IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO.

4, JULY 2008 1945

Discussion on Instantaneous p–q Strategies

for Control of Active Filters
Gabrio Superti-Furga, Member, IEEE, and Grazia Todeschini, Student Member, IEEE

Abstract—Active filters (AFs) compensate harmonic currents was very simple and because it was the first dedicated to the
demanded by non linear or time variants loads, with the aim three-phase system. Two different approaches were proposed:
to draw sinusoidal currents from the source. Different control
the first one [6] achieves a nil instantaneous reactive power
strategies for AF have been proposed: among them, – theory
is the most popular. However, the – strategy presents some without altering the instantaneous active power, the second one
drawbacks. As already reported in the literature, unexpected [7] adds another constraint: the power absorbed from the grid
harmonic frequencies are generated in the source current as a has to be constant. A wide and detailed discussion on – the-
consequence of such a control strategy. ories and different variants in compensation strategies are pre-
The aim of this paper is to provide, by means of the space vector
approach to the – theory, analytical expressions that allow to sented in [8].
determine the order and the amplitude of the harmonic compo- Many applications of the “instantaneous – compensation”
nents in the source current as function of load current or voltage have been proposed: most of them refer to the second approach
harmonic distortion. By means of these expressions the limits and [1]–[3], [5], [10]. However, it is well known that both the
drawbacks of – strategies are quantified and a tool able to quan-
titatively predict the results of such a control is provided. Simula- methods generate harmonics in the source current. Referring
tion results presented in the literature are interpreted according to to the second method, the drawbacks have been shown in early
the proposed expressions; moreover, further simulations are car- papers [11]–[13], by the Nabae–Akagi group [14], [15] and in
ried out in order to analyze some relevant cases not considered in recent ones [1]–[3], [5], [10], [16]–[18].
previous works.
The reason of this phenomenon is easy to understand. In a
Index Terms—Active filters (AFs), compensation, harmonics, three-phase system, if both the voltage and the current are sinu-
– theory, space vector. soidal and balanced, the power is constant. If the voltage wave-
form contains harmonics while the current is sinusoidal and
balanced, the power is not constant. As a consequence, if the
I. INTRODUCTION voltage contains harmonics and a constant power is imposed,
HE latest generations of converters, thanks to the high
T switching frequency and efficient current tracking tech-
niques, can easily shape the waveform of the output current.
some harmonics, in a broad sense complementary to the ones
present in the voltage, are introduced in the current. In [18]
this problem is overcome by preliminary filtering the voltage
The problem of the control and the related current wave shape in order to make it balanced and sinusoidal.
is particularly significant and debated for three-phase active A similar problem of harmonic content in the grid current
filters (AFs). In the literature, many papers appear where the actually exists for active front end converters, such as four-
different approaches to the control are discussed and results quadrant converters, self commutated rectifiers and matrix con-
are compared: [1]–[5] are a selection of the most recent ones. verters. Referring to this, matrix converters are critical [19]: they
In particular, [2] shows a comprehensive Fourier analysis of lack of the energy storage device and, as a result, the instanta-
the simulation results, some of them discussed in detail in the neous power delivered to the load is equal to the instantaneous
present paper. power absorbed from the grid. Moreover, since in many cases
The problem of control can be summarized as follows: given the load voltage is intentionally sinusoidal and balanced and the
the voltage and the power, the waveform of the current absorbed load is balanced, the power delivered to the load is constant,
from the grid has to be determined. The instantaneous power and the power absorbed from the grid must be constant too, ir-
can have oscillations around the mean value, related to current respective of grid characteristics. If the grid supplies a distorted
waveform and depending on the control strategy. Providing that voltage, a situation similar to the second – approach should
the AF comprises an energy storage device, power oscillations arises, although this drawback seems not reported by any au-
on the grid side and load side may be not related. A very pop- thor.
ular strategy is the “instantaneous – compensation,” proposed The – theory had been initially developed for three-wire
in the 1980s by the Nabae–Akagi group: it spread because it systems. Its original formulation is, of course, still valid in four-
wire systems, provided that the fourth-wire current is kept nil by
Manuscript received December 1, 2007; revised March 10, 2008. Published means, for example, of a compensating device. In order to fully
July 7, 2008 (projected). Recommended for publication by Associate Editor
V. Staudt.
handle four-wire systems, a modified – theory has been pro-
G. Superti Furga is with the Electrical Engineering Department, Politecnico posed [20], [21]. Discussions on this subject and comparisons
di Milano, Milano, Italy (e-mail: gabrio.supterifurga@etec.polimi.it). are reported in [22], [23]. This latter approach allows to keep
G. Todeschini is with the Electrical and Computer Engineering Depart- or achieve a nonzero current in the fourth wire on the grid side,
ment, Worcester Polytechnic Institute, Worcester, MA 01609 USA (e-mail:
grazia@wpi.edu). resulting in lower collective rms current [22] and in lower trans-
Digital Object Identifier 10.1109/TPEL.2008.925439 mission losses [24].
0885-8993/$25.00 © 2008 IEEE
1946 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008

A more powerful extension appears to be the – – theory Under sinusoidal and balanced (positive sequence) steady-
[25]–[27] that allows a unified approach in handling three and state of angular frequency , (3) becomes
four-wire systems and in compensation with or without energy
storage element [28], [29]. (4)
In the present paper only the original – theory on three-wire
systems is considered; discussions over four-wire extensions are In this case from (2) and (4) it results that the and com-
beyond the purpose of the paper. The Authors’ choice is justified ponents are sinusoidal and orthogonal
by considering that the original – approach is actually the
most widely used in applications. Moreover, in practical cases,
any grid current in the forth-wire, is very often avoided. (5)
Although some specific cases have been analyzed in the lit-
erature, an analytical explanation of the reason why the – ap- Under distorted periodical steady state of period , the space
proach introduces harmonics in the current has never been pro- vector can be expanded in complex Fourier series [31], [32]
vided in a general form. The aim of the present paper is to fill
this gap by providing analytical formulas that allow to foresee (6)
which new harmonics will be generated, their amplitude and se-
quence sign (positive or negative). Some unexpected results are
shown, and in particular it is explained how dangerous situations where and .
may arise when the dc current is involved. Positive indices correspond to positive sequence terms:
The paper is organized as follows. In Section II mathemat- in particular is the fundamental positive sequence. Nega-
ical tools convenient for the analysis of three phase systems are tive indices correspond to negative sequence terms. Index
introduced. In Section III the two – approaches proposed by corresponds to the constant (average) term. Unbalanced
Nabae–Akagi group are described. In Section IV the drawbacks sinusoidal steady-state is described by the two terms
of the above mentioned approaches are analytically quantified. only.
In Section V some numerical results presented in literature are Periodical distorted but balanced three-phase system is de-
discussed and Section VI carries out further simulation results. fined as . Under this condi-
tion, it has been proved [31] that in (6) only indexes
II. THREE PHASE ANALYSIS ( integer) are present. Moreover, if only odd harmonics occur,
the previous formula becomes
Let us consider a three-phase three-wire system. Let
be the generic phase quantities. The Clarke (7)
transformation (or Park-Gorev transformation on stationary
reference frame) applies to the instantaneous values The (7) provides the usual harmonic components
in distorted and balanced three-phase systems:
, in which the sign gives infor-
mation on the associated sequence type (positive or negative).
(1)
III. INSTANTANEOUS – THEORY
where and are the instantaneous and compo- In [6] the first definition of instantaneous active power
nents. The instantaneous zero sequence component is not con- and instantaneous reactive power is provided
sidered because in three-wire systems the zero sequence current
is nil and as a consequence the zero sequence voltage, if any, has (8)
no effects. Organizing the and components in one complex (9)
quantity, the space vector (or Park vector) [14], [30]–[32] is de-
fined where the , terms are evaluated according to (1).
Two compensation methods have been proposed by
(2) Nabae–Akagi group for the control of shunt AF, based on
(8) and (9).
or, from (1)
A. First Compensation Method
(3) In this approach [6], is fully compensated and is not
altered. Being and the power terms referred to the
The space vector (3) fully represents the three-phase three- load, on the source side the following constraints are imposed
wire quantities in any working condition. The original phase (Fig. 1)
components can be found from the space vector applying the
inverse Clarke transform and the further condition of zero sum (10)
of the phase quantities.
SUPERTI-FURGA AND TODESCHINI: DISCUSSION ON INSTANTANEOUS – STRATEGIES 1947

considered that in the Voltage Source Inverters (VSI) a dc

capacitor is necessary for the commutations: it is cheap to
extend its function to the accumulation of the energy due to
the oscillations , since these oscillations are usually small
in the three-phase systems. As a consequence, with very little
effort the VSI becomes more effective and a constant power is
obtained. Due to these considerations, in the common applica-
Fig. 1. First p–q compensation method. tions the second method is more widely used that the first one.
However, this method presents a drawback, too. If the voltage
is not sinusoidal and balanced, the obtained current is nonsinu-
soidal. In particular the third harmonic is present under voltage
unbalance [1], [2], [15], [16], [18]. Some Authors proposed
corrections to the method: practically, before implementing the
method, they apply a filter to the voltage so that it becomes si-
nusoidal [18].
In spite of the previous described disadvantages, the second
Fig. 2. Second p–q compensation method. Nabae–Akagi compensation method is very popular, mainly for
two reasons.
At first, it was one of the first approaches that exploited
From (8)–(10), the reference currents on the source side are the properties of the three-phase system, while the previous
approaches were single-phase based. With the three-phase
(11) approach, the goal of the compensation becomes the sinusoidal
and balanced waveform, while the simply sinusoidal waveform
achieved when the single-phase approaches are applied.
The reference currents (11) depend on the waveform of the
Secondly, it is a skilful way to perform the complex Fourier
load power and of the common voltage. They do not depend
analysis (6) and to get the fundamental harmonic term of the
directly on the load current.
current, provided that the voltage is sinusoidal and balanced. In
This method actually presents the following drawback: it is
fact the canonical Fourier analysis that is implemented to obtain
well known that if the current is not sinusoidal and bal-
the sinusoidal balanced current is substantially repeated by the
anced, the current is not sinusoidal and balanced too. The
method:
order and amplitude of harmonic components is different in
1) the three-phase currents are multiplied (8) by the sinusoidal
and in : in particular the third harmonic appears in
and orthogonal analyzing functions , (5);
the Fourier series expansion of under unbalanced load [13],
2) the result is averaged on one period, in order to obtain
[14], [16], [17].
(13), that plays the role of the first coefficient of the Fourier
B. Second Compensation Method series;
The second method [7] is based on the decomposition of the 3) the obtained signal is multiplied (14) by the same analyzing
instantaneous active power in the average part and the oscil- functions , to go back to the time domain (the
lating part denominators in (14) are constant under sinusoidal bal-
anced conditions);
(12) 4) the unity p.f. is obtained since the analyzing functions rise
from the voltage itself.
The second method consists in the compensation of and The method loses his validity if the voltage is not sinusoidal
of the oscillating part in (12): as a result, only the mean or balanced, because and are no longer proportional
value of the active power is absorbed from the grid (Fig. 2). The to and . Nevertheless it must be noted that if
constraints on the source side are the actual voltage distortion or unbalance is due only to the
distorting load to be compensated, the compensation method
(13) makes both voltage and current to converge into sinusoidal and
balanced waveforms. Actually only the residual distortion com-
From (8), (9), and (13) the reference currents for the source ponent (back e.m.f. distortion) at PCC reduces the performances
side are of the method; however, considering that nowadays distortions
in the voltage are spread and that, on the other hand, the AF
(14) devices are more efficient and precise, the described problem
becomes more important.
This time the reference currents (14) depend on the average
load power and on the waveform of the common voltage. They IV. ANALYSIS OF – METHODS
do not depend directly on the load current. For the analysis of both methods, the space vector approach
This second approach needs an energy storage device to is chosen because of its efficiency. In particular the complex ex-
handle the oscillations of power . However, it should be pansion (6) is essential because in the following the positive and
1948 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008

negative-sequence harmonic terms, instead of the single phase TABLE I

harmonics, play a determinant role. CORRESPONDENCES BETWEEN THE HARMONIC INDEXES OF THE LOAD
CURRENT i (t) AND THE SOURCE CURRENT i(t) ACCORDING TO (21)
For this reason, in the following all the preceding formulas
will be rewritten in space-vector form and the same approach
will be used the subsequent elaboration.

A. First Approach
Before applying the compensation, the instantaneous powers
delivered to the load are, from (2), (8), and (9)

(15)

After the compensation (10), at

the source side and the space vectors and (Fig. 1) are
in phase. The power absorbed from the grid is therefore

(16)

From (15) and (16), the current is found as a function of

the voltage and of the current absorbed by the load

(17)

Observing that for a generic complex quantity ,

, the (17) can be rewritten as
Fig. 3. Geometric interpretation of (21).
(18)

Equation (18) allows to predict the source current, provided produces a new harmonic of index . The amplitudes of
that the load voltage and current are known. both the harmonic terms are half of the original one .
The simple and important case of sinusoidal and balanced Some correspondences between the harmonic indexes of
voltage is now developed. Under these condition, the voltage the load current and the source current are shown in
space vector (4) is Table I.
Observing Table I, the following comments apply:
(19) • the transformation is reciprocal (the indexes of the load
current and of the grid current exchange each other);
Without losing generality, in (19) is assumed to be real • as expected, a negative sequence component
(i.e., in (4)). Substituting (19) in (18) gives rise to a third harmonic;
• the harmonics and are coupled, as already shown
(20) in [9].
1) Geometric Interpretation of (21): Each couple of har-
Let us apply now the Fourier expansion (6) to both the occur- monic terms in (21) can be interpreted according to Fig. 3, where
rences of the load current in (20) vectors of constant amplitude and rotating at different constant
angular frequencies are depicted in the complex plane (vectors
are indicated in bold uppercase).
Let us consider the voltage : if the three phase voltages are
sinusoidal and balanced, the magnitude of is constant and it
rotates with angular frequency . The generic harmonic cur-
(21) rent rotates with angular frequency ; according to (15),
the related components of the powers and
can be obtained graphically as the product of
The first term on the right hand side of (21) is sinusoidal at with the projections of on the and axes, named
fundamental frequency, balanced and in phase with the voltage. and respectively.
The other harmonic terms are split in two components: the first The – compensation requires to cancel , i.e., to cancel
one preserves the frequencies that appear also in the load the component, while is preserved. This result can be
current, the second one shows the frequencies . In other achieved introducing the vector: since rotates with
words, each harmonic of index present in the load current angular frequency respect to the voltage, has
SUPERTI-FURGA AND TODESCHINI: DISCUSSION ON INSTANTANEOUS – STRATEGIES 1949

to rotate with the opposite angular frequency respect Under the conditions expressed in (28), (27) gives
to the voltage , so that his frequency is , according
to (21).
It is important to notice that the magnitude of and of
after the compensation is half of the original magnitude of ,
otherwise the value of the instantaneous active power would
be doubled.

B. Second Approach
(29)
Equation (15) still applies. After the compensation, the
method provides a constant and (Fig. 2).
As a result, the space vectors and are in phase and the Equation (29) shows that the current is the sum of terms with
power absorbed from the grid is different frequencies. The first term in (29) is the first harmonic,
the other terms are the harmonic disturbances. The following
(22) comments apply:
• the amplitude of the second current term is equal to the rel-
The voltage is known. As a result, the current can
ative amplitude of the harmonic component of the voltage,
be easily calculated
the amplitude of the following terms are the powers of that
ratio;
(23)
• the frequencies of the harmonics in the source current are
different respect to the harmonics of the source voltage;
According to (23), the source current is a function of the ac- in order to determine the order of the current harmonic
tual source voltage and the mean power transferred: the load components it is necessary to know both the order and the
current waveform does not appear in (23). sign of the voltage harmonics;
A significant expression for the current can be found under • in many practical cases only the first two terms (funda-
the following hypothesis: the voltage is the sum of a sinu- mental and first harmonic component) are significant, as
soidal balanced term and of a term representing the the amplitude of the subsequent terms falls down rapidly;
harmonic disturbances • the (29) between the order of the second current harmonic
term and of the voltage harmonics is similar to (21) that
(24)
shows the correspondence between the source current and
The (23) can now be rearranged as the load current harmonics.
Some correspondences between the harmonic components of
(25) the source voltage and source current are shown in
Table II.
Observing Table II, the following comments apply:
Under the condition , the following expansion • as it has to be expected, only the index is not altered by
can be applied in the complex domain: the transformation;
• the fundamental negative component in the voltage
(26) produces a third harmonic component in the current,
as reported in many studies;
• the second harmonic of positive sequence in the voltage
Applying (26) to (25), it results in produces a dc component in the current. This situation is
unusual but not impossible, and may be very dangerous in
presence of transformers or saturable inductances;
• the harmonics and exchange each other, at least
for the first term: since typically both are present under
symmetrical distorted conditions, this may be the reason
why this phenomenon is often not detected.
(27) The previous considerations applies only for a single voltage
harmonic added to the sinusoidal component. If more harmonics
In the general case (27) is not more useful than (23). But are present, the general formulation (27) applies. However, if the
in practical cases (27) allows to determine the most significant amplitudes of the disturbances are small, the effect of the each
terms of . harmonic can be found using (29) and the total disturbance is
Let us consider the following situation, where the term obtained applying the superposition principle.
is constituted by a single harmonic with angular frequency ; 1) Geometric Interpretation of (29): The geometric interpre-
as a result (24) becomes: tation of this second approach is less straightforward than the
one proposed in Fig. 3. For sake of simplicity, (28) and only the
(28) first two terms in (29) will be considered.
1950 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008

TABLE II
CORRESPONDENCES BETWEEN THE SOURCE VOLTAGE v(t) AND SOURCE CURRENT i(t) HARMONIC COMPONENTS ACCORDING TO (29)

this subjects. The recent paper [2] carried out a comparison be-
tween different AF control strategies, among which the second
– method, named in [2] “generalized – strategy”, is consid-
ered.
In [2] a four-wire system is assumed; however the results pre-
sented in [2] applies to the three-wire system too, as long as
the zero-sequence component of the voltage is not involved and
the zero-sequence component of the source current is forced to
be zero. On the load side, the effects of an incidental zero-se-
quence component are taken into account provided that in
(14) is the total mean load power comprising contribution from
load zero-sequence components (see second formula in Table I
Fig. 4. Geometric interpretation of (29). of [2]).
Simulation results provided by [2] referring to the –
strategy are now examined. Case and the related Fig. 2 in
Let us consider Fig. 4 (vectors are in bold uppercase, vector [2] confirms that, under sinusoidal and balanced voltage, the
amplitudes are in uppercase). The vectors and source current is sinusoidal and balanced too, regardless of the
are in phase and rotates with the same angular frequency . The load current waveform. The subsequent cases considered in [2]
voltage rotates with angular frequency . The compensa- are summarized in Table III.
tion requires to cancel the oscillations of the related components The first column of Table III refers to the case as numbered
of the powers and . The in [2]. In order to apply (29), a sign has been associated to the
projection of on the and axes are named and harmonic indexes of the source voltage (second column) using
respectively. The product between and results in ; the further information provided in [2]: in cases B and C (balanced
product between and - results in . conditions) the sign is assigned according to (7); in cases D and
In order to compensate these components, the current E (negative sequence at fundamental frequency) the sign is as-
is introduced: the magnitude of this current is . signed according to the general rule (6). In the last two columns
Since rotates with angular frequency respect to of Table III the harmonic indexes and the amplitude of the corre-
the voltage , has to rotate with the opposite angular sponding harmonic components are evaluated according to (29)
frequency respect to the same voltage, so that his fre- and Table II: the obtained results perfectly agree with values re-
quency is , according to (29). The final result is the ported in [2].
compensation of and of the oscillations of . The fol- In cases B and C the sign is associated to the harmonic in-
lowing relations can be easily verified on Fig. 4: dexes of source voltage (the second column) according to bal-
ance conditions and (7). In cases D and E the index is assigned
according to the general rule (6).
In the last two columns of Table III the harmonic indexes and
the amplitude of the corresponding harmonic components are
evaluated according to (29) and Table II: the obtained results
V. DISCUSSION ON SOME RESULTS PRESENTED
perfectly agree with values reported in [2].
IN THE LITERATURE
In order to compare the analytical expression (29) with the
The analytical results presented above agrees with simula- simulation results provided in [2], the variable named ,
tions and experimental results reported in previous papers on in [2, Figs.3–6] is considered. Moreover, in the comparison the
SUPERTI-FURGA AND TODESCHINI: DISCUSSION ON INSTANTANEOUS – STRATEGIES 1951

TABLE III
COMPARISONS BETWEEN THE CASES ANALYZED IN [2] FOR p–q COMPENSATION AND THE EXPECTED HARMONICS OBTAINED BY APPLYING (29)

sign of the harmonic indexes in Table III have to be neglected,

since in [2] the information on the sequence type (positive or
negative sequence) is ignored.

VI. SIMULATION RESULTS

In order to further confirm the analytical expressions pro-
posed in the previous sections, numerical simulations have been
carried out via Matlab-Simulink package. In the following, the
harmonic spectra of the three-phase quantities are shown in an
unusual but very meaningful form where, according to the com-
plex expansion (6), the harmonic frequencies are relative num-
bers. Their signs correspond with the positive or negative se-
quence type respectively. The system shown in Fig. 5 have been
simulated. The Active Filter is implemented with a three phase
VSI. The currents of the AF have been controlled by means of
the hysteresis band method. The parameters values are:
Grid voltage (line-to-line) 400 V;
50 Hz;
0.08 m ;
5 mH; Fig. 5. Simulated system.
0.4 ;
1.7 mH;
4m ; hystersis band 10 A;
0.25 mH; 800 V.
1952 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008

Fig. 6. First p–q compensation strategy, first simulation. Current on the load Fig. 8. First p–q compensation strategy, first simulation. Current on the grid
side: (a) waveforms and (b) spectrum. side: (a) waveforms and (b) spectrum.

Fig. 7. First p–q compensation strategy, first simulation. Instantaneous active Fig. 9. First p–q compensation strategy, second simulation. Current on the load
power (continuous line) and instantaneous reactive power (dotted line) (a) on side: (a) waveforms and (b) spectrum.
the load and (b) on the grid side.

A. First Method
The Clarke transformation is applied to the voltage and load
current, then and are computed by means of (8) and (11).
1) First Simulation. Fifth Harmonic (Negative Sequence) on
the Load: In this simulation the voltage is sinusoidal and bal-
anced, while the first and fifth negative sequence
harmonics are forced in the load current (Fig. 6). Fig. 7 clearly Fig. 10. First p–q compensation strategy, second simulation. Instantaneous ac-
shows that the instantaneous active power is not affected by the tive power (continuous line) and instantaneous reactive power (dotted line) (a)
on the load and (b) on the grid side.
compensation, while the instantaneous reactive power is almost
completely compensated. In those figures actually the instan-
taneous active power appears to be slightly modified and the
instantaneous reactive power is not perfectly zero: this is due
to the fact that the elaborations and the filters are intentionally
not sophisticated, in order to reproduce a real control. Notwith-
standing, the effect of the compensation is significant. Similar
considerations apply for the subsequent simulations results.
According to (21) and to Table I, the negative fifth harmonic
component in the load current is reflected in an additional pos-
itive seventh harmonic component in the source current Fig. 11. First p–q compensation strategy, second simulation. Current on the
Fig. 8. It is worth to notice that the amplitude of the harmonic grid side: (a) waveforms and (b) spectrum.
components of Fig. 8(b) is equal to half the amplitude of the
fifth harmonic component of Fig. 6(b), according to (21) and to
the graphical interpretation shown in Fig. 3. harmonic components of Fig. 11(b) is equal to half the ampli-
2) Second Simulation. Second Harmonic (Positive Sequence) tude of the fifth harmonic component of Fig. 9(b).
on the Load: In this simulation, the first and second 3) Third Simulation. First Harmonic (Negative Sequence) on
positive sequence harmonics are forced in the load current the Load: In this simulation the first positive sequence
(Fig. 9). Fig. 10 shows the instantaneous active power and the and the first negative sequence harmonics are forced in
instantaneous reactive power. the load current (Fig. 12). Fig. 13 shows the instantaneous ac-
According to (21) and to Table I, the positive second har- tive power and the instantaneous reactive power. According to
monic component in the load current is reflected in an addi- (21) and to Table I, the negative first harmonic component in
tional dc component (harmonic index 0) in the source current the load current is reflected in an additional third positive se-
(Fig. 11). This situation would be dangerous, and should be quence harmonic component in the source current Fig. 14.
carefully avoided. It is worth to notice that the amplitude of the Again the amplitude of the harmonic components of Fig. 14(b)
SUPERTI-FURGA AND TODESCHINI: DISCUSSION ON INSTANTANEOUS – STRATEGIES 1953

Fig. 12. First p–q compensation strategy, third simulation. Current on the load Fig. 15. Second p–q compensation strategy. Voltage: (a) waveforms and (b)
side: (a) waveforms and (b) spectrum. spectrum.

Fig. 13. First p–q compensation strategy, third simulation. Instantaneous active
power (continuous line) and instantaneous reactive power (dotted line) (a) on the Fig. 16. Second p–q compensation strategy. Instantaneous active power (con-
load and (b) on the grid side. tinuous line) and instantaneous reactive power (dotted line) (a) on the load and
(b) on the grid side.

Fig. 14. First p–q compensation strategy, third simulation. Current on the grid
side: (a) waveforms and (b) spectrum.
Fig. 17. Second p–q compensation strategy. Current on the grid side: (a) wave-
forms and (b) spectrum.
is equal to half the amplitude of the fifth harmonic component
of Fig. 12(b).
the first current harmonic, first negative component
B. Second Method with amplitude equal to , second negative
component with amplitude equal to ,
As already stated, this method is the most important, but only etc.;
one simulation will be presented since many examples have • the seventh positive voltage harmonic gives rise to:
been considered in Section V. In order to simulate the second fifth negative component with amplitude equal to
– approach, a low pass filter with cut-off frequency 100 rad/s 15% respect to the first current harmonic, eleventh nega-
has been added in series to the block that calculates (Fig. 5). tive component with amplitude equal to
1) First, Second, and Seventh (Positive Sequence) Harmonics , etc.;
in the Voltage: In this simulation, the first, second and seventh • moreover, since more than one voltage harmonic with a
positive sequence harmonics ( , , ) are forced in the significant amplitude are present, the limits of (29), in
grid voltage (Fig. 15). The amplitude of the second and seventh which interactions between harmonics are neglected, are
positive harmonic are 20% and 15% of the amplitude of the evident. Taking into account (27), the remaining terms in
first harmonic, respectively. Fig. 16 shows that the oscillating Fig. 17(b) are recognized. In the considered case we have:
part of the instantaneous active power is compensated, while the . Developing the third term of
instantaneous reactive power on the grid side is nil. According to (27) (squared term) we get
(29) and to Table II, the harmonic components in the voltage are
reflected in several additional components in the source current
(Fig. 17).
In detail, from (29) it results in:
• the second positive voltage harmonic gives rise to:
dc component (index 0) with amplitude equal to 20% of
1954 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008

The first two terms have been already considered in the pre- [2] M. I. M. Montero, E. R. Cardaval, and F. B. Gonzales, “Comparison of
vious points; from the cross product between the voltage control strategies for shunt active power filters in three-phase four-wire
systems,” IEEE Trans. Power Electron., vol. 22, no. 1, pp. 229–236,
harmonics a further sixth negative harmonic appears, Jan. 2007.
with amplitude equal to , in perfect [3] R. S. Herrera and P. Salmeron, “Instantaneous reactive power theory: A
agreement with Fig. 17(b). comparative evaluation of different formulations,” IEEE Trans. Power
Del., vol. 22, no. 1, pp. 595–604, Jan. 2007.
[4] R. S. Herrera, P. Salmeron, and H. Kim, “Instantaneous reactive
power theory applied to active power filter compensation: Different
VII. CONCLUSION approaches, assessment, and experimental results,” IEEE Trans. Ind.
Electron, vol. 55, no. 1, pp. 184–196, Jan. 2008.
[5] G. W. Chang and T. Shee, “A novel reference compensation current
In this paper the drawbacks of both Nabae–Akagi compensa- strategy for shunt active power filter control,” IEEE Trans. Power Del.,
tion methods applied to the three-phase systems under nonsinu- vol. 19, no. 4, pp. 1751–1758, Jul. 2004.
soidal or unbalanced conditions have been analytically proved. [6] H. Akagi, Y. Kanazawa, and A. Nabae, “Instantaneous reactive power
compensators comprising switching devices without energy storage
In particular, the proposed expressions allow to calculate the components,” IEEE Trans. Ind. Appl., vol. IA-20, no. 3, pp. 625–630,
amplitude, the order and the sequence of the harmonic distur- May/Jun. 1984.
bances that affect the source current after the compensation, [7] H. Akagi, A. Nabae, and S. Atoh, “Control strategy of active power
filters using multiple voltage-source pwm converters,” IEEE Trans. Ind.
starting from the knowledge of the harmonics in the load cur- Appl., vol. IA-22, no. 3, pp. 460–465, May/Jun. 1986.
rent (first method) or voltage (second method). [8] H. Akagi, E. H. Watanabe, and M. Aredes, Instantaneous Power
A sinusoidal current is expected to be absorbed from the grid Theory and Applications to Power Conditioning. New York: IEEE
Press, 2007, ch. 3.
after the compensation: as a result, the first method is seldom [9] H. Akagi, E. H. Watanabe, and M. Aredes, Instantaneous Power
used in the applications, since the harmonics in the source cur- Theory and Applications to Power Conditioning. New York: IEEE
rent are caused by the harmonics in the load current: the latter Press, 2007, ch. 3, pp. 64–68.
[10] V. Soares, P. Verdelho, and G. D. Marques, “An instantaneous active
are in general non negligible and may result in unacceptable and reactive current component method for active filters,” IEEE Trans.
distortions on the source side. The second method is more effi- Power Electron., vol. 15, no. 4, pp. 660–669, Jul. 2000.
cient and it is spread: in fact only the harmonics in the source [11] S. Battacharya, D. M. Divan, and B. Banerjee, “Synchronous frame
harmonic isolator using active series filter,” in Proc. EPE’91 4th Euro-
voltage are reflected in the source current (29). Being the source pean Conf. Power Electron. Appl., Firenze, Italy, Sep. 3-6, 1991, vol.
voltage practically sinusoidal and balanced in many situations, 3, pp. 30–35.
[12] D. A. Marshall, F. P. Venter, and J. D. v. Wyk, “An evaluation of the in-
the obtained source current is usually satisfactory too. Never- stantaneous calculation of load current components,” ETEP Eur. Trans.
theless the limits and possible drawbacks of this method have to Elect. Power Eng., vol. 34, no. 1, pp. 53–59, 1993.
be taken into consideration in the applications. [13] J. L. Willems, “Instantaneous sinusoidal and harmonic active and de-
active currents in three-phase power systems,” ETEP Eur. Trans. Elect.
In this paper, the issue of the grid current waveform has been Power Eng., vol. 4, no. 5, pp. 335–343, 1994.
studied in relation with the AFs and the Power Conditioners; [14] A. Nabae and T. Tanaka, “A new definition of instantaneous active-re-
however, the waveform of the current exchanged with the grid active current and power based on instantaneous space vector on polar
coordinates in three-phase circuits,” IEEE Trans. Pow. Del., vol. 11,
depends, after all, only on the actual voltage waveform at the no. 3, pp. 1238–1243, Jul. 1996.
grid interface and on the waveform of the three-phase instanta- [15] A. Nabae and T. Tanaka, “A quasi-instantaneous reactive power com-
neous power that must, or it is desired to, be transferred from pensator for unbalanced and nonsinusoidal three-phase systems,” in
Proc. IEEE 29th Power Electron. Spec. Conf., May 17-22, 1998, vol.
or to the grid. From this point of view, the problems and the re- 1, pp. 823–828.
sults discussed in this paper can be applied also to three-phase [16] L. S. Czarnecki, “On some misinterpretations of the instantaneous re-
front-end converters with a high switching frequency. The re- active power p–q theory,” IEEE Trans. Power Electron, vol. 19, no. 3,
pp. 828–836, May 2004.
sults presented in this paper are strictly confined to three-wire [17] L. S. Czarnecki, “Instantaneous reactive power p–q theory and power
systems. If four-wire systems are considered, the previous re- properties of three-phase systems,” IEEE Trans. Power Del., vol. 21,
sults are still valid if applied to the three-wire part of the system, no. 1, pp. 362–367, Jan. 2006.
[18] H. L. Ginn, III, in Proc. 7th Int. Workshop Power Definition Meas.
that corresponds to the and voltage and current components Nonsinusoid. Cond., Cagliari, Italy, Jul. 10-12, 2006, pp. 91–96.
and to the powers resulting from these components only. If ex- [19] M. Jussila and H. Tuusa, “Comparison of simple control strategies
tensions of – theories are considered, such as modified – of space-vector modulated indirect matrix converter under distorted
supply voltage,” IEEE Trans. Power Electron., vol. 22, no. 1, pp.
and – – approaches, results and drawbacks similar to what 139–148, Jan. 2007.
discussed in this paper are expected. [20] A. Nabae, L. Cao, and T. Tanaka, “An instantaneous distortion cur-
Finally, the paper shows the effectiveness of the space vectors rent compensator without any coordinate transformation,” in Proc. Int.
Power Electron. Conf., Yokohama, Japan, 1995, pp. 1651–1655.
approach in the analysis of the three-phase systems: the use of [21] H. Akagi, S. Ogasawara, and H. Kim, “The theory of instantaneous
the space vectors allows a clear and straightforward represen- power in three-phase four-wire systems: A comprehensive approach,”
tation and interpretation of the results. In particular it allows to in Proc. Conf. IEEE-IAS Annu. Meeting, 1999, pp. 431–439.
[22] M. Depenbrock, V. Staudt, and H. Wrede, “A theoretical investigation
associate a positive or negative sign to each harmonic compo- of original and modified instantaneous power theory applied to four-
nent, that includes the information on the pertaining sequence. wire systems,” IEEE Trans. Ind. Appl., vol. 39, no. 4, pp. 1160–1168,
Jul./Aug. 2003.
[23] X. Dai, G. Liu, and R. Gretsch, “Generalized theory of instantaneous
reactive quantity for multiphase power systems,” IEEE Trans. Power
REFERENCES Del., vol. 19, no. 3, pp. 965–972, Jul. 2004.
[24] J.-C. Montano, P. Salmeron, and J. P. Thomas, “Analysis of power
[1] P. Salmeron and R. S. Herrera, “Distorted and unbalanced systems losses for instantaneous compensation of three-phase four-wire sys-
compensation within instantaneous reactive power framework,” IEEE tems,” IEEE Trans. Power Electron., vol. 20, no. 4, pp. 901–907, Jul.
Trans. Power Del., vol. PD-21, no. 3, pp. 1655–1662, Jul. 2006. 2005.
SUPERTI-FURGA AND TODESCHINI: DISCUSSION ON INSTANTANEOUS – STRATEGIES 1955

[25] H. S. Kim and H. Akagi, “The instantaneous power theory on the ro- Gabrio Superti-Furga (M’89) received the Ph.D.
tating p–q –r reference frams,” in Proc. Conf. IEEE-PEDS, Hong Kong, degree in electrical engineering from the Politecnico
Jul. 1999, pp. 422–427. di Milano, Milan, Italy, in 1972.
[26] H. Kim, F. Blaabjerg, B. Bak-Jensen, and J. Choi, “Instantaneous Since 1991, he has been a Full Professor of
power compensation in three-phase systems by using p–q –r theory,” Electrical Science at the Electrical Engineering
IEEE Trans. Power Electron., vol. 17, no. 5, pp. 701–710, Sep. 2002. Department, Politecnico di Milano. His current re-
[27] H. Kim, F. Blaabjerg, and B. Bak-Jensen, “Spectral analysis of in- search interests are electrical traction drives, control
stantaneous power in single-phase and three-phase systems with use of converters and modulation techniques, distorted
of p–q –r theory,” IEEE Trans. Power Electron., vol. 17, no. 6, pp. and nonlinear power systems, multiphase systems.
711–720, Nov. 2002. Dr. Superti-Furga is a member of the IEEE Indus-
[28] S.-J. Lee, H. Kim, S. K. Sul, and F. Blaabjerg, “A novel control trial Electronics and IEEE Power Electronics Soci-
algorithm for static series compensators by use of p–q –r instanta- eties.
neous power theory,” IEEE Trans. Power Electron., vol. 19, no. 3, pp.
814–827, May 2004.
[29] H. Kim, S.-J. Lee, and S.-K. Sul, “A calculation for the compensation
voltages in dynamic voltage restorers by use of p–q –r power theory,” Grazia Todeschini (S’07) received the M.S. degree
in Proc. Appl. Power Electron. Conf. Expo (APEC’04), 2004, vol. 1, in electrical engineering from the Politecnico di Mi-
pp. 573–579. lano, Milan, Italy, in 2005 and is currently pursuing
[30] G. J. Retter, Matrix and Space-Phasor Theory of Electrical Ma- the Ph.D. degree at the Electrical and Computer Engi-
chines. Budapest, Hungary: Akademiai Kiadò, 1987. neering Department, Worcester Polytechnic Institute,
[31] A. Ferrero and G. Superti-Furga, “A unified approach to unbal- Worcester, MA.
anced three-phase systems under nonsinusoidal conditions: Some Her current research interests are concerned with
definitions,” in Proc. 4th Int. Conf. Harmon. Power Syst., Budapest, distributed generators, renewable energies, power
Hungary, Oct. 4-6, 1990, pp. 32–37. quality, and control systems.
[32] A. Ferrero and G. Superti-Furga, “A new approach to the definition of
power components in three-phase systems under nonsinusoidal condi-
tions,” IEEE Trans. Instrum. Meas., vol. 40, no. 3, pp. 568–577, Jun.
1991.