You are on page 1of 9

Electric Power Systems Research 60 (2001) 29 37

A new compensation algorithm for balanced and unbalanced

distribution systems using generalized instantaneous reactive power
Mahesh K. Mishra, Avinash Joshi, Arindam Ghosh *
Department of Electrical Engineering, Indian Institute of Technology, Kanpur 208 016, India
Received 2 May 2001; received in revised form 31 July 2001; accepted 12 September 2001

This paper discusses various compensation strategies for shunt active power filters using a generalized theory of instantaneous
reactive power. A general instantaneous vector expression for filter current in terms of active and reactive powers has been
derived. The general time domain algorithm for filter reference currents in terms of source powers has been given. It is shown that
the algorithm works under balanced and unbalanced source voltages while producing a set of balanced three-phase source currents
at a desired power factor. Detailed simulation results using MATLAB have been presented to validate the proposed theory for
balanced as well as unbalanced supply voltages. 2001 Published by Elsevier Science B.V.
Keywords: Shunt active power filter; Compensation algorithms; Unbalanced systems; Generalized instantaneous reactive power theory

1. Introduction
The use of power electronics has grown considerably
over the last couple of decades. This has increased the
reactive power demand and the harmonic pollution in
power industry. To eliminate these harmonics and to
improve the power quality, it is necessary to have the
proper measures and corrections for them. From this
point of view, many definitions and formulations of
reactive power have been proposed in literature [18].
Akagi et al. [1,2] provided elegant definitions of powers
for compensation, but did not consider zero sequence
components. In [5,8] zero sequence power was considered, but the formulations make the use of the h i0
transformation. As a consequence of this, the algorithm
is computation intensive. In [4], the decomposition of
currents into the orthogonal components rather than
power has been proposed.
In [7], the general definitions of reactive and active
powers have been presented. In this formulation, the
* Corresponding author. Tel.: + 91-512-597179; fax: + 91-512590063.
E-mail address: (A. Ghosh).

active and reactive powers are expressed as the dot and

the cross product of voltage and current vectors, respectively. It is further shown that the pq theory [1,2]
is a special case of these general definitions. However,
the instantaneous vector constructed for filter current in
[7] is for a special kind of compensation, where only the
total reactive power of the load is compensated. The
theory is far more powerful and provides an effective
tool for analyzing and understanding of shunt
In this paper, the theory has been extended and a
more general vector equation for filter current has been
proposed. The equivalent time expressions give a general filter current algorithm in terms of the desired
source powers. Several types of compensations can be
obtained by suitable choice of the desired source pow-

Fig. 1. Schematic diagram of a three-phase four-wire system.

0378-7796/01/$ - see front matter 2001 Published by Elsevier Science B.V.

PII: S 0 3 7 8 - 7 7 9 6 ( 0 1 ) 0 0 1 6 6 - 3

M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


ers in the algorithm. Three cases have been analyzed for

a three-phase, four-wire distribution system with balanced voltages. The proposed general filter current
algorithm has been modified for obtaining correct compensation when applied to a system with unbalance in
voltage magnitudes and phase angles. The modified
algorithm (Equal Current Strategy) has been verified by
simulation. For the special case of unity power factor
operation an alternative simple algorithm (Equal Resistance Strategy) has also been derived and verified by

The instantaneous active current vector, ip and the

instantaneous reactive current vector, iq are defined as,

ip = ibp =



q 7
iq = ibq =



The total current vector is the sum of active and

reactive current vectors i.e.

2. Basic definitions


i= ip + iq
The basic definitions of various terms have been
clearly defined in [7]. However, for the sake of completeness, these definitions are reviewed below.
Consider a three-phase, four-wire system with instantaneous voltages and currents as shown in Fig. 1. In the
following, subscripts a, b and c denote the respective phases and vector quantities are shown boldfaced.
The instantaneous vectors, 7 and i are defined as follows.

7 = 6b




i= ib



The instantaneous active power is defined as the

scalar (dot) product of the vectors 7 and i, i.e.
p= 7 i= 6aia +6bib +6cic

Further 7 iq 0, and 7 ip 0, i.e. iq is orthogonal

to 7 and ip is parallel to 7.
The instantaneous apparent power s is a scalar
defined as,
s= 7 i


where 7 =
6 + 6 + 6 and i =
i + i +i are
the norms of the instantaneous voltage and current
vectors, respectively. The instantaneous power factor u
can be defined as ratio of active and apparent power i.e.









The properties of the quantities defined above have

been discussed in detail in [7].


The instantaneous reactive power is a vector defined


3. General expressions for instantaneous real and

reactive powers in terms of symmetrical components

6b 6c

ib ic

6c 6a

q = 7 i= qb =
ic ia

6a 6b

a b

Using theory of instantaneous symmetrical components [9], we can express source voltages and currents in
terms of sequence components. These components are
complex numbers. We have,

where denotes cross product of the vector 7 and

vector i.
Further the norm of vector q =
q 2a +q 2b + q 2c
may be sometimes used as the scalar representing the
instantaneous reactive power. Alternatively, the algebraic sum qsum =(qa +qb +qc)/
3 can also be used as
the scalar representing the total reactive power that
circulates in the three phases. An advantage of this
representation is that it can indicate the polarity of
instantaneous reactive power unlike q , which is always positive.

1 1

6sb =
1 a 2

1 a

1 6a0

a 6a1

a 2 6a2


A similar transformation can be written for currents

1 1

isb =
1 a 2

1 a

1 ia0

a ia1

a 2 ia2


where a= e j120. The subscript s denotes source. It is

to be noted that the instantaneous vectors 7a1 and 7a2
(called instantaneous positive and negative sequence
voltages) are complex conjugate of each other and 7a0 is
a real quantity, which is zero if the voltages are balanced. These facts also apply to ia1, ia2 and ia0.

M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


Using Eqs. (2), (3), (8) and (9), we can compute the
active and reactive powers in terms of instantaneous
symmetrical components. On simplification, we get,
ps = 6a0ia0 + 2Re(6a1i*a1 )



Im 6a1i*a1
qs = qsb =

3 6 i*
a1 a1

j 2y/3

6a0Im ia1e
+ j 2y/3




j 2y/3
ia0Im 6a1e

+ j 2y/3


Fig. 2. A three-phase, four-wire compensated system.


Similarly, as per Eqs. (2) and (11a)

pL = pL + pL

In the above, Re denotes the real part and Im, the

imaginary part and *, the conjugate of a complex
quantity. The relations (Eqs. (10a) and (10b)) are very
useful in explaining the compensator performance when
the system voltages are unbalanced. Similarly, the instantaneous scalar source reactive power is given by
qs sum =

(qsa +qsb +qsc)

= 2Im(6a1i*a1 )

qLa qLa qLa

qL = qL + qL = qLb = qLb + qLb

qLc qLc qLc


It is seen from Eq. (10b), that the terms containing

6a0 and ia0 do not contribute to the qs sum. In general, qs
sum has a dc component, qs sum, and a double frequency
(100 Hz for 50 Hz fundamental) component, qs sum.
Thus only first term of Eq. (10b) contributes to the
average reactive power, qs sum.

In general, the active and reactive powers have two

components i.e.
p=p + p


q= q +q


where, overbar () and overtilde ( ) refer the average

and zero mean oscillating values of the powers. In the
following discussion, the subscripts s, L or l and f
will denote the source, load and compensator (filter),
respectively. Load currents are written with subscript
l, while load powers are written with subscript L. For
example pL denotes the average load power, qs denotes
the zero mean oscillating instantaneous reactive source
power and il denotes the instantaneous load current
vector. It should be noted that as per Eqs. (3) and


Further, qL is the average value of qL sum, the instantaneous scalar reactive load power i.e.
qL = qL sum =

(qLa + qLb + qLc)


Using the above definitions (Eqs. (4a), (4b) and (5)),

the instantaneous vector of filter reference currents in
terms of the filter active power (pf) and reactive power
(qf) is as follows
i*fp + i*fq =

q 7s
+ f
7 s 7 s 7s 7 s


Similarly, the instantaneous source current vector,

written in terms of the source reactive power (qs) and
active power (ps) is as follows
q 7s
+ s
7s 7s 7s 7s

is = isp + isq =
4. General expressions for filter and source currents



The general schematic diagram of a three-phase,

four-wire compensated system is shown in Fig. 2. The
compensator is realized with three ideal current sources
i *,
fa i *
fb, i *.
fc Applying KCL at the three nodes of filter
and load coupling and expressing currents in vector
form, we get,

il is

We expand the cross product in Eq. (16) by applying

the definition similar to Eq. (3). This gives the following
time domain expressions for filter currents
i*fa = ila


(ps6sa + qsb6sc qsc6sb)


(ps6sb + qsc6sa qsa6sc)


6 2sj

j = a,b,c

i*fb = ilb

j = a,b,c


M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


i*fc = ilc

(ps6sc +qsa6sb qsb6sa)


sin 100yt

6sb = 440

sin 100yt

6sc = 440

sin 100yt +

6 2sj

j = a,b,c

In the above, ps is the source active power and qs = [qsa

qsb qsc]t is the source reactive power. The superscript t
denotes transpose operator.


6sa = 440

5.1. Simulation example: balanced 6oltages

The load consists of (i) unbalanced three-phase RL

load: Za = 500 V, Zb = 10045 V, Zc = 7530 V;
(ii) three-phase full bridge converter with firing angle of
For a given system voltages and load configuration,
the steady state load currents are computed. The computed load currents ila, ilb, ilc have been plotted in Fig.
3a. Then, the power terms pL, and qL (whose components
are qla, qlb and qlc) are computed using Eqs. (2) and (3),
respectively. Using the samples of pL, and qL, we compute
p L and q L using Moving Average Filter over half cycle
[10]. Vector q L is computed by subtracting q L from qL
using Eq. (12). The scalar average reactive power of load,
q L, is computed using Eq. (14). The active powers pL, p L
and reactive powers qL and q L are shown in Fig. 3b and
c, respectively. Thus, the power terms pL, p L, qLa, qLb, qLc,
q La, q Lb, q Lc and q L are known. These values are
substituted, according to Table 1, for ps, qsa, qsb and qsc
in Eqs. (18)(20) to generate filter reference currents so
as to achieve different compensation schemes.
All three cases of compensation given in Table 1 have
been considered for simulation. The second terms on
right hand sides of Eqs. (18)(20), corresponding to the
compensated source currents, have been plotted for these
compensation schemes. In the following figures, the
voltage waveform of phase a only has been plotted. It
has also been reduced by a factor of 20, so that its
numerical values are comparable to those of source
currents. In the following discussion, compensated
source currents refer to the currents drawn from source
after compensation.

The compensation schemes given in Table 1, have been

simulated using MATLAB. The following system parameters have been considered for simulation.

5.1.1. Case 1
The source supplies the average load power (p L) and
the total or a part of the average reactive power of the

5. Compensation under balanced voltages

Equations (Eqs. (15) and (16)) give the filter currents
and source currents in a vector form. The alternate
equations (Eqs. (18) (20)) for reference filter currents are
equivalent to Eq. (15), but they use the desired source
powers instead of filter powers. Considering filter current
sources (Fig. 2) to be ideal, the appropriate selection of
source power terms, ps, and qsa, qsb, qsc in Eqs. (18)(20)
gives the desired compensating characteristics. The three
compensating schemes and the corresponding selection
of power terms in Eqs. (18) (20) are proposed and these
have been shown in Table 1.
The power terms in Table 1 have been defined in Eqs.
(12) (14). In Table 1, i is a factor, usually between 0 and
1, which decides the desired power factor of the source.
The relation between i and the desired phase shift
between source current and source voltage is given by,
tan =

Average reactive supplied by source

= L
Average real power supplied by source
p L

The reference currents obtained using the theory of

instantaneous symmetrical components in [10] can accommodate a variable source power factor. However, it
assumes that the compensator supplies the zero-oscillating load active power and a part of or total reactive
power as in Case 1. The other two cases given in Table
1 cannot be readily achieved using the algorithm of [10].

Table 1
Various compensation schemes under balanced voltages

Active and reactive powers drawn from source


Active and reactive powers supplied by








p L
p L
p L

iq L/3
q La

iq L/3
q Lb

iq L/3
q Lc

p L
p L
p L


M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


Fig. 3. (a) Three-phase load currents. (b) Load active power. (c) Load reactive power.

load. The compensator supplies the remaining reactive

power of the load. It is desired that the phase lag
between the source currents and the source voltages be
30 lagging. To achieve this, using Eq. (21) i is 0.67. It
is seen from Fig. 4 that the compensated source currents are sinusoidal and balanced and have a phase
angle of 30. In order to obtain unity power factor, i is
set to zero. If i is varied from 0 to 1, the power factor
angle varies from 0 to 41. The values of i larger than
unity are not desired as the power factor angle increases
even further. The source power factor can even be
made leading by choosing negative values of i.

notches. The unbalance results because unequal reactive

powers are drawn from the phases. It is possible by
averaging to redistribute the total reactive power
equally among three-phases. In this situation, the compensated source currents, though distorted, will be similar in shape. The notches result from discontinuities in
p L and q L, (Fig. 3) due to ideal thyristor bridge. The
simulation of the cases 1 and 2 in Table 1, using pq
theory, has been reported in [5]. For these cases, the
simulated results given above using generalized instantaneous reactive power theory are found to be similar.

5.1.2. Case 2
The compensator compensates only for average reactive load power q L and zero mean oscillating active
power p L, while the source supplies the remaining part
i.e. q L and p L. The results are shown in Fig. 5, where
the fundamentals of source currents are seen to be in
phase with voltages. The source currents are unbalanced and non-sinusoidal due to supply of q L from the

6. Compensation under unbalanced voltages

5.1.3. Case 3
The compensator supplies only the zero mean oscillating active power (p L) of the load and the rest of the
powers of load are supplied by the source. This case
arises if we want to suppress the mechanical vibrations
in the prime movers arising due to supply of p L [5]. The
source currents, shown in Fig. 6, are unbalanced and
non-sinusoidal. The fundamental components of currents also show a phase shift with respect to the source
Thus, we can make the compensator take action on
any of the power terms q L, q L and p L, or their various
combinations. However, the average load power p L and
the losses in a real compensator always come from the
source. We can consider the cases in which the source
supplies the total load active power pL instead of p L but
this is not preferred as it results in additional distortion
in source currents.
It is observed from simulation results that in cases 2
and 3 that the source currents are unbalanced and have

The definitions given in Section 2 are general and do

not assume balanced source voltages. Thus, the proposed theory may be applied in case of the unbalanced
voltages. It has been shown in Section 6.1, that the
application of filter current algorithm (Eqs. (18)(20))
to an unbalance system results in distorted source currents. The algorithm is modified to obtain correct compensation as given in Sections 6.2 and 6.3.

6.1. Simulation example: unbalanced 6oltages

The following magnitude and phase unbalance in
supply voltages have been considered.


6sa = 440

sin 100yt

Fig. 4. Compensating p L, q Land a part of q L (Case 1).

M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937



Fig. 5. Compensating p L and q L (Case 2).

6sb = 1.2 440

sin 100yt + 30

6sc = 0.8 440

sin 100yt + 10

The unbalanced source voltages are plotted in Fig. 7.

The load currents will change accordingly. Considering
unity power factor operation, (Case 1 with i = 0 in
Table 1), the compensated source currents are shown in
Fig. 8. The load active power, pL is shown in Fig. 9.
For balanced three-phase system, with three phase balanced sinusoidal voltages and balanced sinusoidal currents, the three phase instantaneous active power is
always a constant. Similarly, the three instantaneous
reactive powers, qsa, qsb, qsc are also constant.
In a compensated system, when source voltages are
balanced, 6a0 =0. However, the load currents may be
balanced or unbalanced. But due to compensator action, the source currents become balanced, and therefore ia0 =0. Consequently, the first term of Eq. (10a)
and second and third terms of Eq. (10b) are zero and ps
and qs become constant. Thus, the compensated system
becomes a three-phase balanced system.
When voltages are unbalanced, the unbalance is
passed on to the compensator control algorithm (Eqs.
(18) (20)) and results in a particular kind of compensation. The compensator does satisfy the overall conditions imposed on it by the algorithm (Eqs. (18)(20)),
i.e. ps = p L (Fig. 9) and each component of qs is zero.
Since the voltages are unbalanced, and we constrain the
source to supply a constant power, ps, equal to the
average load power, p L, the currents distort from the

Fig. 6. Compensating only pL (Case 3).

Fig. 7. Unbalanced source voltages.

sinusoidal as shown in Fig. 8. As a consequence of

voltage unbalance, both 6a0 and ia0 are non-zero, time
varying and ia0 is also non-sinusoidal. Thus in Eq.
(10a), the first term, i.e. 6a0ia0 becomes non-sinusoidal
whose ac part is equal and opposite to ac part of the
second term. Similarly, the second and third term in
Eq. (10b) in each phase yield non-sinusoidal variations.
These non-sinusoidal variations are measure of the
distortion in source currents seen in Fig. 8. To overcome this problem the algorithm (Eqs. (18)(20)) is
modified as described in the following.

6.2. Modified algorithm for unbalanced 6oltages

Let there be unbalance in magnitudes and in phase
angles of the supply voltages, i.e.
6sa = Vsma sin t





6sb = Vsmb sin t

6sc = Vsmc sin t +


In above, the magnitudes Vsma, Vsmb, Vsmc are unequal. The phase angles qb and qc contribute to the
phase unbalance. The algorithm given in Eqs. (18)(20)
is modified by using a fictitious set of voltages 6 %sa, 6 %sb,
6 %sc in place of the real voltages 6sa, 6sb, 6sc given by Eqs.
(22a), (22b) and (22c). The fictitious set of voltages are
balanced and are given by,
6 %sa = V %sm sin t

Fig. 8. Compensated source currents.


M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


Fig. 10. Compensated source currents for Equal Current Strategy.

Fig. 9. Instantaneous and average active powers.

6 %sb = V %sm sin t



6 %sc =V %sm sin t +



The use of balanced voltages in the algorithm produces balanced compensated source currents as shown
in the example of Case 1 in Section 5, (Fig. 4). For
balanced compensated source currents, both sets of
voltages given by Eqs. (22a), (22b), (22c), (23a), (23b)
and (23c) above, should yield equal average real power,
p s. From this requirement, we obtain the magnitude
V %sm as,
IsmV %sm cos = (VsmaIsm cos +VsmbIsm cos( + qb)
+VsmcIsm cos( + qc))
The above equation leads to
V %sm = (Vsma +Vsmbhb +Vsmchc)


In the above equation,

hb =

cos( + qb)


hc =

cos( + qc)


and is the desired phase angle between the source

voltage 6sa and the desired compensated source current
isa. In the case of unbalance in magnitudes only i.e.
qb = qc =0, the factors hb and hc in Eq. (24) reduce to
unity and consequently V %sm is the average of the unequal magnitudes Vsma, Vsmb, Vsmc.
Based on above considerations, the modified algorithm for filter reference currents is as follows,

i*fa = ila

(ps6 %sa +qsb6 %sc qsc6 %sb)


(ps6 %sb + qsc6 %sa qsa6 %sc)


(ps6 %sc +qsa6 %sb qsb6 %sa)


Under balanced voltages i.e. Vsma = Vsmb = Vsmc and

qb = qc = 0, the algorithm given above converges to
that given in Eqs. (18)(20). Under balanced source
voltage conditions the algorithm given in Eqs. (26a),
(26b) and (26c) converges to that given in [10] when
operated under unity power factor condition.
The power terms in Eqs. (26a), (26b) and (26c) are
obtained from Case 1 in Table 1 and measured load
powers by the same procedure as in the case of the
balanced voltages as discussed in Section 5. With this
algorithm, considering the same parameters as in Section 6.1, the simulated results for unity power factor
operation i.e. i=0 are shown in the Figs. 10 and 11. It
is seen that the compensated source currents obtained
from the modified algorithm are balanced sinusoids,
unlike the distorted waveforms of Fig. 8.
It is to be noted that if we consider 6 %sa, 6 %sb, 6 %sc,
fictitious balanced supply voltages then the fictitious
source would supply only average load power and no
zero mean oscillating active and reactive power. But
since the actual supply has unbalance in magnitudes, so
the source supplies some zero mean oscillating active
and reactive powers as shown in Fig. 11.
The above algorithm may be referred to as Equal
Current Strategy. Similar strategies have been reported
as Sinusoidal Current Source Strategy in [8] and Equal
Current Criteria in [11]. However in [8], to find the
fictitious balanced set of voltages for control algorithm,
the complex hi0 transformations and elaborate
computations are used. In [11] the synchronous detection method is used but it is limited to the case of unity



j = a,b,c

i*fb = ilb


j = a,b,c

i*fc = ilc

j = a,b,c



Fig. 11. Instantaneous active and reactive powers for Equal Current

M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937


power factor and magnitude unbalance only. The

modified algorithm given above, using generalized theory is simple and considers unbalances in magnitudes
and/or phase angles. The algorithm also provides the
facility of setting the desired power factor angle, .

6.3. An alternate algorithm for unity power factor

Equal Resistance Strategy has been reported in [11].
This strategy can also be realized using generalized
algorithm (Eqs. (18) (20)) with appropriate choice of
power terms ps, qsa, qsb and qsc. One way to remove the
non-sinusoidal variations of the terms in Eqs. (10a) and
(10b) described above is to impose the following condition to the source currents.
6sa 6sb 6sc
= = = Req
isa isb isc


where Req is equivalent resistance of each phase as seen

by the source. Inverting Eqs. (8) and (9), and applying
Eq. (27), it can be shown that 6a0 =ia0Req, 6a1 =ia1Req
and ia0 is sinusoidal. Therefore, the first term in Eq.
(10b) is zero by itself and second and third terms cancel
in each phase. Hence the vector qs and its components
qsa, qsa and qsc in Eq. (10b) become zero. Further, real
power ps Eq. (10a) consists of a dc and a sinusoidal
part in each term. Because the compensated system is
equivalent to a resistance in each phase, the power
drawn from the source is given as,
ps =

6 2sj

j = a,b,c



Using ps from Eq. (28) and qsa =qsb =qsc =0, we get
from Eqs. (18)(20):
i*fa = ila



i*fb = ilb





i*fc = ilc

Fig. 12. The compensated source currents for Equal Resistance



6 2sj=

V 2smj

j = a,b,c

j = a,b,c


Using Eq. (31), the value of Req is given as,

Req =

V 2smj

j = a,b,c

2p L


where Vsmj ( j= a, b, c) is the peak value of voltage in

phase j. Substituting Req from Eq. (32), into Eqs. (29a),
(29b) and (29c) the instantaneous reference filter current vector is given as:
2p L sa

% 6 2smj 6sc
j = a,b,c

Under balanced source voltages, the algorithm (Eq.

(33)) tends to algorithm (Eqs. (18)(20)) for case 1 with
i= 0. The performance of the improved compensator
with voltage unbalance has been simulated using same
parameters as in Section 6.1. The compensator uses Eq.
(33) to generate the reference filter currents. The compensated source currents are shown in Fig. 12. The
active source and load powers are plotted in Fig. 13.
Comparing the source current waveforms of Figs. 8
and 12, it is seen the later has no distortion. The source
currents are purely sinusoidal and in phase with the
respective phase voltages. However, this is achieved at
the cost of drawing a particular value of zero mean
oscillating active power from the source. This is seen in
waveform of ps, which contains 100 Hz ac component
in addition to p L as shown in Fig. 13. Thus if we use the

To determine Req, the average value of ps in Eq. (28)

is equated to the average load power. This gives:
p s =

j = a,b,c


6 2sj
= p L


where the symbol refers for the average value over

half cycle. For sinusoidal voltages (with or without
unbalance in magnitudes and phase angles), the following relation is true

Fig. 13. Source and load powers for Equal Resistance Strategy.

M.K. Mishra et al. / Electric Power Systems Research 60 (2001) 2937

condition Eq. (27), we get improvement over the distorted source currents of Fig. 8 by redistribution of the
zero mean oscillating active power between the source
and the compensator.


To maintain the capacitor voltage constant despite the

switching losses, a feedback loop is usually included.
The reference current generation scheme can then be
suitably modified to incorporate the feedback signal as
in [10].

7. Conclusions
In this paper, shunt compensation algorithms based
on generalized instantaneous reactive power theory
have been proposed. The general filter current expressions (Eqs. (18)(20)) have been derived in terms of
desired source powers. Different choices of these powers in terms of the load power give rise to different
compensation schemes (Table 1). These filter current
expressions are very general and give correct compensation for balanced voltages, with or without zero sequence components. The proposed theory has been
extended for unbalanced source voltage operation by
introducing a fictitious set of system voltages, which
gives the same average load power. The modified filter
current algorithm gives correct compensation to obtain
any desired source power factor for unbalance in source
voltage magnitudes and/or phase angles. This can be
called Equal Current Strategy, as the compensated
three-phase source currents are balanced sinusoids. For
the special case of unity power factor operation, an
alternative Equal Resistance Algorithm has also been
All these algorithms have been verified by simulation
using MATLAB. The detailed simulation results have
been given for each case. Note that algorithm can not
be directly be used when the source voltages are distorted. In this situation, fundamental extraction of distorted source voltages must be used. In the proposed
work, an ideal compensator is considered. However, a
practical compensator is usually supplied by a dc storage capacitor. The capacitor voltage is pre-charged to a
voltage that is higher than the peak of system voltage.

[1] H. Akagi, Y. Kanazawa, K. Fujita, A. Nabae, Generalized

theory of instantaneous reactive power and its application, Electrical Engineering in Japan 103 (4) (1983) 58 65.
[2] H. Akagi, Y. Kanazawa, A. Nabae, Instantaneous reactive
power compensators comprising switching devices without energy storage components, IEEE Transactions on Industry Application 20 (3) (1984) 625 630.
[3] T. Furuhashi, S. Okuma, Y. Uchikawa, A study on the theory of
instantaneous reactive power, IEEE Transactions on Industrial
Electronics 37 (1) (1990) 86 90.
[4] J.W. Willems, A new interpretation of the Akagi Nabae power
components for nonsinusoidal three-phase situations, IEEE
Transactions on Instrumentation and Measurement 41 (4) (1992)
523 527.
[5] E.H. Watanabe, R.M. Stephan, M. Aredes, New concepts of
instantaneous active and reactive powers in electrical systems
with generic loads, IEEE Transactions on Power Delivery 8 (2)
(1993) 697 703.
[6] M. Aredes, E.H. Watanabe, New control algorithms for series
and shunt three-phase four-wire active power filters, IEEE
Transactions on Power Delivery 10 (3) (1995) 697 703.
[7] F.Z. Peng, J.S. Lai, Generalized instantaneous reactive power
theory for three-phase power systems, IEEE Transactions on
Instrumentation and Measurement 45 (1) (1996) 293 297.
[8] M. Aredes, J. Ha fner, K. Heumann, Three-phase four-wire shunt
active filter control strategies, IEEE Transactions on Power
Electronics 12 (2) (1997) 311 318.
[9] W.A. Lyon, Transient Analysis of Alternating-Current Machinery, Wiley, New York, 1954, pp. 28 42 Chapter 2.
[10] A. Ghosh, A. Joshi, A new approach to load balancing and
power factor correction in power distribution system, IEEE
Transactions on Power Delivery 15 (1) (2000) 417 422.
[11] C.L. Chen, C.E. Lin, C.L. Huang, Reactive and harmonic
current compensation for unbalanced three-phase systems using
the synchronous detection method, Electric Power System Research 26 (1993) 163 170.