Abstract— The paper is focused on the presentation of the current and voltage space vectors in the stationary reference
different theories of powers and current decomposition methods frame.
in the compensating current calculation for three-phase three-
wire shunt active filtering systems. The decomposition methods Regardless of theory, this allows the decomposition of the
are grouped into two categories, depending on the type of distorted current generated by a nonlinear load and highlighting
coordinate system used in the processing measurements involved the useless component. That is why the main practical
in calculations. The former is based on the phase coordinates application of these power theories is the shunt active power
system and includes: the theory of Fryze, the theory of Fryze- filtering [19]-[25].
Buchholz-Depenbrock, the so-called generalized theories of
instantaneous reactive and non-active powers, the Currents’
As a main principle, starting from a distorted load phasor
Physical Components Theory (CPC) and the Conservative Power current (iL), the shunt active power filter (SAPF) is able to
Theory (CPT). The latter refers to the orthogonal reference inject such a compensating phasor current (iF) in the point of
frame–based approaches such as the p-q theory and id-iq common coupling (PCC) so that the current drawn from the
method. The applicability of each theoretical approach in the network has the desired shape and zero passing (ides),
control of a shunt active power filter was analyzed. A summary
highlighting the applicability under non-sinusoidal voltage and
the main performance of each method is performed. Next, our i F = i L − i des . (1)
platform for testing the control strategies of shunt power filters,
based on DSP DS1103 system, is presented. An implementation From the point of view of practical applicability, active
on a dSpace 1103 DSP system of the theories presented above and filtering can be of two types: partial filtering and total filtering.
some experimental results are given in the second part of the In the partial filtering case, the supply phase currents are
paper. balanced and sinusoidal or they have the same shapes as the
voltages. In the total filtering case, the supply currents are
Keywords— Active current, DSP, Active filter, Control methods
balanced and sinusoidal or have the same shapes and zero
for electrical systems, Optimal control
crossing as the voltages.
I. INTRODUCTION On the other hand, if the current is sinusoidal, the goal of
filtering is “zero distortion factor” and if the current has the
The time-based description and analysis of complex same shapes and zero crossing as the voltage, the goal of
phenomena in electrical power systems is a long-standing filtering is “unity power factor”. So, the desired current can be
concern of specialists, since the first decades of the 20th the sum of the active and reactive components in the partial
century. In 1931, Fryze proposed the instantaneous non-active filtering case and the active component in the total filtering
power theory for the operation under non-sinusoidal conditions case.
[1].
This paper, first presents the main theories which have
After that, many theories which can be considered as applications in computing the reference compensating current
instantaneous power theories were developed over time. All of a three-phase SAPF. Two, kinds of formulations are
these theories are based either on phase coordinates system as outlined, depending on the type of coordinate system used in
extensions of Fryze’s idea [2]-[8] or on an orthogonal calculations, i.e. phase coordinates system and orthogonal
stationary reference frame as extensions of so-called p-q theory reference frame.
introduced by Nedelcu in 1978 [9], [10] and developed by
Akagi in 1983 [11]-[15], for active filtering application. In the second part, sections III, IV and V are dedicated to a
short presentation of a shunt active system platform developed
A different definition of the instantaneous non-active power by authors in their laboratory.
in the direct-quadrature (d-q) plane was introduced by Ferrero
and Superti-Furga [16], [17]. In [18], the instantaneous active Next, the control implementation on the dSpace 1103 DSP
and reactive components of the current are calculated in the system is presented and some experimental results for both
synchronous reference frame based on the components of balanced and unbalanced loads are illustrated.
G p (t ) = pΣ (t ) / uΣ2 ; (10)
P
ia (t ) = ⋅ u (t ) , in (t ) = i (t ) − ia (t ) (2)
2
u - the instantaneous phase power currents,
ikp = G p (t ) ⋅ uk ; (11)
where ║u║ is the rms value of the voltage. As the ratio
2 - the instantaneous phase zero-power currents,
P u is constant, the active current has the same waveform
as the supply voltage and the same zero-crossing instants. ikz = ik − ikp ; (12)
The mutual orthogonality of ia(t) and in(t) is expressed by:
- the equivalent conductance,
in (t )ia (t )dt = 0 ;
T 2 2 2
∫0 i = ia + in (3) PΣ
G= ; (13)
2
uΣ
Based on the nonactive current, Fryze defined the
nonactive power as: - the phase active currents (Fig. 1),
ika = G ⋅ uk ; (14)
Q = u ⋅ in . (4)
- the phase non-active currents,
Note that the Fryze’s nonactive power is an average value, ikn = ik − ika . (15)
whereas the Fryze’s active and nonactive currents are
instantaneous quantities. Expression (15) provides the current to be compensated in
order to minimize the supply current without modifying the
B. Buchholz’s Extension collective active power.
In order to extend the Fryze’s definitions for multi-phase
D. Generalized Instantaneous Reactive Power Theory
circuits (m phases), in 1950, Buchholz introduced the
collective instantaneous values of voltage and current as [2]: The generalized instantaneous reactive power theory for
three-phase systems was proposed by Peng by defining at first
the power components and then by decomposing the distorted
m m current. [4].
uΣ = ∑ uk2 ; iΣ = ∑ ik2 . (5) Based on the instantaneous space vectors,
k =1 k =1
⎡u a ⎤ ⎡ia ⎤
u = ⎢ub ⎥; i = ⎢⎢ib ⎥⎥ ,
⎢ ⎥
and the associated collective rms values as: (16)
⎢⎣uc ⎥⎦ ⎢⎣ic ⎥⎦
m 2 m 2
uΣ = ∑ uk ; iΣ = ∑ ik . (6) the instantaneous active power is the inner product
k =1 k =1
p = u ⋅ i = u aia + ubib + uc ic , (17)
Accordingly, the apparent power was defined as:
and the instantaneous non-active power vector is the following
cross product:
SΣ = uΣ ⋅ iΣ . (7)
⎡qa ⎤ ⎡u2i3 − u3i2 ⎤
q = u × i = ⎢⎢ qb ⎥⎥ = ⎢⎢ u3i1 − u1i3 ⎥⎥ . (18)
C. Fryze-Buchholz-Depenbrock Theory ⎢⎣ qc ⎥⎦ ⎢⎣ u1i2 − u2i1 ⎥⎦
In 1993, making use of Fryze’s and Buchholz’s definitions,
Depenbrock published the so called Fryze-Buchholz-
Depenbrock (FDB) theory [3]. The following quantities were
introduced:
ua / 50 (V); isa (A)
(a)
iFa (A)
Time (s)
Time (s)
ua / 20 (V); isa (A)
Fig. 2. Supply voltage and active current calculated in accordance with the
(b) generalized instantaneous non-active power theory, under non-sinusoidal
conditions.
P(t )
{
u = [u A u B uC ]T = 2 Re [U A U B U C ]T e jω1t = } , (26)
i p (t ) =
U 2p (t )
u p (t ) ; i q (t ) = i(t ) − i p (t ) , (22) = 2 Re Ue { jω1t
}
and the current components, except the generated current, are
where up(t) is a reference voltage which depends on included in the current fundamental harmonic, i.e.
compensation objective, and
i L1 = i a + ir + iu , (27)
1 ∞
P(t ) = ∫t −Tc p(τ )dτ ;
t
(23) i g = ∑ iLk ; (28)
Tc
k =2
∞
1 i L = i L1 + ∑ iLk . (29)
U p (t ) = (τ )dτ
t 2
∫t −Tc u p . (24) k =2
Tc
By considering the equivalent passive load of delta
It was pointed out that the averaging interval Tc can be zero, structure, with respect to active and reactive powers, whose
one fundamental cycle, one-half cycle, or multiple cycles, equivalent admittance is Ye = Ge + jBe , the current
depending on the compensation purpose [5]-[7]. We must note
that the active currents given by expression (22), have the same components can be expressed as follows [26], [27], [30]:
shapes and phases as the voltages (Fig. 2). Whereas Peng’s
theory was developed only for a system of three phases, a more { }
i a = 2 Re GeUe jω1t ; (30)
2 Re{jB Ue ω };
generalized theory of instantaneous reactive power was given j 1t
by Dai [8]. ir = e (31)
{
iu = 2 Re A[U A U C U B ]T e jω1t . } (32)
x' (t ) =
1T
∫ x' (t )dt . (44)
T 0
Starting from the general definitions of the equivalent
admittance (Ye) and unbalanced admittance (A), convenient In poly-phase systems, the current components (active,
fictitious line-to-line admittances of the load can be expressed reactive and void) can be defined for each phase [33].
in the three-wire system in order to reduce the amount of Active current (Gn = equivalent phase conductance):
calculation [31], [32]:
un , in Pn
Ye = Y AB + YBC + YCA ; (33) ian = un = un = Gn un , n = 1 ÷ N . (45)
2
un U n2
(
A = − YBC + αYCA + α *Y AB ; ) (34)
Reactive current (Bn = equivalent phase susceptance):
α = e j 2π 3 , α * = e − j 2π 3
. (35)
un , in Wn
As the measured line-to-line voltages in the experimental irn = 2 un = U 2 un = Bnu n , n = 1 ÷ N . (46)
setup are uBA and uCA, the phase-A is chosen as a reference and un n
YBC is assumed to be zero. Therefore, the expressions of the
equivalent and unbalanced admittances become: Void current:
i vn = i n − i an − i rn , n = 1 ÷ N . (47)
I I
Ye = YBA + YCA = B + C ; (36)
U BA U CA Active and reactive current terms can also be defined
collectively, i.e. by making reference to an equivalent balanced
( )
⎛ I I ⎞
A = − αYCA + α *YBA = −⎜⎜ α C + α * B ⎟⎟ . (37)
load absorbing the same active power and reactive energy as
the actual load. Balanced active currents (minimum collective
⎝ U CA U BA ⎠
currents needed to convey the active power P corresponding to
the equivalent balanced conductance - Gb):
Then, the equivalent conductance (Ge) and susceptance Be)
of the load are calculated as: u, i P
i ba = 2
u= u = Gb u . (48)
u U2
Ge = Re{Ye } ; Be = Im{Ye } . (38)
i = i a + i r + i v = i ba + i ua + i br + i ur + i as + i rs + i g ; (53)
- active currents are the sum of balanced active currents a p (t ) = u(t ) ⋅ i* (t ) , (58)
b u
(i a ) and unbalanced active currents ( i ); a the instantaneous real and imaginary powers were introduced
- reactive currents are the sum of balanced reactive currents as follows:
( i br ) and unbalanced reactive currents ( i ur ); { }
p p (t ) = Re al a p (t ) = ud id + uqiq ;
- void currents are the sum of scattered active currents q p (t ) = Im ag {a p (t )}= uq id − ud iq .
s s
(59)
( i a ), scattered reactive currents ( i r ) and load-generated
harmonic currents ( i g ). If expressions (55) and (59) are compared, it can be seen
that the instantaneous imaginary power in (59) is the minus of
expression of q in (55). However, both of them were used in
III. ORTHOGONAL REFERENCE FRAME–BASED the compensating currents calculation. As the authors have
METHODS chosen the non-power invariant transformation from a-b-c
coordinates system to α-β reference frame, the instantaneous
Since its proposal, at the beginning of the 1980s, the
complex power is expressed as:
Akagi’s instantaneous reactive power theory, known as the p-q
theory, has been applied in the control of three-phase SAPFs. s = 3 / 2 ⋅ u ⋅ i* = p + jq , (60)
A. p-q Theory and the current space vector can be expressed as,
In 1983-1984, professor Akagi and his coauthors proposed
the p-q theory for three-phase circuits [11], [12], [34]-[36]. 2 u
i= ⋅ ⋅ s* , (61)
Compared with the previous theories of instantaneous reactive 3 u2
power, the instantaneous active and reactive powers were
expressed by using the components of voltage (uα, uβ) and 2
current (iα, iβ) vectors in the stationary reference frame α-β. The u = uα2 + u β2 . (62)
Clarke’s power invariant transformation was applied to
transform the three-phase coordinates quantities to the α-β two- If the DC components (P and Q) and the AC components
phase equivalent quantities, i.e. (p~ and q~) of the instantaneous powers p and q are taken into
account, the following current vectors can be defined for the
2⎡ 1 −1 2 −1 2 ⎤ operation under sinusoidal voltage conditions [24]:
T= ⎢ ⎥. (54) - the active current vector,
3⎣ 0 3 2 − 3 2⎦
2 P
ia = ⋅ u; (63)
Thus, the conventional instantaneous power (p) and the 3 u2
instantaneous imaginary power (q) were introduced as:
- the reactive current vector,
p = uα iα + u β iβ ; q = uα iβ − u β iα . 2 Q
(55) i r = − ⋅ 2 ju ; (64)
3 u
To give a physical meaning to the instantaneous power
- the supplementary useless current vector on account of p~,
components, the components of the current were expressed as a
2 p
function of p and q: i sp = ⋅ ~2 u ; (65)
uβ 3 u
uα
iα = p− q;
2 2
uα + u β uα + u β2
2
- the supplementary useless current vector on account of q~,
. (56)
uβ uα 2 q
iβ = 2 p+ 2 q. i sq = − ⋅ ~2 j u ; (66)
uα + u β2 uα + u β2 3 u
It was pointed out that, in order to eliminate the
instantaneous reactive power at the power supply side, the - the total supplementary useless current vector,
instantaneous compensating components of the current must 2 ( p~ − jq~ )
be: is = ⋅ u. (67)
3 2
uβ u u
iCα = 2 2
q; iCβ = − 2 α 2 q.
uα + u β uα + u β The above detailed decomposition allows expressing the
(57)
compensating current for any compensation objective. A
Another approach in defining the instantaneous powers was mathematical inconsequence of the original p-q theory under
based on the use of the direct-quadrature plane (d-q) instead of non-sinusoidal voltage was pointed out by authors [24], [25],
(α-β) plane [16], [17]. After defining the instantaneous [37] – [39], mainly related to the fact that the active current
complex power as
given by (63) has not the same shape as the voltage, under B. id-iq Method
nonsinusoidal voltage conditions (Fig. 3). As an alternative of the current decomposition based on the
In order to obtain an active current whose waveform has the original p-q theory in the stationary reference frame α-β, Soares
same shape as the supply voltage, in accordance with Fryze and his coauthors proposed to calculate the id and iq
definition, the denominator in (63) must be constant and equal components in the synchronous reference frame d-q [18]. The
to the rms value of the voltage vector modulus [25], axis d follows the direction of the voltage space vector, so that
the rotation angle (θ) can be expressed as a function of voltage
1 t 2 components in α-β reference frame, i.e.
U=
T
∫t −T
u dt . (68)
uβ u
sin (θ ) = ; cos(θ ) = α . (71)
So, the expression of the true active current becomes u u
30
20
active filtering system composed of a two-level VSI which is
10
connected to PCC through an inductive coupling filter to
0
prevent the high order switching harmonics from propagating
-10
into the power supply, an inductive distorted current source and
-20 an industrial PC. The VSI inverter based on SKM100GB123D
-30 IGBTs power modules (IC=100 A, VCES=1200 V), having a
0.09 0.095 0.1 0.105 0.11 0.115 0.12 DC-capacitor of 1100 µF and an interfacing reactor of 4.4 mH,
Time (s) acts as SAPF to generate the compensating currents. The line-
to-line supply voltage is 380 V rms and the apparent power of
Fig. 3. Distorted supply voltage and active current calculated by Akagi’s VSI is 15 kVA.
relation.
Two types of distorted balanced / unbalanced loads were
used, i.e. an a.c. voltage controller which supplies a three phase
Curent [A]; Voltage /10 [V]
30
uA inductance (the resistance is very low) and a full controlled
20 three phase bridge rectifier. The first load is the dedicated
10 inductive distorted current source and it has some important
0
iLaA facilities:
-10 - The current can be high distorted (THD over 100%);
-20
- The control angles of the each phase group of thyristors can
-30
be modified independently, so that the load can become
unbalanced;
0.09 0.095 0.1 0.105 0.11 0.115 0.12
Time (s)
- It allows dynamic modification of the load current, by
Fig. 4. Distorted supply voltage and active current calculated by the sinusoidal modulation of the control angle with settable
proposed expression. frequency.
a isa
b isb
Nonlinear
c isc load
ua iLa
ub iLb
uc iLc
iFu shape generation Optimal dc voltage
calculation Reference current
calculation
x u*DC iFCa iFCb iFCc
x
Voltage +
x
controller -
iFua +
Σ - Current
iFub +
Σi - + controller
Fuc
Σ -
iFa
iFb
iFc
L2 L1
uDC
C
Cf
It has the disadvantage that the current is almost purely divided by its amplitude. However, when the voltage supply is
reactive and, when total filtering is achieved, the remained distorted, the denominator’s ratio is not clearly defined and it
current network is very small. The acquisition system based on can become a source of errors. Another strategy is to monitor
LEM sensors measures two line-to-line supply voltages, two the voltage zero-crossing and then to generate the appropriate
load line currents, two inverter line currents and the DC-link sinusoidal signal.
voltage. The industrial PC is equipped with a dSPACE 1103 The third possibility is to use specific phase-locked loop
DSP board which is used for the monitoring of entire SAPF (PLL) techniques and circuits [36], [41], [42]. The synchronous
system. The control board is equipped with a PowerPC 750GX reference frame-based PLL introduced in [31] uses only one
processor running at CPU clock of 1 GHz for fast floating- phase voltage and the PLL control loop does not make use of a
point calculation. PI controller. In [42], a multiple-complex coefficient-filter-
based synchronization technique is used to estimate the
B. Control system fundamental positive and negative components of the distorted
The cascaded control loops, which include the optimal DC- and unbalanced supply voltages. In the adopted control system,
link voltage loop outside the inner current loop, are first the circuit providing the shape of the additional reference
designed in a Matlab/Simulink model. The dSPACE Real- current has the structure shown in Fig. 6.
Time Interface (RTI) together with Real-Time Workshop
(RTW) automatically generate real-time code. Thus, through
the interface between Simulink and DSP, the controller board
is fully programmable from the Simulink block diagram
environment. To obtain high switching frequencies, the
programmable digital I/O channels are used to generate the
required six IGBTs’ gate signals. In addition of the computed
reference current in accordance with the methods discussed
above, an additional component (iFu) is required to cope for
losses in the power circuit and keep the DC-capacitor voltage
at its set value (Fig. 5). A PI controller was chosen to generate
the amplitude of this additional reference current, whereas a
specific circuit provides its shape (uFu) based on the supply
voltage. There are several ways to obtain this signal. The
simplest one could be to use the measured voltage signal Fig. 6. PLL circuit configuration
A fictitious power (pf) is calculated on the basis of real [44]. The implementation of a specific control system for
voltages and two feedback currents generated by sine functions an optimal prescribed DC-voltage is originated by extensive
at the output of a PI controller after integration. The cancelling analysis and experimental results on the active filtering system,
of pf guarantees the synchronized output currents of the PLL when the coupling interface and DC-storage circuit are well
circuit [36]. defined. It has be pointed out that, for each value of the
apparent power to be compensated, there is an optimal value of
The resulting reference current is accurately tracked by DC-voltage which minimizes the total harmonic distortion
using a hysteresis-band current control whose main advantages factor of the supply current after compensation (Fig. 8) [45].
are related to simple hardware implementation, quick current The authors have found an appropriate 4th degree polynomial
controllability and robustness under load parameters variation function for the optimal DC-voltage (UDCo) curve fitting, which
[9], [28]. For the compensating current calculation, a lot of the is defined as follows,
models were build (corresponding to each method).
V. OPTIMAL DC-VOLTAGE CONTROLLER DESIGN U DCo = −0.0073 ⋅ SC4 + 0.32 ⋅ SC3 − 5.5 ⋅ SC2 +
. (77)
The cascade control system is composed of two control loops + 48.17 ⋅ SC + 460.3
with the voltage loop outside the inner current loop (Fig. 7). It The compensating apparent power (kVA) that appears in
must calculate the reference current for each phase of VSI, (77) is,
maintain the DC-bus voltage constant and generate the inverter
2
gating signals. The two essential parameters to be controlled SC = S Load − P2 . (78)
are the inverter output current and the DC-bus voltage at the
inverter input from the viewpoint of active power balance. A PI Thus, only the supply voltages and load currents are needed
controller is adopted to control the voltage across the capacitor. to calculate the optimal DC-voltage [43].
Assuming that the equivalent transfer function of the current
loop is one (the current controller is of hysteresis band type), VI. CONTROL IMPLEMENTATION ON DSPACE 1103
the voltage controller is tuned according to the Modulus SYSTEM
Optimum (MO) criterion for an efficient disturbance rejection
To perform the real-time control of the active filtering
[43], [44]. In addition, the passband frequency (fp) of the unity
system, the control algorithm previously described has been
feedback system must be imposed. If the transfer function of
built under Matlab/Simulink environment combined with the
the voltage controller is written as
RTI and RTW tools provided by dSPACE 1103 system (Fig.
9). After normalizing, the digital inputs supplied by ADC
1 + θ1V ⋅ s
GVC (s ) = . (74) blocks are used according to the adopted control strategy. Since
θV ⋅ s the hysteresis control has been chosen, the three-phase SLAVE
DSP PWM block cannot be used for the IGBT’s gating signals.
If KVT and KCT are the proportional constants associated to the Consequently, two options remain for gating signals transfer to
the IGBT’s drivers, that is either through digital to analog
voltage and current transducers and Us is the rms value of the
(D/A) channels or through digital output channels. A detailed
phase voltage, the following expressions can be used to experimental analysis on the analog outputs signals shown that
calculate the two time constants [44]: the accurate transfer through D/A channels is guaranteed for
0.36 signal frequencies up to 3.5 kHz. Therefore, the generated
θ1V = . (75) switching signals are taken out of the DS1103 with the help of
fp six digital outputs through the DS1103BIT_OUT block of
Master PPC library. A specific block has been created to
3 ⋅ KVT ⋅ U s 1 control the start-up process of the shunt active power filter and
θV = 0.6495 ⋅ ⋅ 2 . (76)
K CT ⋅ C ⋅ U DC f p the associated DC-capacitor charging. Two digital to analog
converters are used to control two line-contactors (named K1
A passband frequency of 20 Hz, which leads to θ1V = 0.018
s and θV = 0.0043 s, was taken into consideration to obtain a
good voltage response without worsening the current response
and K2) which allow a two-stage process of the DC-capacitor The high order harmonics corresponding to the switching
charging. In addition, some protections were taken into frequency (about 20 kHz) can be seen.
consideration and validation conditions were used to avoid
unexpected behaviors during the system operation. As far as B. Results by CPC method
the sampling time is concerned, it was reduced as much as Now, the load is an AC voltage controller that supplies an
possible without reaching a critical value associated to overrun unbalanced three-phase inductive load (Fig. 12). The voltage is
errors. Although a sampling time of 20µs allowed the sinusoidal too, but the currents are high distorted (over 100%)
implementation of all control strategies taken into and unbalanced. The compensating current corresponds to the
consideration, a decrease of 10 % was possible by reducing the unbalanced current and to the load generated component. After
amount of calculation when using expression (15) to calculate compensation, the line currents are balanced and sinusoidal
the reference compensating current. (Fig. 13). The current is not in phase with the corresponding
voltage because its reactive component is not compensated.
VII. EXPERIMENTAL RESULTS 40 400
Using the experimental platform described in sections III,
IV and V, all the methods presented in section II, have been
implemented. Next, some results obtained by three of the 20 200
[A]
u [V]
i ,i ,i
0 0
a
current;
Lc
u [V]
i ,i ,i
0 0
- Figures 15 and 17 correspond to the improved p-q method
La Lb
a
(the expressions proposed by authors);
-20 -200
- It is clear that the phase current and the phase voltage have
the same waveform; Moreover, in the total compensation
-40 -400 case (Fig. 17), the zero crossings of the supply current and
0 0.01 0.02 0.03 t [s] voltage are identical.
Fig. 11. The three-phase controlled rectifier current powered by an 40
unsymmetrical voltages system (CPT) iL
u/10
20 is
uA iLA uB iLB uC iLC iF
-20
-40
0.03 0.035 0.04 0.045 0.05
Fig. 14. Supply voltage - black, load current – red, active filter current - green
and line current after compensation – bleu, on one phase, obtained by the
classic p-q method and partial compensation
Time (s)
40
Fig. 12. Supply voltages and load currents drawn by the AC voltage controller u/10 iL
(unbalanced)
20 is
iF
uA isA uB isB uC isC
0
Supply voltages (V)
Supply currents (A)
-20
-40
0.03 0.035 0.04 0.045 0.05
Fig. 15. Supply voltage - black, load current – red, active filter current - green
and line current after compensation – bleu, on one phase, obtained by the
improved p-q method and partial compensation
Time (s) 40
is
Fig. 13. Supply voltages and currents after compensation (CPC –unbalanced) u/10
20
C. Results by p-q method iL
In this case, the both types of load have been used under
0
low distorted voltage conditions (about 5%). For the AC
voltage controller that supplies a balanced three-phased
inductive load, the goal has been harmonics compensation. -20 iF
Opposite, for the three-phase controlled rectifier balanced load,
the full compensation has been imposed. In all the cases, very -40
good performances of the active filter system are achieved (the 0.03 0.035 0.04 0.045 0.05 0.055 0.06