutral current waveforms before and after using the APF. Fig. 12 shows the harmonic spectrum of the 'a' phase current before and after using the APF. The average THCD for the phase currents before using the APF is 63% for the first 50 harmonics, while after using the APF it has been reduced to 4.75%. The RMS neutral current is also reduced from 6.2 A to 0.4 A after using the APF which means a reduction by a factor of 15.5. These values confirm the great reduction of phase and neutral current harmonics due to the use of the APF. Although it was expected that the fundamental component of the phase current would reduce as a result of power factor correction, it is slightly increased after using the APF. The difference is mainly due to the increase of the DC voltage and also to losses related to the APF switches and power diodes. Fig. 13 shows source voltages and currents in a and p coordinates which are drawn by using data downloaded by means of the DSP. From these figures it is clear that the source voltages are not purely sinusoidal. The current vector
PM3384. FLUKE 8 PHIUPS

.-

4LL--,
*T
0 2

ST

(b)

Fig. 12: Harmonic spectrums of 'a' phase current (a) before and (b) after using the APF

i s also in phase with the voltage vector after using the APF, showing that unity power factor has been achieved.

The average measured DC voltages of the single phase diode rectifiers before using the APF was 318 V and for the three-phase diode rectifiers it was 548 V. The APF has increased these DC voltages have been increased to 330 V for the single phase diode rectifiers and 570 V for the three-phase diode rectifier.

B. Unbalanced Nonlinear Load

For this part, the 'b' phase diode rectifier has been

001

0.02 Tim 1Sec.l

00 .3

(a)

001

0 02 Time tSn )

003

(b)

(b) Fig. 13: a and p components of the source voltage and current vectors (a) before and (b) after using the APF

Fig. 1 1: Phase and neutral currents for balanced load (a) before and (b) after using the APF (10 A/div & 5 ms/dlv)

1374

disconnected to make the nonlinear load unbalanced. Fig. 14 (a) presents the source phase and neutral currents and Fig. 15 (a) is the frequency spectrum of these currents. As can be seen from this figure the neutral current consists of the fundamental and triplen harmonics, with the 31d being dominant.
Fig. 14(b) shows the APF performance for this unbalanced load where three phase currents are well balanced and the neutral current has been significantly reduced which can be seen in Fig. 15(b). The average current THD of the three phase currents has been reduced from 71% before using the APF to 5.4% after using the APF. The neutral current RMS value is also reduced from 4.9 A to 250 mA which is a reduction factor of about 20 times. Due to the boost action of this circuit and the voltage set point being 330 V, the average DC voltage has been increased from 320 to 330 V which is about a 3% increases. This results in a 7% increase in output power, since the load is resistive.

..:L
~ ~

0 0
500 1000
1500

2000

2500

e k " .
0

."L;
0
500
I

1000

1500

2000

2500

0 0
500 1000

IS00

2000

2500

*T

The average fundamental current for the three phases after using the APF is slightly higher than the average fundamental currents before compensation. This is consistent with an increase in the real power drawn to account for the higher output power and also some losses in the APF.

so0

1000

1500

2000

2500

Frequency (Hz)

(a)

*T

500

1000

1500

2000

2500

0 0 2
500

1000

1500

2000

2500

8 1

.q,0
500
--; 0
0

1000

1500

2000

2500

- ;

-~

' I

500

1000

1500

2000

2500

F r e q u e n c y (Hz)

(b) Fig. 15: Harmonic spectrums of the phase and neutral currents in unbalanced load condition (a) before and (b) after using the APF

(b)
Fig. 14: Three phase and neutral currents for unbalanced load condition (a) before and (b) after using the APF (10 Ndiv & 5
msldiv)

VII. CONCLUSIONS
The work presented in this chapter describes the implementation of a 2.5 kVA, 4 15V three-phase four-wire

1375

APF controlled by three-dimensional control methodology. Hardware requirement of the benchtop design are explained and the experimental results of the proposed APF for different nonlinear load conditions are presented. The results show that the proposed APF and its control system is able to reduce the harmonic contents of the source current and reduce the current THD to around 5% which meets IEEE-5 19 standard requirement. The experimental results for unbalanced nonlinear load also showed that this APF can compensate the unbalanced current and reduce the neutral current by a factor of 20. It is also shown that the APF can achieve unity power factor. VIII. REFERENCES Gruzs T. M., A Survey of Neutral Current in ThreePhase Power Systems, IEEE Transaction on Industry Applications, Vol. 26, No. 4, 1990, pp. 719-725. Liew A., Excessive Neutral Current in Three-phase Fluorescent Lighting Circuits, IEEE Transactions on Industry Applications, Vol. 25, No. 4, pp. 776-782, 1989. Quinn C. A., Mohan N., Mehta H., A Four-Wire, Current-Controlled Converter Provides Harmonic Neutralisation in Three-phase, Four-Wire Systems, IEEE APEC93, pp. 841-846, 1993.

[4] Sutanto D., Bou-Rabee M., Tam K. S., Chang C. S.,

Harmonic Filters for Industrial Power Systems, IEE International Conference on advances power system control, operation and management, Hong Kong, pp. 594-598, 1991.
151 Akagi H., Kanazawa Y., Nabae A., Instantaneous Reactive Power Compensator Comprising Switching Devices Without Energy Storage Components, IEEE Transaction on Industry Applications, Vol. IA-20, No. 3, 1984, pp. 625-630.

[6] Watanabe E. H., and Stephan R. M., New Concept of Instantaneous Active and Reactive Power in Electrical Systems with Generic Loads, IEEE Transactions On Power Delivery, Vol. 8, No. 2, pp. 697-703, 1993. [71 Dastfan A., Platt D., V. J. Gosbell, Control of ThreePhase, Four-Wire Active Power Filter Using ThreeDimensional Vector Control, in Proc. of Australian Universities Power Engineering Conference (AUPEC96), Melbourne, Australia, p. 223-228, 1996.

P I

Dastfan A., Active power filter with minimum components, PhD Thesis, University of Wollongong, 1998.

1376