Electrical Power Quality and Utilisation, Journal Vol. XII, No.

2, 2006

Unified Power Quality Conditioner

in a Grid Connected Photovoltaic System
Marcelo C. CAVALCANTI, Gustavo M.S. AZEVEDO,
Bruno A. AMARAL, and Francisco A.S. NEVES
Federal University of Pernambuco, Brazile

Summary: This paper presents a system that provides photovoltaic generation as well as the Key words:
functions of a unified power quality conditioner. The system can be controlled for current control systems,
harmonics and reactive power compensation simultaneously by using a converter operating as converters,
active shunt filter. The other converter is used as active series filter and it compensates voltage energy conversion,
harmonics or voltage sags and swells. Using only an inverter in photovoltaic energy conversion photovoltaic power
process, the system presents increased efficiency when compared to the conventional systems. systems,
The synchronous reference frame method is used to control the three-phase converters. power quality,
Simulation results demonstrate the good performance of the proposed configuration. Experimental solar energy
results corresponding to the operation of the series filter as voltage sag compensator are
presented.

1. INTRODUCTION reactive power compensation, flux control of active power

and voltage regulation. In this case, it is called Unified Power
Photovoltaic (PV) energy has great potential to supply Flow Controller (UPFC) [6]-[7].
energy with minimum impact on the environment, since it is Conventionally, grid connected photovoltaic energy
clean and pollution free [1]. A large number of solar cells conversion systems are composed of a dc-dc converter and
connected in series and parallel set up the photovoltaic or an inverter [1]-[2]. The dc-dc converter is controlled to track
solar arrays. One way of using photovoltaic energy is in a the maximum power point of the photovoltaic array and the
distributed energy system as a peaking power source [2]. inverter is controlled to produce current in such a way that
On the other hand, strict regulations have been applied to the system current has low total harmonic distortion (THD)
the equipment connected to the utility lines. Some of these and it is in phase with the utility voltage. The efficiency of
regulations are related to harmonics distortion and power the conventional system is low because the dc-dc converter
factor. However, with the development of power electronics, and the inverter are connected in series. The purpose of this
much equipment tends to increase the levels of harmonic paper is to design a photovoltaic generation system for
distortion. The line current at the input to the diode bridge connection in a three-phase system using only a dc/ac
rectifier deviates significantly from a sinusoidal waveform inverter. The proposed system increases the conversion
and this distorted current can also lead to distortion in the efficiency and also provides useful function any time,
line voltage. Moreover, much modern equipment uses digital operating as power supply as well as harmonic and reactive
controllers, based on microprocessors sensitive to variations power compensator when the sun is available. At low
in the voltage and current waveforms. Therefore, to increase irradiation, the system operates only as harmonic and reactive
the PV system utilization the power conversion can be power compensator. Other dc/ac converter is used to provide
designed to also provide functions of a unified power quality voltage harmonic compensation. Cost estimation shows that
conditioner. the use of additional components increases the cost in less
The utilization of two dc/ac fully controlled converters than 12% to have another function to improve power quality.
makes the system have the most versatile structure of Also this converter does not change the efficiency of the PV
converters applied as energy conditioner. In this case, energy conversion since the converters are connected in
depending on the controller, the converters can have different parallel. The control was implemented with the synchronous
functions of compensation. For instance, they can realize reference frame (SRF) method. The system and controller
active series and shunt filters combined to compensate were design and simulated. Different pulse-width-modulation
simultaneously load current and harmonics of the supplied (PWM) techniques have been compared to suggest a
voltage [3]. In this way, the equipment is called Unified Power configuration with optimal efficiency. The system provides
Quality Conditioner (UPQC) [4]. approximately 2.8 kW of photovoltaic generation.
An active shunt filter is a suitable device for current-based
compensation [5]. This configuration includes current
harmonics and reactive power compensations. The active 2. GRID CONNECTED PHOTOVOLTAIC SYSTEM
shunt filter can also balance unbalancing currents.
The active series filter is normally used for voltage-based The proposed photovoltaic (PV) energy conversion
compensation [5]. In this case, voltage harmonics and voltage system has high efficiency, low cost and high functionality.
sags and dips are compensated. Figure 1 shows the block diagram of the proposed system.
Other applications can be found in literature for purposes The converter 1 (PV converter) in Figure 1 is responsible to
of compensation of the fundamental frequency, such as convert the PV energy to the grid as well as to compensate

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System #'
 For instance, they can realize active series and shunt filters
combined to compensate simultaneously load current and
harmonics of the supplied voltage.

3. MAXIMUM POWER POINT TRACKING

The controller of converter 1 in Figure 1 has to track the

maximum power point of the photovoltaic array as well as to
compensate harmonic and reactive power. When the system
is operating as photovoltaic energy generator, the maximum
power point tracking (MPPT) controller is used to calculate
the reference voltage. When the system is operating only as
harmonic and reactive power compensator, the reference
Fig. 1. Proposed system: PV generation with UPQC function voltage is constant [1].
It is important to operate the photovoltaic system near the
maximum power point to increase the efficiency of
photovoltaic arrays. A MPPT method often used is the
perturbation and observation method [9]. However, in this
paper, it is used the slope of power versus voltage, which
decreases the oscillation problem and it is easy to implement
[1]. The output power of the PV array and the differential of
the output power to the output voltage can be expressed as:

P =V ⋅I (1)

dP ≅ I + DI ⋅ V
(2)
dV DV

where:
P — PV array output power.
V — PV array output voltage.
Fig. 2. Characteristic diagram of the solar array
I — PV array output current.
DI — Increment of the PV array output current.
DV — Increment of the PV array output current.
In the method, (2) is used as the index of the maximum
power point tracking operation (Fig. 2), where S is the solar
irradiation. When dP/dV<0, decreasing the reference voltage
forces dP/dV to approach zero; when dP/dV>0, increasing
the reference voltage forces dP/dV to approach zero; when
dP/dV=0, reference voltage does not need any change [1].
A voltage sag is simulated between 0.02 and 0.12s and the
reference voltage generated by the MPPT algorithm has small
oscillation around the ideal voltage at the beginning and the
end of the short circuit (Fig. 3). The PV array voltage stays
near to the ideal voltage that is around 121V. Even during the
voltage sag, the MPPT algorithm presents very good results
with at least 99% of the PV maximum output power.

Fig. 3. MPPT controller: solar array output voltage 4. CURRENT BASED COMPENSATION

Figure 4 shows the converter 1 connected to the grid. The

current harmonics and reactive power. The converter 2 converter is responsible to convert the PV energy to the grid
(Dynamic Voltage Restorer – DVR converter) in Figure 1 is as well as to compensate current harmonics and reactive
responsible to compensate voltage harmonics or voltage power. Under balanced operating conditions, it is possible to
sags. The utilization of two controlled converters makes the express the inverter phase output voltages in terms of the
system to have the most versatile structure applied as energy inverter output voltages with respect to the negative dc bus:
conditioner. In this case, depending on the controller, the
converters can have different functions of compensation. vkn = vkN − vnN k = a,b,c (3)

$ Power Quality and Utilization, Journal • Vol. XII, No 2, 2006

Each phase voltage can be written as:

diCk
vkn = vSk − Lc (4)
dt

In a three-phase, three wire load:

vnN = 1 (vaN + vbN + vcN ) (5)

3

Substituting vnN from (5) into (3):

v1n = 2 vaN − 1 vbN − 1 vcN (6)

3 3 3

Similar equations can be written for phase b and c voltages.

The phase voltages can be also written as:

Fig. 4. Converter 1 connected to the grid

van   2 / 3 −1 / 3 −1 / 3  T1 
     
 vbn  = V ⋅  −1 / 3 2 / 3 −1 / 3 ⋅ T2  (7)
 vcn   −1 / 3 −1 / 3 2 / 3  T3 

where the variables Tk represent the states of the inverter iCd 

 
upper switches. Tk is 0 for opened switches and 1 for closed  iCq  =
switches. Defining dk as swithing state function [8]: i 
 C0  (11)

da   2 −1 −1 Ta   cos wt cos ( wt − 2 p / 3) cos ( wt + 2 p / 3)  iCa 

 d  = 1 ⋅  −1 2 −1 T  2  − sin wt 
 b 3   b = − sin ( wt − 2 p / 3) − sin ( wt + 2 p / 3) ⋅ iCb 
(8) 3
 dc   −1 −1 2  Tc   1/ 2  i 
 1/ 2 1/ 2   Cc 

On the dc side:
where:
iCd, iCq — D-axis and q-axis converter currents,
dV = idc = 1 T ⋅ i + T ⋅ i + T ⋅ i
dt
(
C C a Ca b Cb c Cc
) (9) i C0 — zero-sequence converter current,
w — supply angular frequency.
The complete model of the system (Fig. 4) in the abc referential Taking into account the absence of the zero-sequence
is shown in (10) [8]: components in the currents in a three-wire system, the
simplified transformation matrix can be used:

 da 
iCd   cos ( wt − p / 6 ) sin wt  iCa 
 0 0 −
LC   = 2 ⋅  (12)
 − sin ( wt − p / 6 ) cos wt  iCb 
 Ca 
i   iCa   vS1 
d i  =  db    1    iCq 
0 0 − ⋅ i + v (10)
dt  Cb   LC   Cb  LC  S 2 
 V    V    The model in the dq frame is as in (13) [8]:
    0
 2 d a + db d a + 2 db
 C 0 

 C
 dd 
 0 w −
where: LC 
iCd    i  v 
iCa, iCb, iCc — three-phase converter 1 currents, d i  =  − w d q   Cd  1  Sd 
− ⋅ i +
dt  Cq   LC   Cq  LC  Sq  (13)
LC — inductance of the transformer, 0 v
da, db, dc — three-phase switching state functions, V  V   0 
  d dq     
C — capacitance of the dc link,  d 0 
vSa, vSb, vSc — three-phase supply voltages.  C C 
In (10), the steady state fundamental components are
sinusoidal. To reduce control complexity, the dq frame in (11)
where:
rotating at the supply frequency can be used. With this frame,
dd, dq — D-axis and q-axis switching state functions,
the positive-sequence components at fundamental frequency
vSd, vSq — D-axis and q-axis supply voltages.
become constant [8]:

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System $
The current equations in the model (13) can be written as: where:
ud — D-axis output of the current PI compensator.
diCd uq — Q-axis output of the current PI compensator.
LC = Lc ⋅ w ⋅ iCq − V ⋅ d d + vSd (14) The voltage equation in the model (13) can be written as:
dt

C dV = d d ⋅ iCd + d q ⋅ iCq (20)

diCq dt
LC = − Lc ⋅ w ⋅ iCd − V ⋅ d q + vSq (15)
dt
Defining:
Defining:
u pv = dd ⋅ iCd + dq ⋅ iCq (21)
ud = Lc ⋅ w ⋅ iCq − V ⋅ d d + vSd (16)
and considering that the voltage control is realized by using
a PI compensator, the equation (22) in Figure 5 is:
uq = − Lc ⋅ w ⋅ iCd − V ⋅ dq + vSq (17)
idcr = 2 V u
3 Vsup pv (22)
and considering that the current control is realized by using
PI compensators, the equations (18) and (19) in Figure 5 are:
where:
vSd + LC ⋅ w ⋅ iCq − ud u pv — output of the voltage PI compensator,
dd = (18) Vsup — amplitude of the supply voltage.
V
Figure 5 shows the controller block diagram of converter 1
in Figure 1. The phase-locked-loop (PLL) circuit detects the
vSq − LC ⋅ w ⋅ iCd − uq amplitude and the position of the supply voltage vector.
dq = (19) When the system is operating as photovoltaic energy
V
generator, the MPPT controller is used to calculate the

Fig. 5. Control of the proposed system: PV generation / current harmonic and reactive power compensations

$ Power Quality and Utilization, Journal • Vol. XII, No 2, 2006

Fig. 6. Tracking performance of the d-axis current loop Fig. 7. Tracking performance of the q-axis current loop

reference voltage. When the system is operating only as

current harmonic and reactive power compensator, the
reference voltage is constant [1]. In the SRF, the fundamental
positive-sequence components of the load appear as dc
quantities [8]. Since the grid currents must have a sinusoidal
waveform and be in phase with the grid voltage in the
proposed design, the grid current has only d-axis component.
Therefore, the q-axis reference current for the inverter is the
q-axis load current [2]. The d-axis reference current is
composed of three parts: reference d-axis load current (Idhr),
reference dc link current (Idcr) and reference photovoltaic
array current (Ipvr).
The d-axis load current is passed through a low pass filter
that removes the high frequency components in the d–q
reference frame. A first order low pass filter with a cut Fig. 8. Current harmonics and reactive power compensations
frequency of 20Hz is used. Subtracting the d-axis load current
of the filtered d-axis load current, the result is the negative of
the d-axis harmonics. This value is used as the reference d-
axis load current since the positive current flows into the
inverter. The dc component in the d–q reference frame
corresponds to the fundamental component of the real power
flowing to the load. The inverter dc link voltage controller
calculates the current to maintain the dc link voltage by
passing the dc link voltage error through a PI compensator.
The current obtained by the MPPT controller corresponds
to the real power available from the photovoltaic array and it
is subtracted from the current components to maintain the dc
link voltage and to compensate harmonics.
Using d–q reference frame, the coupled dynamics of the
current tracking problem have been transformed into
decoupled dynamics. By adjusting PI compensators, a fast
tracking and zero steady state errors can be achieved. The Fig. 9. Current harmonics for the load
results (Fig. 6 and Fig. 7) show that the oscillating current
harmonics injected by the inverter track their references with fundamental current is used as magnitude 1 and it is not
high accuracy even during the voltage sag. The dc link current shown to allow that the harmonics can be analyzed. In Figure
increases to keep the maximum power in PV array. 10, the supply fundamental current is used as magnitude 1.
The controller can compensate harmonic and reactive
power effectively as shown in Figure 8. Figure 8 shows that 5. VOLTAGE BASED COMPENSATION
the supply current is displaced by 180 degrees because it is
generated more power from the PV array than that needed for Figure 11 shows the converter 2 in Figure 1 connected to
the load. Therefore, in this case the PV array supplies the the grid. The voltage compensator is a system based on
load and injects power for the grid. In Figure 9, the load power electronics that detects the feeder voltage and in case

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System $!
 v0 N 2 = 1 (vaN 2 + vbN 2 + vcN 2 ) (25)
3

The variables Tk2 represent the states of the converter 2

upper switches. Tk2 is 0 for opened switches and 1 for closed
switches. Defining dk2 as swithing state function:

d a 2   2 −1 −1 Ta 2 
  = 1 ⋅ −  
 db 2  3  1 2 −1 Tb 2  (26)
 dc 2   −1 −1 2  Tc 2 

In the filter output voltage:

Fig. 10. Current harmonics for the supply

dvK 0
dt Cf
(
= − 1 i fk − iLk ) (27)

The complete model of the system (Fig. 11) in the abc

referential is shown in (28):

 0  da 2 
0 1 0  − L V 
 Lf 
 f

 i fa   1 
 i 
fa  db 2 
   0  
L f   i fb   − L f V 
0 0
d  i fb  =  ⋅  +  
dt v   1 (28)
 0a  − C 0 0 0  v0 a   1 iLa 
 v0b      C
  v0b   f 
f

 0 − 1 0 0   1 i 
 Cf   C Lb 
 f 

where:
ifa, ifb, ifc — three-phase converter 2 currents,
Lf — inductance of the filter,
da2, db2, dc2 — three-phase switching state functions,
Fig. 11. Converter 2 connected to the grid Cf — capacitance of the filter,
v0a, v0b, v0c — three-phase filter output voltages.
In (28), the steady state fundamental components are
sinusoidal. To reduce control complexity, the dq frame rotating
the voltage is different of the desired voltage, it supplies the
at the supply frequency can be used. Taking into account
necessary voltage to compensate the voltage error. It can be
the absence of the zero-sequence components in a three-
used to compensate voltage harmonics at the point of
wire system, the model in the dq frame is as in (29):
common coupling or voltage sags, keeping the load voltage
around its rated value.
In this paper, it is presented the voltage sag compensation
 0  dd 2 
control. Its use is more justified when many sensitive loads w 1 0  − L V 
 Lf 
are connected to the same feeder [10].  i   
f
 i fd    dq2 
   −w
Under balanced operating conditions, it is possible to 1  fd
 
L f   i fq   − L V 
0 0
express the inverter phase output voltages in terms of the d  i fq  =  ⋅  +  f 
inverter output voltages with respect to the negative dc bus: dt v0 d   1  
  − 0 0 w  v0 d   1 i  (29)
 v0 q   C f  v  Ld
C
vk 0 = vkN 2 − v0 N 2 k = a,b,c (23)    0q   f 
 0 − 1 −w 0   1 i 
 Cf   C Lq 
Each phase voltage can be written as:  f 

di fk where:
vk 0 = vK 0 − L f (24)
dt dd2, dq2 — D-axis and q-axis switching state functions.
v0d, v0q — D-axis and q-axis filter output voltages.
In a three-phase, three wire load:

$" Power Quality and Utilization, Journal • Vol. XII, No 2, 2006

Fig. 12. Control of the proposed system: PV generation / current harmonic and reactive power compensations

The current equations in the model (29) can be written as: The voltage equations in the model (29) can be written as:

di fd dv0d
Lf = L f ⋅ w ⋅ i fq + v0 d − V ⋅ d d 2 (30) Cf = −i fd + C f ⋅ w ⋅ v0q + iLd (36)
dt dt

di fq dv0 q
Lf = − L f ⋅ w ⋅ i fd + v0 q − V ⋅ d q 2 (31) Cf = −i fq − C f ⋅ w ⋅ v0 d + iLq (37)
dt dt

Therefore:
Defining:
i fdr = iLd + C f ⋅ w ⋅ v0q − uvd 2 (38)
ud 2 = L f ⋅ w ⋅ i fq + v0d − V ⋅ d d 2 (32)

i fqr = iLq − C f ⋅ w ⋅ v0d − uvq 2 (39)

uq 2 = − L f ⋅ w ⋅ i fd + v0q − V ⋅ dq 2 (33)
where:
and considering that the current control is realized by using u vd2 — D-axis output of the voltage PI compensator,
PI compensators, the equations (34) and (35) in Figure 12 are: u vq2 — Q-axis output of the voltage PI compensator.
The voltage sag compensator (DVR) is composed
v0 d + L f ⋅ w ⋅ i fq − ud 2 basically of an energy storage system, a dc-link, a dc–ac
dd 2 = (34) converter and a transformer (Fig. 11).
V
To find out which voltage reference value should be in the
load when voltage sag occurs, it is needed to determine the
v0 q − L f ⋅ w ⋅ i fd − uq 2 voltage value before the sag. The output signals of the PLL
dq2 = (35) are used to inform in a very short time the values of Vd, Vq
V
and q. During the voltage sags, the variations in the angle q
where: originate another synchronous referential rotated in relation
u d2 — D-axis output of the current PI compensator, to the original reference angle q* (Fig. 13).
u q2 — Q-axis output of the current PI compensator. One of the methods for generation of reference voltage is

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System $#
 Vq — Q-axis grid voltage,
VDVR,d — D-axis DVR reference voltage,
VDVR,q — Q-axis DVR reference voltage,
VS — supply voltage vector,
q — supply voltage vector angle,
q* — load reference voltage vector angle.
To determine which the value of reference angle q*, instead
of filtering the angle q that is supplied by the PLL, it should
be filtered the value of w*. The output of this filter passes by
an integer, supplying the reference angle.
The controller can compensate voltage sags effectively
Fig 13. Reference axes system during a voltage sag as shown in Figure 14. Figure 14 shows the phase 1 voltage
for the supply and the load. Phases 2 and 3 also present
effective compensation, keeping the rated voltage in the load.
The prototype of the DVR is composed of a capacitor set
(dc link), an inverter, an output filter and a series transformer.
It is also included a three-phase resistive load. The
experimental tests have been realized with 380V (line-to-line
voltage). The control system consists of a microcomputer
with interfaces dedicated to measure the electrical variables
and to command the converters switches.
To verify the performance of the DVR system, the following
parameters are selected for experimental implementation.
1. Input filter: 1.8mH, 10A
2. Rectifier and inverter: 6 switch + diode (1200V, 50A)
3. Dc-link capacitance: 4.7mF, 900V
4. Output filter: 5mH and 40mF
5. Series transformer: 3 single-phase (220V:220V)
The controller can compensate voltage sags effectively
Fig. 14. Phase a voltages during the simulation
as shown in Figure 15. Figure 15 shows the three-phase
voltages for the system.
by using a low pass filter in the output signals of the PLL.
This filter has a slow dynamic, in such a way that the difference
between the reference signal and the grid signal is the voltage 6. PWM TECHNIQUES
value that should be injected by the DVR to compensate the
sag: Space vector modulation (SVM) is nowadays the PWM
technique most used to control inverters. SVM is based on
(40) the concept of approximating a rotating reference voltage
DVd = Vd* − Vd space vector with those realizable on a three-phase inverter.
An optimal pulse-width-modulation is obtained on a
DVq = Vq* − Vq
voltage second average basis if only four switching states
adjacent to the reference vector are used (SVPWM). In this
From Figure 13:
case each phase is switched in sequence by switching only
one inverter leg at each transition from one state to the next
( )
VDVR ,d = Vd* − VS cos q − q* one. One possibility to reduce the number of switching is to
use the two-phase modulation in which only two phases are
(41)
VDVR ,q = Vq* − VS sin ( q − q )* modulated while the third phase is clamped to the positive
(DPWMMAX) or negative (DPWMMIN) dc rail [11]. Since
clamping implies in no switching losses, this technique
Knowing that Vq* should be zero, (40) is simplified for: reduces losses in each modulation interval.
SVM techniques can also be implemented by using digital
(
VDVR,d = Vd* − VS cos q − q* ) scalar PWM. In this approach, non-sinusoidal modulating
waveforms are introduced in a simple way. In digital scalar
(42) PWM the split and distribution of the zero space vectors
(
VDVR,q = −VS sin q − q* ) duration V0(t01) and V7(t02), inside the sampling interval,
can be represented by the apportioning factor m = t01 / (t01 +
where: t02) [12]. When 0 < m < 1 the modulation is known as
V d* — D-axis load reference voltage, continuous modulation. The case m=0.5 is equivalent to
V q* — Q-axis load reference voltage, SVPWM. When m = 0 (DPWMMAX) or m = 1 (DPWMMIN)
Vd — D-axis grid voltage, the modulation is known as discontinuous modulation.

$$ Power Quality and Utilization, Journal • Vol. XII, No 2, 2006

 Because of the simple implementation, the digital scalar
PWM is used to control the inverters. The switching state
functions in (18, 19, 34 and 35) are transformed to normalized
reference phase voltages.
For digital implementation, time th is equivalent to the zero-
sequence component of the pole voltage. By considering
times ta, tb and tc for the three sinusoidal reference signals,
th can be written as [12]:

th = (1 − m) ⋅ (TS − Tmax ) − mTmin (42)

in which TS is the switching interval, Tmax is the largest time

interval among the three phases, and Tmin is the smallest
time interval among the same three phases. The Tmax and
Tmin can only be defined after ta, tb and tc are determined by:

( )
tk = v*k + 1 / 2 ⋅ TS (43)

To verify the design of the proposed conversion system,

simulations using different loads were realized. Inverter
efficiencies have been calculated from the component models
used with curve fitting techniques [13] for different load
conditions.
DPWMMIN and DPWMMAX present the best
efficiencies among the PWM techniques. It is expected since
that only two phases are modulated at each modulation
interval for those techniques. The MPPT algorithm presents
very good results with at least 99% of the PV maximum output
power. The total harmonic distortion (THD) of the supply’s
currents is kept below 5% for all situations. The sinusoidal
PWM (SPWM) presents the highest THD
Based on characteristics discussed above, the criteria to
be used to compare PWM techniques for the PV conversion
system is the inverter efficiency since the differences among
the MPPT efficiencies are very small when changing the PWM
technique. Also, except the SPWM, the differences among
the THD of the system currents are very small when changing
PWM. Therefore for the proposed design, DPWMMIN or
DPWMMAX should be used to control the switches in the
converters 1 and 2 to improve efficiency.

7. COST ESTIMATION

The proposed system is composed basically of a PV

generation system and a DVR system (Fig. 1). For the
evaluation of cost, the main components are analyzed.

A. PV Generation System
Fig. 15. Three-phase voltages during the experiment
The PV generation system is composed of a PV array, a dc
link and a dc–ac converter. In the dc side it is installed a
capacitor set.
In the configurations used in this paper, the PV array is Three possibilities of parallel connections have been
always composed of 7 panels in series in such a way that the tested to evaluate cost in according to generated PV power.
ideal dc link voltage is around 120V. When a PV generation It has been used PV arrays of 7 panels in series (700W), 14
system is connected to the grid, it is important to have the dc panels (1,400W) and 28 panels (2,800W).
link voltage higher than a single panel (17V) to allow the good The dc link is composed of a capacitor of 300mF for all
performance of the control. The grid has a phase rms voltage situations. Therefore the cost of this part of the system is kept
of 220V, which means that a transformer is used. A turn ratio of constant. Also the cost of the dc capacitor is very low when
10:1 has been used to connect the system to grid. compared to the PV array (Table 1). All prices are in US$.

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System $%
Table 1. Cost of the PV System without Voltage Sag Compensation. 11.4%. Increasing the PV array power, the cost of the extra
components becomes less significant for the system. For the
Components
PV array power PV array of 1,400W, the additional cost is 6.5% while for the
PV array of 2,800W the additional cost is only 1.1%.
700W 1400W 2800W
PV array 3,594 7,188 14,376
8. CONCLUSION
Dc capacitor 22 22 22

PV converter 82 132 322

The proposed design introduced in this paper improves
PV transformer 238 287 369 functionality in grid connected photovoltaic generation
systems. The design was used to do a comparative study of
PWM techniques for this specific situation. The system can
Table 2. Cost of the PV System with Voltage Sag Compensation.
be connected to a three-phase system of any electric utility if
a matching transformer is used. The excellent performance of
PV array power
Components the system is verified from simulated results using Matlab.
700W 1400W 2800W The voltage waveform in the photovoltaic array follows
the reference voltage for all irradiation conditions. Besides
PV array 3,594 7,188 14,376
that, the controller also compensates harmonic and reactive
Dc capacitor 22 22 22 power. Using the design based on simulation results, it is
PV converter 132 210 396 possible to make a comparative study of different possibilities
of control.
PV transformer 271 336 450
The good performance of the DVR system is verified from
DVR converter 82 82 82 simulated and experimental results. The voltage waveform in
DVR transform. 222 222 222 the load follows the reference voltage keeping the load
voltage in the rated value. Cost estimation also has been
DVR filter 61 61 61 realized. It is shown that the additional components are not
significant for the total cost of the system, especially if the
PV array power is high.
The PV converter is composed of six power switches with
their gate drive circuits and heat sink. MOSFETs of 200V ACKNOWLEDGMENT
have been used. The price of transformers with turn ratio
10:1 is also considered in the total cost. The power ratings of
The authors thank Conselho Nacional de
the transformers are 600, 1,500 and 3,000VA for the PV array
Desenvolvimento Científico e Tecnológico (CNPq) for its
power of 700, 1,400 and 2,800W, respectively.
financial support.
B. DVR System
The DVR system is composed of a dc-ac converter, a filter REFERENCES
and a series transformer.
The DVR converter is composed of MOSFETs of 200V, 1. K u o Y . C , L i a n g T . J . , a n d C h e n J . F : Novel
5A. The currents even during the voltage sag are much smaller maximum power-point-tracking controller for photovoltaic
energy conversion system. IEEE Trans. on Industrial Electronics,
than the currents of the PV converter. 2001, 48, 3, pp. 594–601.
The filter is composed of a 50mF capacitor and a 5mF 2. L e s l i e L . G . , J r . : Design and analysis of a grid connected
photovoltaic generation system with active filtering function,
inductor for all situations. The LC filter has a cut frequency Master Thesis, Virginia Polytechnic Institute and State University,
of 318Hz that is approximately 5 times the grid frequency. Blacksburg, Virginia – USA, 2003.
3. A k a g i H . , K a n a z a w a Y . a n d N a b a e A :
Transformers with turn ratio 10:1 are also considered in the Instantaneous reactive power compensator comprising switching
total cost. The power ratings of the series transformers are devices without energy storage components. IEEE Trans. on
600VA for all conditions since the currents have small variation Industry Applications, 1984, 20, 3, pp. 625–630.
4. A r e d e s M . a n d W a t a n a b e E . H . : New control
when the PV array power changes (Table 2). algorithms for series and shunt three-phase four-wire active power
It is worth noting that in the proposed system there are filters. IEEE Trans. on Power Delivery, 1995, 10, 3, pp. 1649–
1656.
additional components, increasing the cost of the global 5. S i n g h B . , A l - H a d d a d K . , a n d C h a n d r a A . : A
system. Also the PV converter and the PV transformer must review of active filters for power quality improvement. IEEE Trans.
have higher ratings to support the higher values of currents on Industrial Electronics, 1999, 46, 5, pp. 960–971.
6. G y u g y i L . , S c h a u d e r C . D . , W i l l i a m s S . L . ,
during the voltage sag. However, due to the high cost of the R i e t m a n T. R . , To rg e r s o n D . R . a n d E d r i s A . :
PV array when compared with the other components used in The unified power flow controller: a new approach to power
transmission control. IEEE Trans. on Power Delivery, 1995, 10,
the system, the additional cost can be justified. 2, pp. 1085–1093.
The critical case for the cost of the additional system occurs 7. M o r a n S . : A line voltage regulator/conditioner for harmonic
for the PV array of 700W. In this case, the PV system has a sensitive load isolation. IEEE Industry Applications Society
Conference, 1989, pp. 947–951.
total cost of approximately US$ 3,936 without voltage sag 8. M e n d a l e k N . a n d A l - H a d d a d K . : Modeling and
compensation. With voltage sag compensation, the cost is nonlinear control of shunt active power filter in the synchronous
reference frame. IEEE Harmonics and Quality of Power
approximately US$ 4,384 that means an additional cost of International Conference, 2000, pp. 30–35.

$& Power Quality and Utilization, Journal • Vol. XII, No 2, 2006

 9. H u a C . , L i n J . , a n d S h e n C . : Implementation of a Marcelo Cabral Cavalcanti
DSP-controlled photovoltaic system with peak power tracking. was born in Recife, Brazil, in 1972. He received the
IEEE Trans. on Industrial Electronics, 1998, 45, 1, pp. 99–107. B.S. degree in electrical engineering in 1997 from the
10. C h a n g C . S . , H o Y. S . a n d L o h P. C . : Voltage quality
Federal University of Pernambuco, Recife, Brazil, and
enhancement with power electronics based devices. IEEE Power
Engineering Society Winter Meeting, 2000, 4, pp. 2937–2942. the M.S. and Ph.D. degrees in electrical engineering
11. H a v a M . , K e r k m a n R . J . , a n d L i p o T . A . : Simple from the Federal University of Campina Grande,
analytical and graphical tools for carrier based pwm methods. Campina Grande, Brazil, in 1999 and 2003,
IEEE Power Electronics Specialists Conference, 1997, pp.1462– respectively. Since 2003, he has been at the Electrical
1471. Engineering and Power Systems Department, Federal
12. J a c o b i n a B . , L i m a A . M . N . , d a S i l v a E . R . C . ,
A l v e s R . N . C . , a n d S e i x a s P. F . : Digital scalar pulse University of Pernambuco, where he is currently a Professor of
width modulation: a simple approach to introduce non-sinusoidal Electrical Engineering. His research interests are power electronics,
modulating waveforms. European Conference on Power renewable energy systems and power quality.
Electronics and Applications, 1997, pp. 100–105.
13. C a v a l c a n t i M.C., da Silva E.R.C., Gustavo Medeiros de Souza Azevedo
B o r o y e v i c h D . , D o n g W. , a n d J a c o b i n a C . B . : was born in Belo Jardim, Brazil, in 1981. He received
Comparative evaluation of losses in soft and hard-switched the B.S. degree in electrical engineering in 2006 from
inverters. IEEE Industry Applications Society Conference, 2003,
pp. 1912–1917. the Federal University of Pernambuco, Recife, Brazil,
where he is currently working toward the M.S. degree.
His research interests are power electronics and
renewable energy systems.

Bruno Aguiar Amaral

was born in Recife, Brazil, in 1981. He received the
B. S. degree in electrical engineering in 2005 from
the Federal University of Pernambuco, Recife, Brazil.
His research interests are power electronics and
renewable energy systems.

Francisco de Assis dos Santos Neves

was born in Campina Grande, Brazil, in 1963. He
received the B.S. and M.S. degrees in electrical
engineering from the Federal University of
Pernambuco, Recife, Brazil, in 1984 and 1992,
respectively and the Ph.D. degree in electrical
engineering in 1999 from the Federal University of
Minas Gerais, Belo Horizonte, Brazil. Since 1993, he
has been at the Electrical Engineering and Power
Systems Department, Federal University of Pernambuco, where he is
currently a Professor of Electrical Engineering. His research interests
are power electronics, renewable energy systems and power quality.

Marcelo C. Cavalcanti et al.: Unified Power Quality Conditioner in a Grid Connected Photovoltaic System $'