2004 11th International Conference on Harmonies and Quality of Power

Review of Harmonic Current Extraction

Techniques for an Active Power Filter
A. M. Massoud, S . J. Finney, and B. W. Williams

Abslrad-- Since the early SOS, many harmonic extraction and guides for researchers and engineers in the field of active
techniques have been published in the literature hut there is a filters. In this paper, harmonic current extraction techniques
lack of publications rhich review harmonic current extraction (time-based and frequency-based) are classified, giving a
techniques for researchers and engiueen in the field of active
description for each technique and classification under either
filters. In this paper, classification of harmonic extraction
techniques (time-based and frequency-based) is presented, giving category, along with the reference. Fig. 1 Presents the
a brief explanation of the techniques classified under each Classification of harmonic Current eXtraCtiOII techniques.
uttegory and their reference source.
n. HARMONIC EXTRACTION TECHNIQUES
Index Terms-Active filter, power quality, harmonics
A. Instantaneous reactive power theor).
I. INTRODUCTION
UE to rapid power electronics developments, non-linear A. I 3-phase 3-wire
Dloads in many applications which inject In 1983, Nabae et al. [11,[21 proposed instantaneous reactive
harmonics into the power system. High levels of harmonics Power theory based on the instantaneous value concept. The
lead to distribution system To these conventional instantaneous active power for a three-phase
problems there are two solutions, either use a converter with a-P
circuit can be defined in as:
unity power factor sinusoidal input current hence avoiding the p = e , . i , +ep.is (1)
harmonics introduced by the converter itself, or by using a where p is equal to the conventional equation. The
filter (passive or active). Advantages of passive filters are easy instantaneous reactive power is defined as:
maintenance, less expensive, and less complexity. But they q = e , xia + e p x i , (2)
present several disadvantages; parallel and series ,."er F~~a non-linear load, if the reactive and harmonic
aging of passive components, and dependency on the source producing currents are to be compensated, the compensating
impedance. Active filters are used to overcome these current can be expressed as:
-
disadvantages. There are two types of active filters; shunt and
series active filters. The shunt active filter is the most
commonly used. This type of fiter depends on the injection of
anti-phase but equal currents to the harmonic components, to
c:;]=(2p I: ~1

(1')
where 17 is the AC conlponent of the instantaneous active
(3)

produce cancellation. Since the early XOs, many harmonic power.

current extraction techniques have been published but there is A.2 3-phase 4-wire
a lack of reviews on harmonic current extraction techniques The instantanews reactive power theory can be extended to

rvcrgc

Fig. I . Classification of harmonic cumnt extnction techniques

three-phase circuits which include zero-phase sequence
The authorsare with Depanmem ofcomputing andEIecuical Engineering components [1],[2]. A new term is introduced which is the
Heriot-Wm UrUuer~ity,Riccmon Campus, Edinhurgh, UK.EH14 4.4s.

0-7803-8746-5/04/$20.00 02004 IEEE. 154

instantaneous zerephase sequence power po and is defined as: are the compensating currents
p0 = %.io (4) C. Rotating p-q-r theory
Now the three independent quantities po , p , andq can be Three power components are defined as linearly independent
exnressed as: on the p-q-r reference frames [SI. Thus, the three current
components can be controlled independently by compensating
for the three instantaneous power components on the p-q-r
reference frames. As shown in Fig. 2a, a new a'-p\-O reference
frame is established by rotating the &axis, of the a-p-0
Since the zero-sequence current must be compensated, the reference frame by 81, aligning the a-axis with the projection
reference compensation current in the zero-coordinate is io line of the voltage space vector to the a-p plane. The current
itself. space vector on the a\-p'-O reference frames can be described
A.3 Nonsymmetrical volfages in p-q theory as:
If the system voltage is nonsymmetrical [5]. The new
instantaneous reactive power q is defined by: (10)
# .
q=e:i,+eLi,+e,i, (6)
where e: , e;, and e: have 90' (or -904 delay with respect to
e,, eb and e , respectively.
'
B. Cross vector theury
Next, as shown in Fig. Zb, the p-q-r reference frames can be
The cross vector theory in [3] [4] defines an instantaneous formed by rotating the p\-axis of the a\-p\-Oreference frame by
real power p and three instantaneous imaginary powers. The e2,aligning the a\-axis with the voltage space vector. The
instantaneous reactive power theory can be defined in the a-b-c current space vector on the p-q-r reference frames is described
reference frame as: as:

And it can be defined in the a-p-G reference frame as: and e,# =m (13)

The instantaneous activelreactive powers can be defined

respectively by the scalarlvector products and the current space
vectors.
- . -
Similarities and differences between the "3-phase 4-wire" p-q P = eppr. I,, - epip (14)
theory and the cross vector theory can be summarized as ~ = z ~ , , x F ~=[o,-e,i,,e,i,]'
~, (13
follows: The first considers the zero-sequence circuit as a P-ZUk
single-phase circuit independent of the a-phase and P-phase P'W.
W-arkl
circuits and deals with the zero-sequence current as an
instantaneous active current only. This means that no
instantaneous reactive current exists in the zero-sequence
circuit. For the second arrangement the zero-sequence, a-
phase, and P-phase circuits are all considered. This implies
that the zero-sequence current can be divided into zero-
sequence instantaneous active and reactive currents.
Assuming that the active filter is ideal, the neutral is isolated,
and the voltages are balanced, so the active filter current can be
Fans
I
expressed as [6],[7]:

(9)
(q-ZUUl
(b)
Fig. 2. Physical meaning of p-q-r reference frames (a) relation between a-PO
references frame and a' - $ 4refernce bames (View f"the top of the 0-axis)
(b) relation between a\ -$'a reference framer and w-r reference framer
where i k , i , , and i k are the load currents and ,i , iCa , and i , (view bom the b n o m of the paris)

155
Note that p-q theory does not observe power conservation since reference input and the load current is used as the primary
the instantaneous zero sequence reactive power cannot be input, that is, the fundamental component acts as the noise
defined hy the theory. While cross vector theory defines one while harmonics act as the signal.
instantaneous real power and three instantaneous imaginary Rlwiinplt
lfu~menW+hm~~~ ii0
powers that observe power conservation, the three
instantaneous imaginary powers are linearly dependent of each
other which means they cannot he compensated separately.
The p-q-r theory [9],[10] takes advantage of both p-q theory =j
Rlfcnncrri8hl
shlplq tircuz, I I I
and cross vector theory. The defmed instantaneous powers D%ln(mt)
(r""dunnal)
observe power conservation. Both instantaneous real and
Fig. 3 . Adaptive noise cancelling as a harmonic extraction technique
imaginary powers can he defined in the zero-sequence circuit
in three-phase four-wire systems. The three power components F. Capacitor voltage contrul
are linearly independent of each other. This technique depends on regulating the dc-bus voltage of the
D. Synchronous reference frame power converter [ 151. Harmonic extraction is performed using
capacitor voltage control as in [16],[17]. This technique is
D.1 lntroducrion based on power balance where the supply real power must be
A synchronous reference frame [11],[12] was suggested by
equal to the load real power plus the inverter losses. To
Divan. The load currents are transformed from a-b-c stationary maintain power balance, the capacitor must instantaneously
reference frame to a d-q synchronously rotating reference compensate for the difference between the supply and the load
frame as power (Fig. 4a). So controlling the capacitor voltage using a PI
controller results in an output which is proponional to
instantaneous power balance changes. Multiplying this output
hy the per unit voltages of the common coupling point results
in the reference supply current. With this concept, the control
circuit can be significantly simplified. For obtaining a balanced
where e is the angle of rotation of the d-q coordinates and is current under non-ideal mains voltages, only one phase of the
equal to wt, where w is the power supply frequency. The d- mains voltage is used as the phase reference to calculate the
current and q-current components represent the active and desired mains current, as shown in Fig. 4h. For an unbalanced
reactive power components of the current respectively. The supply voltage, the per unit voltage of the supply voltage is
currents can be decomposed into: calculated with resoect to a common base 1311 defined as
- I

id = i d + id (17)
- + -iq
i, = i, (18)
where Td and-
- -
the fundaniental active and reactive
current components, while id and iq are the harmonics active
and reactive current components.
D.2 Filrerirzg and Moving average process
A low pass filter is used to extract the DC components while
the moving average process is another method [13]. In [13] it
is assumed that even harmonics are not presented in the load
currents. The filters are substituted by blocks that calculate the
moving average
6 ' (b)
= - Jiddt Fig. 4. Capacitor volWge ~ontroltechnique
T ,.f
This procedure can be performed because every cdd order G. Correlation function technique
harmonic component is a multiple of six after transformation Enslin proposed division of the apparent power into two
to the d-q reference frame. So every transformed cdd order orthogonal components namely real power and fictitious power
harmonic has a zero average value over 116 of the fundamental [IX]. Fictitious power is also sub-divided into two orthogonal
period. components: reactive and deactive power using correlation
techniques. Enslin uses auto-correlation to calculate the r.m.s.
E. Aduptise interference canceling technique
values of the measured currents and voltages over the period T.
The adaptive interference canceling technique [I41 keeps Cross-correlation between the current and voltage over the
the system in the best operating state by continuously self- period dT is used to calculate the active power. The auto-
tuning. The fundamental components in the load current and correlation is defined as:
the AC source voltage are mutually correlated. In the following
detecting system in Fig. 3, the AC source voltage is used as the

156
From @ I ) , the voltage r.m.s. is defined as: J. Synchronous detection method
(22) In this method, for compensating component calculation,
v = 07, (0));
The cross-correlatjon is: there are three approaches [221: equal power synchronous
detection (PSD), equal current synchronous detection (CSD),
R,,,(T)=T I lv(r)i(t-r)dr
n (23)
and equal resistance synchronous detection (RSD).
0
From (23), the active power can be defined as: J.1 Equal power approach
P=Rv(0) (24) The three-phase instantaneous active power is defined as
So the active current component is: p = EmIm + Ed[* + E,I, (28)
i , ( t ) = TP. v ( r ) (25) For equal power synchronous detection
V EmIm = E,I, = E,I, (29)
The line currents after compensation are expressed as:
H. Filtering the load current
In this method, the distorted load current is filtered,
extracting the fundamental current component [19]. Thus the
filter output current is equal to the fundamental component of
the load current. In order to provide the active power required
by the load, the current signal obtained from the filter is (30)
synchronized with the respective phase to neutral source
voltage. This technique [20] suffers from the phase and
magnitude errors introduced by the signal active filter. Two
main categories emerge. Using the high-pass-filter method is where p is the real total power of the compensated three-phase
straightforward for removing low-order frequencies in the power system, e&) is the voltage source of phase a, E,, is peak
load-current signal. Because of differentiation, this technique is amplitude of the 'a' phase voltage, i,(i) is the compensated line
vulnerable to noise. Using the low-pass-filter method is current and so on for the other two phases. Based on the equal
preferred over the high-pass-filter method because it reduces power approach, the active power after compensation in each
the effect of differentiation in the resulting filtered component. phase is the same. The resulting line currents are different,
As mentioned above, the system suffers from magnitude and depending on each corresponding phase voltage.
phase errors. J.2 Equal ciirreiif approach
For equal current synchronous detection, CSD,
I. Sinu.roidal tracking rriodel
I , =I* = I , (31)
In 1211, the detected load current is applied to the sinusoidal
tracking model without any transformation. The output of this The line currents after conmensation are
model is used as the fundamental component of source current.
The resonant frequency is set to the fundamental frequency of
the source voltage. Fig. 5 shows an equivalent circuit of the
sinusoidal tracking model. It is assumed that 0, = = g.
The state equation of this circuit is

The equal current approach compensates the load current to

obtain equal line currents. The resulting current waveforms are
sinusoidal, equal, and in the correct phase sequence.
5.3 Equal resistance approach
For wual resistance svnchronous detection. RSD.

(33)
The line current after conmensation are exmessed as:

[r]
Fig. 5 . Equivalent circuit ofsinusid4 tracking madel
The output equation is

y = k , (- sin 8, cos 8, (27)

(34)
~,
where W,and k, are the resonant frequency and the control ir(r) = b)
2.17.e~
gain of sinusoidal tracking model, and 0, is the phase angle of (E; + E : ~+ E A ]
sinusoidal tracking model at the resonant frequency.

157
The equal resistance approach assumes that the load resistance N. Fasi Fourier TraIhfomI ( F I T ) and Discrete FT (DFT)
will be equal after compensation. The line currents vary There are basic assumptions for the application of DFT and
depending on the phase voltages. FFT 12x1. These assumptions are: (i) the sampling frequency is
K. Darrieus definition of reactive power greater than twice the highest frequency in the signal to be
analysed and (ii) each frequency in the signal is an integer
In 1970 Darrieus extended the concept of reactive power to
multiple of the fundamental frequency.
non-sinusoidal cases [23]. Emile Pillet puts the definition of
The FIT approach is applied if (1) the number of samples is an
Darrieus in the form:
integer power of two over an integer number of cycles and ( 2 )
the waveform is stationary. When the number of samples does
not satisfy the basic assumption in the DFTFFT algorithm, a
Therefore the three-phase reactive power can be defined as Kalnian filter based algorithm is applied. The fast Fourier
transform (FIT) was developed to reduce the computational

I
q = o i , J e , d r + i, J e , d r + i, l e , d r
In ~ $ 0 coordinates using an invariant power form, the
I burden of the D R .
0. Recursive Discrete Fourier Truiisfonn (RDFT)
instantaneous reactive nower is given bv: Traditional hamionic analysis approaches such as DFT
(Discrete Fourier Transfomi) and FFT mast Fourier
Transform) algorithms are steady stated concepts and have a
slow response because of the time delay caused by the
L Neural neiwork computational complexity [29]. However, the Recursive
A method to generate the current reference for shunt active Discrete Fourier Transform (RDFT), presented in [30], can
renew the spectrum values immediately after input of new
power filters is presented in 1241 which uses a neural network
to extract the fundamental sinusoid fiom a distorted load datum. Thus the RDFT is more suitable for real-time
implementation than DFT and FIT.
current waveform. At the same time, if the supply voltages are
unbalanced or distorted, the input of the neural network is not P. Ka/inan Filter
the supply voltage, but a unity vector of the fundamental supply The Kalman filter is a recursive optimal estimator, well
voltage. Thus the active power filter attains good performance suited for on-line applications. It requires a state variable
even with distorted or unbalanced source voltages conditions. model for the parameters to be estimated and a measurement
In [ZS], an artificial neural network (ANN) with a large equation that relates the discrete measurement to the state
number of inputs and neurons is used, where the harmonic variables (parameters). The Kalman filter uses a mathematical
estimation depends on training to get the weights of different model of the states to be estimated and a Kalman filter
neurons. In [26], an adaptive neural network (ADNN) is used recursive estimation algorithm is presented in [28]. If the
to determine adaptively the fundamental and harmonic harmonic has a time-varying magnitude, the Kalman filter
component, instead of training the neurons. algorithm tracks the time variation after the initialisation
M . Ideniificafion by Fourier series period (half a cycle).
A periodic current with a zero dc-value i(i) of any form can Q. Wavelet buses
be represented [27] by: This method is based on the definition of the active and
i ( t ) = %[a, cos(nw,i) + b. sin(no,i)]
(38) reactive power in the time-frequency domain using the
"=I complex wavelet transform. The voltage and current are
where w, is the fundamental angular velocity. With a transformed to the time-frequency domain using the complex
numerical implementation of a moving Fourier series, the wavelet with scaling and translation parameters to set the
coefficients become: kequency range and localize the frequency respectively [32].
The sub-band of interest is the one that cover the fundamental.
Similarly as in Fourier, the active power is defined using the
complex wavelet.
(40)
111. CONCLUSION
where T, is the sampling period and N an integer. Or A review of the most significant harmonic current extraction
recursively: techniques has been presented. This classification may help
a , , ( k ) = o , ( k - l ) + - -2P i k T , ) ti(+ N )T,)]cos(no,kT,) (41) researchers and engineers to find suitable techniques for their
N applications in the active power filter area.

1V. REFERENCES
with N = T i I(ZT,) . The identification response-time [ I ] H. Akagi, Y. Kanarau'a, and A. Nabae, "Generalired theory of the
corresponds to half the fundamental period. In this way the insMntaneous reactive power in three-phase circuils." IPEC'83-Inl. Pow-
ElecuonicsConf..Tokyo. Japan 1983. p.1375-I386
individual current harmonics can be identified.
121 H. Aka& Y. ffinawua. and A. Nabae, "lnsmmnm~sreactive power

158
 compensators comprising swirching devices without energy stordge I211 Park, J.H.: Shin. D.R.: Roh. T.K: Ah% I.M.: Lee, H.W.; Woo, 1.1.:
componcns." IEEE Tram. Ind. Appl., Vo1.20 , p.625630.1984 "Deadbeat control of three phase voltage source active power filter using
131 Fang Zheng Peng: JihSheng h i : "Generalized instantaneous reactive power sinusoidal vacking model",TENCON 99. Proceedings of the I E E Region
theory for three~phase power systems", Lnsmmentation and Measurement, 10Conference,Dec 1999, vo1.2,pp: 1442-1445
IEEETransactianson,Vol.45.Issue:I .feh. 1996 .pp:293-297 1221 C. L. Lin, C. L. Chen C. L Huang: "Calculating approach and
141 Fang Zheng Peng: On.G.W., Jr.: Adams, D.J.: " H m o n i c and reactive power implementation far active filfers in unbalanced three-phase system using
compensation bawd on the generalized instantaneous reactive power theory synchronous detection method.., Industrial Elecuonics. Control.
for three-phase four-wire systems", Power Elecuonics IEEE Transactions on Insuumentatian. and Automation, 'Power Elecuoonics and Motion Control'.,
,Vol. 13, Issue: 6 .NOV.1998, pp: 1174- 1181 Pmceedings of UV 1992 Lntemaxional Conference on, Vol.1, pp: 374 -380
[SI Komatsu, Y.; Kawabata. T.?A conw-al method of anive power filter where 1231 G. U. Marques, "Detection of reference currents of active filters using the
system voltage conuins negative-phasesequence component 01 m e p h a s e - Danieus defmitian of rednive power", Industrial Wenronio, 2WO. N E
sequence component". Power Electmnics and Drive Systems, 1995. Vol. 2, 2WO.ProceedingsofUle2000IEEE,Vol.I,pp:73-77
pp:583 -586 [?41 Dswei Gao: Qingchun Lu; Xiaorui Sun: "Design and performince of an
161 T. Furuhzhi, S. O k u m , and Y. Uchkawa."A study on the theory of active power filter for unbalanced loads", Power System Technology. 2002.
instantaneous reactive power". Indun. Ellen. Trans., IEEE, 1990. Vol. 37. pp. htemationa! Conference O n . V01.4, pp: 2496 -25W
86-90 1251 S. D.Round, N. Mohan,"Comparison of frequency and time domain neural
171 M. K. Mishra. A. Jorhi, A. Ghosh "A new algorithm far active shunt filters network controllers for an active power filter". pmceedings of IEEE Ind
using inctanmeous reactive power theory", IEEE Power Engineering Review, Elect., Control and Inarumentation, vol. 2, 1993, pp. 1099-1104
2000, pp: 56-58 1261 Rukanuuaman M.: Nishida, K.; Nakaoka M.7 DSP mnud shunt APF
181 H. Kim, H. Akagi, "The instantaneous power theory on the rotating pq-r with harmonic extraction by adaptive neural network, Industry Applications
reference tames", IEEE 1999 lntemational Conference on Power Electronics Conference, 2003. 38th IAS Annual Meeting. Conference Record of the ~

and Drive System, PED'99, July 1999, H K. p p 4 2 2 4 2 7 Volume:2, 12-16Oc~2003,Pages:1215-12?1"01.2

191 Kim. H.. Blaabjerg, F., Bak-Jensen. B., Chai. J.; "Instantaneous Power 1271 Abrahamxen, f.: David A.; "Adjustable speed drive with active filtering
Compensation in Three-Phas Syaems by Using p q - r Theory", IEEE capability for harmonic current compensation", Power Elenronics SpeciaPss
TranwctionsonPowerElecuonics,Vol. 17, No.5,2W?,Qp:701-710. Conference, 1995. 26th Annual IEEE. V01.2. 18-22 June 1995. pp:1137 -
[IO] Kim, H., Blaabjerg, f.. Bak-lensen. E., Choi, J.: "Specml Analysis of 1143
lnstmtaneous Powers in S i n g l e - F h s and Three-phase Systems With Use of 1281 Girgis, A.A.: Chang, W.B.: Makrdm E.B.; "A digital recursive mezuremenl
p q - r Theory", IEEE Transanions on Power Electmnics, Vol. 17, NO. 5, scheme for online tracking of pawer synem harmonics". Power Delivery,
2002, pp:711-720. IEEETransimionson, Vol. 6 , July 1991. pp:I I S 3 -1 160
[II ] M. Takeda K. keda, A. Teramota and T. Anauka, ''Harmonic current and 1291 Rechka S.: Ngandui, T.: limhong Xu: Sicard. P.: "A mmparatix smdy of
reactive power compensation with an active filter": PESC 88, April 1988, harmonic detection algorithms for active fdters and hyhtid active filters",
p.1174-I 179 Power Elenranis Specialists Conference. 2002. p e s 02. 2002 IEEE 33rd
1121 V. Scares, P. Verdelho, G. D. Marques'% instantaneous activeand reactive Annual , Vol. I , pp: 357 -363
current mahod for actiw Glters", IEEE Transactions on Power Electronics, 1301 M. Dole%R.D.Larem, "An industrially useful means for decomposition and
Vol. 15, No. 4, ?OOO, pp: 660-669 differentiation of harmonic wmponents of periodic waveforms", 2000, IEEE.
[131deLacerdadeOliveim,L.E.:daSilva.L.E.B.:duSPva,V.F.:Torr~s.G.L.; PP1016-IO23
Pinto, J.O.P.: "Improving the dynafic response of active power filters bwcd 1311 B. Singh, Anuradha, U. P. Kothari, "Analysis of a navel active filter In
on the synchronous reference f n m e method... Applied Power Eleclronics balancing and reactive power compensation", Power Electronics and Varaible
Conference and Exposition, 2002. Seventeenth Annual IEEE. Vol. 2 . pp: 742 Speed Drives, 1998. Conference h h l . No. 456, IEE, p p 5 7 6 2
-148 [32] Driesen, 1.: Belmans, R.?Actiw power filter cantml algorithm using
1141 L. Shigua: H. Zhengcheng: L. 2hixiang:"An adaptive detecting method for wavelet-based power definitions", Harmonics and Quality of Power. 10th
harmonic and reactive currents", TENCON '93. Proceedings. Computer, lntematioml Conference o n , Vol. 2 ,6-9 Oct. 2002. pp: 466 - 471
Communication. Canvol and Power Engineering.1993 IEEE. vol.5, Oct
1993, pp. 334-338 V. BIOGRAPHIES
1151 Shyh-lier Huang: JinnChang Wu: "A conml algorithm for three-phase
three-wlred active power filters under non-ideal mains voltages". Power Ahmed M. Massoud received an BSc. degree (fmt class honours) and M.Sc.
ElectronicsTrans.. IEEE. V01.14. l u l 1999. pp: 753 -760 from the faculry of Endneekg. Alexandria University, Egypt in 1997 and 2000
1161 T. F ~ ~ h a ~S. h iOkuma.
, and Y. Uchikawa,"A audg on the theory of and is currently pursuing a PkD. degree in Elecuical Engineering in the
Computing and Electrical Depanmant, Heriot-Watt Univerbity, Edinburgh, UK
instantaneous reactive power", Indus. Elen. Trans., IEEE, 1990. Vol. 37, pp.
His research interests include power quality. active power filtering. and multilevel
86-90
conveners.
1171 B. Singh, Anuradha. D.P. Kothsri, "Analysis of a novel active filter for
Stephen J. Flnney received an M.Eng. degree from L o u g h h o u g h University of
balancing and reactive power compensation", Power Electronics and Varaihle
Technology, Loughborough, U.K., in 1988, and a PhD. degree from Heriot-Wau
SpeedDrives, 1998. Conferencehbl. No. 456. IEE, p p 5 7 4 2 University, Edinburgh, U.K., in 1995. For two years he %'as with the Electricity
1181 Plltorino. L.A.; H o m A.: Enslin. J.H.R.:"Power theory evaluation far the Council Research Cenue Pdhorataier near Chester, U.K. He is currently a Lecturer
control of an active power filter", AFRICON, 1996.. IEEE AfRlCON 4 t h . at Heriot-Watt University. His areas of interea are Sofl-swilChmg techniques.
Vol.2,24-27Sep1996.pp:676-681 power semiconductor p m t e d o ~energy recowry snubber circuits and low-
1191 Mom". L.: Dim, M.:Higucra, V.: Wallace. R.: Dixon, J.?A three-phase distortion rectifier topologier.
active power filter operating with fined switching k q u e n c y for reactive power BarryW.Williamn received an M.Eng.Sc. degree from the Unirerdty of Adelaide.
and current harmonic compensation", Industrial Electronics. Conuol, Adelaide, Australia, in 1978, and Ph.D. degree from Cambridge Univerdty,
Instrumentation, and Automation, 1992. 'Power Electronics and Motion Cambridge, U.K., in 1980. After seven y w s as B Lecturer at Imperial College,
Conuol'.. Proceedings of the 1992 lnremvtional Conference on. V01.l. pp. University of landon, U.K., he was appointed 10 the chair ofelecuical engineering
362 -367 at Heriot-Watt Univesity, Edinburgh, U.K., in 1986. His m w c h activity
1201 EIHahrouk, M.: Da-sh. M.K.; Mehw. P.; "Active power filters: a review", includes power semiconductor modeling and protection mnvener topologies and
Elecric Power Appli~;ltions,IEE Prmeedings- , Vd.147. Issue: 5. 2000, soflswitching techniques, and the applialion Of ASlCs and ficmprocewrs to
pp:403 4 1 3 indusuial electronics.

159